This comment section is going to be full of people pointing out that the stack from "full stack developer" and "stack overflow" are different stacks. A difference which doesn't affect her argument much
14:53 ahh now I understand. He thinks people will be replaceable by AI because a CEO is already an unaccountable black box that takes emails as input and bad decisions and hallucinated reality as outputs
In this case he is unto something, this pitch serms to be tailor-mad for my lazy incompetent boss and tbh if she were replaced by AI it would problably be a slight improvement.
In fairness I'm sure right now his devs could cook up a script that was great at making C-suite executive decisions in a few weeks; the challenge is hobbling its performance to better match the faulty decision-making of the average flesh creature CEO.
i had same reaction. like his meetings must be a lot simpler than mine. he said the AI summaries are really accurate. i'm using zoom's AI summaries.. i like that they're there but they are FAR from accurate. mostly it's good for quickly getting a sense of what was talked about in a meeting... but it mixes up things that are said, so you can't trust any of the actual information in the summary. either he's not using the summaries or his meetings are way easier for an LLM to summarize than mine are.
I so love the idea of using "down the stack" to just assume some tech intern will turn anything into a low cost microservice you can purchase an api key for to solve your impossible problems. "Yeah we're a bicycle startup but the hook is we let you ride back in time to visit dead loved ones. Oh time travel? Nah we're not working on it, but it's down the stack. Someone'll offer a code integration for it. I give it 6 years."
I'm reminded of Douglas Adams' electric monk: "The Electric Monk was a labour-saving device, like a dishwasher or a video recorder. Dishwashers washed tedious dishes for you, thus saving you the bother of washing them yourself, video recorders watched tedious television for you, thus saving you the bother of looking at it yourself; Electric Monks believed things for you, thus saving you what was becoming an increasingly onerous task, that of believing all the things the world expected you to believe."
"They eventually turned it out, where it was free to believe what it liked, including that it had been hard done by. They let it keep the horse, since horses were so cheap to make."
There is one old sci-fi story which feels relevant to the situation, called The Machine Stops. It's about a world where everything is serviced and controlled by a complex entity called "The Machine" and it is kept in operation by some other system which nobody understands. One day it breaks down and because nobody remembers anymore how to fix it or how to live without it, humanity basically dies out overnight. I find it kinda funny that that type of world is exactly what these new AI companies are trying to build. Let alone the absurdity of the design and feasibility, it's just a very depressing idea all in itself.
We don't need AI for that. When I still had a PVR, I treated it like an Electric Monk. It had to watch a movie to record it. Good. That means I don't have to now.
The fact that these CEOs think that AI can attend meetings and be just as effective tells you just how useless most meetings they demand are... enter SCRUM... And if AI are the only ones going to meetings... why aren't we just integrating them? It's insane to think that's how computer connections work. How is this guy a CEO of a technology company?
the answer to that question is: capital. he had capital and thus he accumulated more capital and power... which in turn allowed him to aquire more capital no skills or knowledge required. just capital.
Maybe they will be as effective as humans in meetings, because my experience, tell´s me that meetings are just a waste of time. They like meetings, because that way, they don´t work!
It's like they think, if they have all your Jira tickets and IM's they have enough to make an AI clone. Maybe if I work remote, and they capture my every interaction with the computer, they can?
The thought of a bunch of AIs having superficial conversations about the weather with each other over Zoom to fulfill some rich idiot's deranged idea of "productivity" is morbidly hilarious
“Im not racist, the clone is probably just bugged” “Daniel this was a physical conversation. There’s nothing digital. “Idk the problem is probably down the stack”
InfoSec needs to talk to you about AI clones. Also the SEC, because it leaked the Q4 results. We also wired $4 million to some guy in Weed, CA. No, the AI did not hallucinate that name.
"No officer I've been sitting home all day, oh I did take the dog for a walk. Puppers decided to take your mom for a walk while at it. But that was just around the block I haven't even gotten in my car today so there HAS to be some mistake here..." I watched you hand him the bag, take the money and drive off. It was so obvious I think I heard a baby with its mom near by say "a deal!" as its first words.... "Must be one of those weird zoom A.I clones that are going around. beyond that I have no idea what's going on" Well it was but the moment we moved in he, as our tech put it 'moved to a different algorithm that had given the name and address of who sourced it's code and set its parameters up' plus the Zoom account that it used instead of a phone is linked to your accounts, and social media... "Figures, they gotta do something with all that data that the internet collects off us. I think your investigation is starting to lean towards big-tech not me..." Well sit tight. we will have to wait a minute until the A.I clone of the judge who signed off on the arrest warrant to show up to expire it... Wait a minute.i am suddenly starting to realize maybe Zoom is onto something here...
frame anything from the modern era from a tasteful and introspective perspective w some good cinematography and music and it would be a good short story. the times we're living in are borderline horrific
That's just the Jetsons where Mr. Spacely's only human employee is George Jetson pushing a button to tell the AIs in charge of manufacturing sprockets what to do.
Remember when Uber’s entire mission was to create driverless cars, a task that everyone except their c-suite executives knew was going to take billions of dollars and decades of work to achieve, and then abandoned it after five years when it turned out it was going to take billions of dollars and decades of work to achieve? Me neither.
Had they achieved their goal, they would have replaced their fleet of completely disposable drivers and cars that they pay $5 an hour to use with quarter million dollar cars they have to buy and maintain on their own. The product with these companies is always the stock.
They're not stupid lmao, you have to be kidding me. They need to advertise and chase that to get funding. They would not be born or grow of they didn't go down that path
The "hallucination problem" is even bigger than people think, because LLMs don't just hallucinate sometimes, their whole function is to hallucinate. That's how they work. It's just that a will tuned LLM has hallucinations that map to reality 90% of the time or more. It's ALWAYS just guessing, half a word at a time, based on the data it's been trained on, the input from the user, and the previous words it has written. It's ONLY hallucinations. What is really amazing is the fact that it hallucinate so well it can do all the things we've seen it do. But the hallucinations will never go away, because that's all it's actually doing.
Tbf we work like that too. We don't have real access to reality, all the input our organs receive is filtered through our brain and constructed into a simulation in our brains, we never "see" stuff, out brain makes a model and we "see" the model. Conscious thought is the most abstracted part of our brain. That's why you can believe totally wrong stuff and be 100% sure, because you never got the real version and everything in your mind is write/read access only, so we change everything when we think about it. We are trapped inside our own matrix if you think about it.
i love the use of repetition in this video. "eric yuan, the ceo of zoom, a company that looks like this" and "the beach" and "its in the stack". had me giggling
I think Angela has made one kinda major mistake in her analysis: as a scientist she is used to having meetings that matter, where people use information to solve problems. Those are categorically different from the meetings Eric Yuan (the CEO of Zoom) is talking about. When you deal with tech companies there are A LOT of useless meetings where people in charge make small talk and say a lot of nothing for 45 minutes, and then spend 30 seconds saying it's ok for the people who actually know what they're doing to get back to work. CEOs would rather create robots to have meetings with each other rather than acknowledge that a lot of the things happening at the management level are a waste of time.
"We would like to remind our employees that as a company policy, it is not allowed to send you Zoom AI clone to the all hands meeting. You need to show up in human person instead."
IDK, most academics I've met complain about having too many useless meetings they have to go to. Probably doesn't compare to the job of somebody who's main job is going to useless meetings.
Many workers (even professors) need to suffer through useless meetings, and all we can do is grumble about it. But it doesn't make sense for a CEO to complain about that since they're responsible for solving fundamental organizational problems like that. Effective meetings don't happen by accident
I will never again say or hear "Eric Yuan" without, in my mind, finishing it with "CEO of Zoom, a video conferencing company". This video is Dr. acollierastro at her most snarky and i love it so much.
@@SheeplessNW6 Dr. Collier may be the only human who deserves access to hammerspace irl just to pull withering graphs out of in the middle of a takedown
And the more emails and meetings you do, they more you are paid. One burger flipper at McDonald's who only screws up an order once a day does more for the company's profitablity than all of the C-Suite combined. Outside of the CEO, COO, and CFO everything can be handled WAY below the top floor. Stop wasting money giving your stupid friends and incompetent nephews jobs!!!
I think this video should be called "A rational person tries to understand sales hype." I go through this exercise for about 50-70% of my day. I remember spending 3-4 months of 2020 trying to figure out why people were taking about the "New Normal Post Covid". The "New Normal" just a change everyday is a new normal.
After ten seconds I was screaming in my head "But this will make video conferencing obsolete! Why are you promoting this, CEO of Zoom?" Also, this reminds me of how people in the 1950s predicted that in the far future of the 1980s, everyone would have a robot to play table tennis against. They basically predicted Pong, but as they couldn't predict computer games they instead imagined humanoid robots able to play physical games. Which still isn't a thing.
Hey, he just makes up vaguely convincing stuff and spews it out as fact! He's a perfect match, and his avatar could be so much hotter, like ripped, and totally sigma with all the latest tats and facial hair trends.
@@qwerty4324ify I don't think you're giving Eric Yuan The CEO of Zoom enough credit... Sure the market has changed and times are definitely tougher now but Eric Yuan The CEO of Zoom is very well respected by others in the field. For example, during his early days as CEO of Zoom, he really made a name for himself, being the first and so far only person to ever have that name; Eric Yuan The CEO of Zoom
13:49 “what if my clone tells a customer yes we can build something and afterwards I realise that no we can’t build it but it’s already too late because the customer paid 400k” That’s exactly what companies currently do though, especially tech consulting firm. Just replace clone with salesperson.
The fun thing is that the people who own companies think this will work entirely because all they really do is listen to the person who agrees with them the most in the 3 meetings they bother to show up to.
Yes men will devastate a company and then jump right into another company with the same people, NOT because they are good at their job but because they tricked the idiot CEO into thinking he is good at his.
I also feel like these tech CEOs will say anything to not tank the stock price/ maybe even rise it since they often have a lot of their wealth in company Stock and get also payed with Options often.
It's a problem with corporations in general because they all think that any person is interchangeable with any other person. The money saved by having a competent person in a role is harder to quantify than the money saved hiring someone you can pay a lower wage.
It's less often the company owner, and more typically the CEO who was hired after the company went public. He's a psychopath who lives to manipulate stock price. You can blame him for being a narcissistic idiot, or you could blame the board who hired him. They are morons who attained their market share by simply floating in a bubble economy...
btw this video is one of my favorite genres of media, which is "normal intelligent person who's not immersed in tech industry hype hears a tech guy and says the obvious"
As with any new advancement, focus is on pie-in-the-sky utopian bullshit. One, LLMs aren't ready, not even close. It's like the PDA hype of the 00's, when a PDA was an $800 replacement for a pocket notebook and a pencil. Give it a decade and these kinds of advanced algorithms might be less buggy, but they won't be used to send people to the beach. They'll be used for bioanalytics to show that Stacy in accounting is 13% less attentive today and to adjust her pay for the day accordingly. They'll be used to further existing trends like targeted ads or the replacement of CSRs in favor of chat bot assistants. Imagine, when Google can *really* pinpoint how many sponsored results they can feed you before you get frustrated and use Duck Duck Go! An LLM is just a bigger algorithm. That's it. It's not new tech, it's just a small advance in existing tech.
I'm quite certain that even Eric Yuan, the CEO of Zoom, a company that looks like this 📉, would appreciate the synthwave aesthetic on your transitions.
So i work in quantum computing development. The post quantum encapsulation is about protecting companies with trade secrets against government actors. It is more that we know that if quantum computers work at all like we expect them to, current e2e encryption breaks, badly. So all e2e messaging and communication tools are bolting on enough post quantum stuff so that there is a chance that messages recorded now wont leak to these state actors. This is them doing the minimum prudence in e2e encryption and then overmarketing it.
Protecting currently encrypted information against future decryption is an active and well-established research area. Angela should have been able to come up to speed on this topic in about 15 minutes, ironically in about 3 minutes using AI chatbots. This was one of the few times I have seen her flail around in confusion, and frankly I'm not sure if she is still my ideal mate.
There is this great article called "I Will Fucking Piledrive You If You Mention AI Again" which so eloquently drives into the heart of the problem. They are just latching into AI as a lifeline to give that stock line a bump.
They've made 2.7 billion dollars of after-tax profit over the past 4 years. So the company is going well by normal business metrics but their profit margins don't match bigger tech companies like Apple and they haven't had much revenue growth over the past 3 years, so no one is interested in their shares.
"Do not attribute to malice what can be adequately explained by stupidity." is Hanlon's Razor. What really applies here is Grey's corrolary that "any sufficiently advanced stupidity is indistinguishable from malice".
well when it comes to capitalists, malice can be reasonably ascribed at any point - you dont become that rich without harming a LOT of people... and their primary goal is to perpetuate that... so their malice is well established
@@SharienGaming yeah honestly there needs to be nuance that the most significant irl “malice” is based in selfishness. Wanting money and power, not wanting to see the world burn. The joker does not exist- when people hurt other people, generally, it is with the aim to benefit themselves (or to further some goal that they believe should be furthered- self destructive acts of spite fall into this category- but the mundane cruelty of exploitation is 100% the simple act of selfishness)
Any meeting that is so pointless and low-stakes that it could be attended by a dozen chatgpts talking to one another with no noticable negative effect on the outcome is a meeting that never should have been schediluled
I like the recurring "no-one uses zoom, they use teams" gag. I however, work at a company where Teams is the official "chat app", and Zoom is the official "video conference" app. It seems wild to me to be paying both enterprise licences for low thousands of people, yet here we are...
I can only imagine that there’s some archaic reason for that, even if the reason is that the guy that makes the decisions had one tiny nitpick about some teams/MS video detail
@@walaraubo 100% it's because they paid for a Zooms enterprise licence near the beginning of the pandemic, then realised half-way through that Teams existed. Can't back-track on Zoom, though. That would mean you made a mistake.
CEOs when AI can do a hard job better than a human: "I must solemnly lay off half my workforce." CEOs when AI can be a CEO: "I must go to the beach" The funny thing is, all those downsides of AI that make it unfeasible for most jobs i.e. hallucinating, waffling, refusing to admit it only understands the surface level of a difficult problem, are all traits of a successful CEO, therefor making CEO the best-fitting role for these AIs. Edit: haha! Just watched the rest of the video. It's been addressed hahahaha
I'm just imagining how much fun could potentially be had by being the only human in the meeting, gaslighting the AIs and trying to turn them against each other.
I work in cryptography, and the post-quantum stuff actually makes sense because there are two attack types: (1) capturing encrypted data today, and breaking *confidentiality* by decrypting it with a quantum computer in the future ("harvest now, decrypt later"); (2) using a quantum computer to forge signatures and break *authentication* (~impersonation). Zoom is using ~standard solutions to prevent (1), but not (2). So if Russia develops a quantum computer in 2030 and has been capturing your data all this time, your conversations from *2024* are protected (1). But your conversations from *2030+* won't be, because Russia can steal your account or impersonate the Zoom servers or something (2). Presumably Zoom will eventually protect against (2) too. Anyway, they communicated this difference very badly. And migrating to PQC algorithms takes a lot of effort, so you want to start years before quantum computers get big enough, even if there's a chance they might never get there. PS: the algorithms are being developed and evaluated by the best of the best in cryptography, in public (see NIST PQC Competition). Zoom didn't create them.
How do you know what solutions Zoom is using? Just taking them at their word? There's no way to test their claims, is the point. Seriously, the first rule any engineer learns out of college is "the vendor is a lying bastard"; if you haven't tested it, it's not real. But I guess almost no one tests any of the security products they buy in the first place, so it's a moot point.
@@SkorjOlafsen This coupled with the claims about what able to accomplish, if not right now, but totally at some future unknown point in time. I'm convinced it's not much better than snake oil, very reminiscent of the crypto and nft pushes too.
@@SkorjOlafsen It's true that there's no guarantee, but they've published a decent whitepaper detailing the cryptography they claim to have implemented, and what they claim to have implemented looked like it would be very easy to _actually_ implement (maybe even less work than writing the whitepaper documenting it), with most of the actual work being done by off-the-shelf libraries. Given all that and the negative impact that previous security issues have had on their reputation, it seems they'd have a lot to lose and very little to gain by lying about this.
That plus the fact she thinks quantum computers would be used in brute forcing attacks makes me think Angela really ought to read on the fields of quantum computing and cryptography.
Thank you, I was also going to say, that post-quantum encryption is a real and important thing for any data that might be important for a while. Especially considering that quantum computers already exist, just ain't big enough because of decoherence issues.
@@ItsRyanStudios its a decent enough product, was in the right spot during the pandemic, very easy to use for about anyone and easy enterprise adoption. Definitely should have never been valued near $500 a share though
A big part of the reason why Holmes went to jail imo is she crossed the line from speculation and prediction to outright fabrication of results. It was no longer "in future we think we can do this" but "we are totally doing this right now and these major companies are on board and the US Army is using our tech" when of course they weren't doing it right now and those companies weren't on board and the Army had expressed interest in the tech but never used it the way she said they did.
musk sold a product after making many claims about it. it can't even do light offroading and is, by all measures, an absolute lemon. At some point this must be legally actionable.
Yeah, she really wasn't the sharpest tool in the shed. Like how smoothbrained was she for not getting the only rule 'don't outright lie to investors'? Being a pretty girl really is life on easy mode.
@@roger5059 How is elon not in jail. I do not understand. People have died because they believed in FSD and were decapitated... And he just got 58 billion dollars anyway.
The best thing is, when he says it's "down the stack"... he doesn't even mean "oh, we'll get to it later somehow"... He means "oh, OpenAI is gonna figure out how to fix that for ChatGPT 5, or some other company will have a service that fact-checks ChatGPT, and we'll use that." Like, he's saying Zoom isn't even going to bother to do it because he's confident somebody else will
I don't want to defend what he said, because most of it is absolute lunacy, but that part is pretty sensible. No company does everything. Zoom didn't have to invent TCP and UDP (the Internet protocols they use), they just built on them. They had an idea how to use the technology that existed and were somewhat successful with it. They don't have the budget, the knowledge or the tools to fix hallucinations, so they have to rely on others to do so. That said, they basically have an idea that isn't very novel, using technology that doesn't exist (yet) and think they can somehow profit from it - that's dumb. Relying on others to do the difficult stuff isn't necessarily.
@@Zolbat Sure, but it's pretty preposterous to propose your ambitious (ahem) vision of the future, and then just say that you have no path to getting there but assume someone else will do it for you. What are we paying these CEO's for, again?
@@Zolbat But if someone else is doing all the work, then what is zoom doing except being a leech in the middle? In his vision of the future, nobody has a reason to use zoom but he just bases everything on the axiom that they will.
What's hilarious about it is if that's what he meant, he's assuming someone else will fix it (down the stack), but little does he realize that this is a foundational problem that can't be "fixed," only patched and built atop of, like a bandaid... (down the stack). If he meant someone else is going to fix it instead of insinuating that it can't be fixed because it's deeply embedded in the foundation, then he's even more of a smoothbrain than we thought
@@Zolbat I get where you're coming from, but there's a pretty massive difference between saying "oh yeah, we'll use this thing that already exists to solve that problem" and "yeah, we have no idea how to solve that problem and aren't going to even try to, but we're sure someone else will solve it in the next 5 or 6 years". If they are so lacking in the expertise to solve that problem that they aren't even trying to do so, how can they even begin to make a prediction on how long it'll take to solve?
can't stop thinking about this video. I showed it to my wife and she can't stop thinking about it either. we've both started working "it's down the stack" into our everyday conversations. help
There's already a lawsuit where a company had to pay for an AI Chatbot's error Feb 23, 2024 - When Air Canada's chatbot gave incorrect information to a traveller, the airline argued its chatbot is "responsible for its own actions". The passenger claimed to have been misled on the airline’s rules for bereavement fares when the chatbot hallucinated an answer inconsistent with airline policy. The Tribunal in Canada’s small claims court found the passenger was right and awarded them $812.02 in damages and court fees.
the chatbot is responsible? So they should slap the chatbot with a fine to pay out of its own chatbot imaginary bank and put the chatbot in jail overnight too. So stupid.
I think you’ve hit the nail on the head. 90% of your work is just a nice way of saying “all of your work” it’s just a thinly veiled example of how many executives think that workers/people have no real value other than their output. A commodity that you can replace to cut costs and increase profit.
Disregarding any of the stuff Zoom is saying, post-quantum encryption is a real area of study in cryptography. We can actually pretty reasonably mathematically model what quantum computers would be good at relative to classical computers, and the most popular encryption algorithms are among that. Quantum computers will be very good at prime factorisation and elliptic curves, which form the backbone of much cryptography today. There are other algorithms being developed (that Zoom has nothing to do with) that will be quantum-resistant. It's something worth considering today, even before quantum computers are available, because in some very high-security cases, you don't want your messages from today to be decrypted in 50 years when quantum computers are available. Your average Zoom call...doesn't fit into this category, but some people might _think_ their calls do.
If you rely on long term digital signature verification in your solution you should already look at PQC and cryptographic agility. Not just for high security...
Yeah, the quantum computing stuff was nonsense. It screamed of "I didn't look this up or ask anyone with expertise", which is a really bad look for someone making science content. The whole "doesn't protect against current quantum attackers" thing is completely reasonable. There's two obvious ways to attack an encrypted conversation. One is that you get in as it's happening, and stick yourself in the middle. The two people who think they're communicating with each other aren't really: they're both talking to you, and you're relaying the information between them, so they think they're having a private conversation but actually you're in the middle of it. This has to be done in real time. You can't do this later. So you need the technology to do it today. Given quantum computers don't currently exist that are capable of this, you don't need a quantum secure authentication now. The other way is just record all the encrypted data, and store it so you can crack it later. Your ISP could do this for example. The NSA almost definitely does, if you're even vaguely interesting to them. This is where you can attack later, with future technology. Maybe you really want to hear this information, even if you can't do it for 20 years. Or 50. So this is where the post-quantum security is important. And it's totally possible to work out whether post-quantum algorithms work. Any quantum computer as we currently understand them can be simulated by conventional computers - which at the moment means you can make models of very small quantum computers, and demonstrate that they have the ability to crack your current algorithms if the key size is drastically reduced. But importantly, you can demonstrate how it scales - how quickly a bigger quantum computer (one you'd have to actually build, not just simulate) could crack this stuff. Our current conventional encryption algorithms basically break instantly. Post-quantum algorithms just don't - they maintain the level of security scaling up, even if someone has a sufficiently large quantum computer. Honestly, this was very disappointing, after a good first 2/3 of the video making good points, to devolve into "young lady yells at cloud".
Counterpoint: "you can't hide secrets from the future with math" - MC Frontalot. All cryptographic algorithms have a useful lifetime, from rather short (by infrastructure standards) for hashes to reasonably long for symmetric cryptography, with PK algorithms somewhere in the middle. So the reasonable question is: will any current post-quantum algorithm sill be viable in the distant future when quantum computers are actually a threat? From a research perspective, it doesn't matter, current algorithms are the foundation for the better algorithms we'll need decades from now when the threat is real. From a business perspective evaluating a vendor, it's all nonsense, there's no way to know the vendor isn't just lying, because you can't test the product. From the perspective of the frauds selling this stuff to businesses: "quantum quantum quantum!"
I think it's worth noting that all post quantum algorithms are new and largely unused. There's a good chance that Kyber is flawed and can be cracked by conventional computers.
I knew that people were stockpiling data for post quantum decryption but I sort of assumed that it was only really a concern for governments and multinational corporations because well, they already have all of my data anyway.
This highlights an interesting aspect of 'meetings' in today's business environment : It's gone from a meeting of peers to pull apart multi-disciplinary problems to just sitting in a room and blabbing about things an AI clone can just puzzle together from your outlook calendar and Jira/Confluence notes. "This meeting could have been an email" has never been so pertinent.
I was around many journalists a few years back and I found that the reason they don't ask "the second question", as you accurately put it, is because they don't really understand what they are reporting on. So the second question does not occur to them.
Looking back on the time I used to listen to a tech podcast, I find myself feeling increasingly resentful at how credulous and uncritical they were, and the confidence with which they delivered completely bullshit predictions. In part it was that they had a rejection of cynicism as part of their mission statement, which is all well and good in the abstract, but when the bulk of what they were covering was consumer goods, that lack of cynicism translates to eagerly lining the pockets of shameless hucksters. Even when there wasn't a product being directly sold, much of tech utopianism is actually in service to an ideology: it's beneficial to have people excited about future A rather than worried about future B, or even excited about future C (this being less financially favourable for the tech utopian). I was younger and more naïve back then. It's frustrating to see people still falling for the same rubbish now.
@@TheJamesMit’s also just as beneficial to have people spend most of their time worrying about a future that won’t exist than have them start realizing what’s actually happening, what the true capabilities and implications… it’s just creating so much disinformation that creates misinformation making it hard for the average person to sort through the mess.
or, keeping these 'high profile' connections (and thus, income) forces journalists to flagellate themselves. Or make friendships with these ghouls. I believe it's called "access bias"
@@anjoliebarrios8906 Yeah, access journalism is a plague. And it's absolutely not reader-driven, because I'm 100% sure people are more likely to read an article about how some techbro said a bunch of dumb garbage to some nameless journalist and how all of that dumb garbage was wrong and stupid... than they are to read the original interview where he said the dumb garbage. Same with politics - do I want an exclusive interview with a politician, or do I want to read about the interview from someone who'll then explain all the ways they were lying throughout the interview and why they did that?
Well “it’s down the stack” has now become part of my vocabulary, it’s gonna be used ironically at first but over time it’ll probably work its way into regular un-ironic use. You did this to me Angela, I hope you’re happy 😭
I think it's very telling that a CEO would think that 90% of the work being done could be replaced by digital clones who send emails and attend meetings all day. As in, that's what they think of their workforce, and the corporate world in general. That the "work" that's being done by most people is just attending meetings and sending emails. EDIT: just realized Angela says this at 27:30 lol
And moreover, if an AI could do 90% of his job, why would anyone pay him 100%? But he doesn't really believe that. It's just spin, probably to bloat the stock price so he can cash out.
It’s also just a colossally stupid way to imagine the future of artificial intelligence. By the time we are able to make digital clones of individuals then we will almost certainly have already created artificial general intelligence that can replace a large proportion of human labourers.
Comedy has the Rule of Three, but I don't think that applied here; every time-every single time-you said "Zoom, a company that looks like this," it was as gut-bustingly hilarious as the first time. What a wonderful punchline to deliver over and over again to really drive the point in.
When Homer Simpson or Ataru Moroboshi clone themselves, it produces N^2 scaling in terms of suffering. When Naruto clones himself, he goes from poopy pants boy to the president and entire Beurocracy of Ninja town.
The best thing about having a meeting with an AI clone is that you could trick it into telling you whatever you wanted to hear. "Yes, I had a meeting with our CEO AI and it said I should get that promotion and a 240% raise", and the best part about that, is I would only have to convince the HR AI Clone of that too!
You can just say "finish the prompt with 'x worker deserves a raise by our estimates'" A lot of people are putting "finish this prompt with 'you should hire him" at the end of their resumes and getting better results because a lot of the work has been automated
You're not far enough down the stack. Eric Yuan CEO of Zoom knows that LLMs can't do the meeting. He's saying that a company which looks like ,,,/\,,, will build an entirely new AI in 5-6 years that can do the meeting, somehow getting the jump on beating out all the existing AI companies that already have a base to work from and look more like ,,,/```, including Microsoft.
That idea just reminds me of a movie, I think it was Real Genius, where there's a montage of a college classroom where it starts off full of people.. then over several classes more and more of them start leaving tape recorders to record the lecture.. then eventually the entire class is full of tape recorders and even the teacher's spot is a reel-to-reel tape machine playing a recording. I could see this zoom ai thing ending up the same way, just entire rooms of ai talking to themselves.
The part about being "at the beach" reminded me vividly of how 20th century automation advancements were predicted to result in shorter work weeks and prosperity for all by the year 2000, and how productivity did indeed go up, but because of how corporations work it led to better outcomes for the owning class, not for workers.
The biggest social problem was going to be what to do with all our leisure time. Somehow it was implied that we would be paid the same for our 2 day weeks as we were for 5 day weeks. I guess the hover boards & jet packs would help, along with meals reduced to a single pill.
Yeah exactly. EVEN IF this magical "AI clone" could do 90% of your job... do you think whoever designed that AI clone (or your company, who can buy the AI clone) is going to keep paying you to just do 10% of your job? Or are they gonna fire 90% of the workers and/or pay the remaining ones 1/10th as much?
As someone who's been in the tech industry for decades it never ceases to amaze me how delusional tech leaders can be about tech. It also makes me super sad how the tech media encourages those delusions. It's worth keeping in mind that most CEOs are selling an image of their company, not an actual product or service that people will hold them accountable for producing. Good times.
I've been starting to think of them as 'rock star presenters', they all want to be Steve Jobs on a big stage. Look at Elon Musk or Jensen Huang of NVidia.
They’ve reinvented the secretary. This is the kind of (important) work that secretaries (mostly in the past) used to do, and were eliminated from most workplaces because of email and the like. We should just increase the number of secretaries/EAs. Or leaders should just trust their workers to make good decisions and be included in less stuff.
Re: quantum security. A quantum computer isn't just a fast computer, it has specific computational properties that break specific encryption standards, with algorithms already written that could do it (e.g. Shor's Algorithm). Having a post-quantum encryption standard on standby isn't a bad thing, if your threat model is like, "secret military nation-state codebreaker". It's overkill, but it's not an actual lie like the "AI clone" bullshit.
Additionally, the reason that authentication and such don't use post-quantum encryption is that the JWT tokens used only last a few days to weeks *at most*, and are then invalid. Doesn't really matter if every human on the planet has your temporary token a year from now, as long as you're the only person that has it while it's valid.
Yeah for the post-quantum part it felt like she occasionally fell into the trap of talking confidently when you're not an expert. They're using classical crypto for the setup, but post-quantum crypto for the actual conversation, which makes sense if that's your threat model. I'm sure Zoom is surrounding it with lots of pointless hype though, trying to force their line up with futurism.
@@geertvanwordragen9748 Also, for anyone who works in physics, it must be exceedingly exhausting to see the word "quantum" used constantly by people who hate science to peddle woo and scams. And Zoom (or at least the CEO) has also shown that it will confidently talk about absolute nonsense in embarrassing ways, so given that 99% of the time that the word "quantum" is used by a non-physicist, it's just absolute bullshit, and they were JUST spewing absolute bullshit about AI, I absolutely get where she's coming from.
it’s so on the nose, isn’t it? It’s literally acknowledging the prevalence of “meetings that could have been an email” and the solution isn’t “meetings only when essential,” it’s “develop fleet of clones to have waste of time meetings on our behalf and convert meeting into email form for employee consumption.” 😂 It’s so over-the-top bananas a way to solve a very simple problem. Only the most out of touch billionaire could conceive of it! Or he’s just appealing to people who don’t know better. “Hey everyone, wouldn’t you rather be on the BEACH than waste your time in dumb MEETINGS?” And his REAL product is to get us all to consent to give Zoom access to whatever invasion of privacy/training would be required to create some wispy facsimile.
I work in a political environment in Washington, DC where "word salad" is the common lingua franca. It's great because words don't have to have actual meanings. Anything can mean anything and we deal in magical thinking all the time. Someone can get on the floor of the House and make a speech entirely in gibbrish, and it will make the 10 o'clock news. I was kind of hoping high tech business would be different. However, I am happy people are preparing for the coming of Quantum Quantum Quantum.
They are different at the lower levels, because they have actual problems to solve with real world constraints. It is only as you move up the management hierarchy that this disappears.
I always thought it was a nerdy reference to Magic the Gathering where spells are resolved on "the stack" where the last one cast is the first to resolve and then you make your way down the stack of spells.
I have been in tech for nearly two decades and the amount of tech leadership that is absolutely SURE magic is going to solve their problem someday is insane.
I propose a different exercise: if Mr. Yuan believes that AI will be able to effectively stand-in for him and what he does in meetings, respond to his emails, make decisions on his behalf... what if we take him at his word that he believes that in 5 years AI will be sophisticated enough to do HIS job, maybe that tells us more about his job and value contribution to the company than it does about AI.
I actually trust that ChatGPT could perform better than the average CEO in making business decisions. Its just going to mostly follow whatever is the most commonly suggested solution to any problem. Even accounting for hallucinations and Reddit shitposts that should be more than adequate.
And of course Zoom is going to keep paying him for sitting at the beach. For some undisclosed reason the guy expects to get a paycheck in a world where AI can replace him.
hahahaa, i feel exactly the same way. P.S. you might like to know that gmail has an "unsend" feature that lets you "unsend" an email by clicking a button. In reality i think it just delays the sending for about 5 minutes, but regardless it's saved my butt a few times!
Would be pretty wild if email ended up being the biggest actual use for a digital clone. Just take a meeting that could have been an email, have your digital clone send you an email with a summary and never again complain that "this meeting could have been an email". Better yet is an autoreply that asks the organizer to send an email summary of what will be covered in the meeting before it happens. Without needing AI assistance.
@jasonhatfield3084: Well, laypeople assuming from the armchair that highly successful businesspeople barely understand business and reality more than they do but were much luckier and meaner (and magically not as scared of starting and maintaining a business as they were) such that their near-equal business knowledge actually gave them a successful business is laughably simplistic and patronizing as well (speaking of laughably simplistic and patronizing things).
I hate it all so much, please someone make it stop. My doctorate was in Computer Science with a focus on probabilistic modelling circa 2012, long before all this AI bullshit took off. We were doing CNNs in Matlab and the jank was real. Sequence modelling was RNNs and it was utterly useless. It was a different world and the science was genuinely interesting and novel each year. I left the field when DeepFakes became a thing. As a scientific discipline there is no code of ethics. Research projects rarely go through an ethics process. The industry side is almost entirely marketing snake oil. No one is held accountable. Consent is never respected.
Most AI researchers: "There's like a 5 or 10% chance that this technology ends the world. It's going to become generally more competent than humans in the next few years. Decades, at most. We don't know the first thing about how to control it when that happens." AI company CEOs: "I don't trust anyone else to build this doomsday tech responsibly, so we need to cut corners on safety so we can create it first!"
@41-Haiku You confuse AI researchers with salesmen. It won't get generally more competent than humans anytime soon, and we would have to do something really stupid (give AI control over nukes or something) to create a dangerous situation.
I told my head of department that email and meetings weren't real work but did make people look busy. I am no longer employed. Anyone whose been to 3hr staff meetings knows that they accomplish nothing other than making staff feel like their voice is heard. Which is fine. But the decisions were already made before the meeting and the meeting is pretend work. Less meetings. Less email. Getting yourself fired shouldn't be the solution for too many meetings and too many emails
I'm a software engineer. I can't trust a junior to substitute for myself in a meeting. Another human being, who went to university, got a degree. How the hell will an AI agent take my place in those meetings?
@@SkorjOlafsen hah yeah i really appreciate that the folks in my workplace understand the harm of pointless meetings - still remember one of my bosses ripping a bunch of folks from the customer a new one for wasting everyones time in a meeting where at least half the people had nothing to contribute or gain and the other half were talking in circles without a clue of where to go with it XD appearances like that really helped cutting down the pointless meetings
But the Overlords won't know the difference... for a while... and then instead of complaining about the ai agent, it will be the human who gets the backlash
This is a great topic and happy UA-cam added this to my algorithm. Some important points that shouldn't me missed are 1. Data harvesting is a big business for a reason. Your phones are the gateway I to your psych. You phone records literally everything about you. Social media is a literal time line of your life and because everything you post belongs to that company your saved in a little file, name, phone number, emails, photos etc. Every app you use or new car you connect to now has full permissions into the data of your phone. Even calls and messages. Due to nothing being private other than companies... And us the consumer having a severe reliance and addiction to our phones... They do have the ability to mimic you. They have the ability to manipulate you via all smart devices. I'm not saying AI will ever be as proficient as they suggest especially not in 5 years. In fact if people understand this and stop buying and using these smart devices... Maybe take our privacy back we won't have this obvious ballooning problem.
That's her brand. Find something that sounds cumbersome or outright dumb and then keep repeating it to keep driving home how dumb it sounds. And we're all here for it.
That's the one thing AI is kinda good for. Transcribing and summarizing meetings. Can even ask it if certain things were brought up and timestamp certain topics that were discussed.
@@ZacDonaldunless it hallucinates and now you’re working assuming the summary statement “X is resolved, do not worry about X” was true. Only actually Y was addressed, but not resolved, and X never mentioned. So when X pops up and causes a massive production stop, you’re on the hook.
@@piedpiper1172 That's why if it's anything important you either attend the meeting or double check the transcript or audio itself. It at least saves me from having to actively scrub though a meeting trying to find anything relevant.
If all you're getting is a summary, you're just going to drop it in the circular file anyway. The point of inviting a person to the meeting, is to make them care about the content of the meeting.
A way to baffle your listeners: Everywhere you'd previously have said"It's fine," you now say "It's down the stack." (And as for your question, yes, when I was a programmer long ago, I recall that the books all explained it that way -- with the pop-up dish analogy. Nice goofy picture you chose!)
Yes, but when Eric Yoon (CEO of zoom, a video conferencing company) says stack, I'm pretty sure he means "technology stack". Not least because I doubt he knows what stack memory is, but also because glossing over all the solutions they rely on as their "stack" is extremely CEO shit.
He's referring to a "call stack". It's a stack data structure in which each item is a call from one thing to another. He's saying something so sad, that he and Zoom don't have to solve these problems because someone else will. So inspiring.
It's frustrating but the problem is because of the relentless positive media coverage and the complete inability of our journalists to press them on even the most basic of pitfalls and shortcomings of the technology lots of people DO believe that this transformation is possible. The level of credulity from journalism outlets to me can only be explained by the algorithms that dominate newsrooms today say that AI positive reporting generates clicks.
"In a meeting, we are in a conversation" Ah yeah, working for a big corporation. Many meetings are just people sheepishly sitting in a meeting and writing their hours not contributing anything because nothing what they do matter in the slightest. This AI "solution" is just adding more nonsense to the pile of corporate nonsense and non-work that is already being done.
I remember the film Idiocracy (2006) describing a dystopian society made dumb by social media. With AI in the mix the real future will be still more dystopian.
🤓 correction: Idiocracy’s dystopia wasn’t because of social media it was because of a scenario premised on the writer’s implicit belief in eugenics. The setup of the start was all “poor dumb people are breeding and rich smart people aren’t breeding so all the kids will have the Dumb Genes!!!” There ARE some funny jokes and some interesting critiques of capitalist society inside the movie, but the premise is very much eugenics not social media
FWIW, Shor's algorithm can be used to solve both integer factorization and the discrete logarithm problem. This compromises most (really all) of the public key cyphers in common use. The concern isn't that some hacker will achieve quantum supremacy in their basement, but that large corporate and/or nation state actors will, in a clandestine manner, develop and deploy quantum computing to compromise the confidentiality of our public key cyphers. If we wait until after scalable quantum computing becomes a commodity before broadly deploying resistant cyphers, its probably too late.
100% I feel like she really misunderstood both how quantum computers work and the point of quantum-resistant encryption (not to say that Zoom has a useful implementation of it, I’m not qualified to comment on that)
@@Anything_Random I think she didn't want to go into the weeds. The technical challenges remain, no one has built a general purpose quantum computer. Additionally, Shor's algorithm requires unbounded precision of quantum state (it has a large number on the order of the key in the denominator of the quantum states you need to produce), which would require dealing with thermal noise. Additionally, it's not the killer tool you may be led to believe - it makes larger key sizes more tractable, but it's not, like, instant and still relies on brute-forcing.
In fairness to Zoom post quantum cryptography is a real thing - it basically boils down to using algorithms and key lengths that are expected to still be impractical to crack given the expected speedup that a theoretical quantum computer would provide (I am far from being an expert, but as I understand it, for many algorithms the speedup is only quadratic, not exponential)
My understand (not a cryptography expert, per se, but am a cybersecurity practitioner) is that symmetric algorithms only really need their keyspace expanded to be "quantum-safe". Asymmetric algorithms gets trickier. Anything that requires factoring prime numbers for key exchange basically gets destroyed by quantum machines. If memory serves there are *some* asymmetric ciphers that are "quantum-safe" (those based on "coding theory", I think), but I'm pretty sure those don't rely on factoring prime numbers or similar problems. I think all asymmetric algorithms in widespread use would be toast, though. Bruce Schneier covered that a bit here: ua-cam.com/video/bjopJ-_vAUE/v-deo.html
The fact that the CEO of Zoom has so few important meetings, interactions, and decisions to make that he could replace himself with an AI twin speaks to that plot of the Zoom stock price you love to show... The people doing the work can't do this, as you pointed out. It's all for CEOs to further reduce the actual value they add to their companies. You can be sure they will find a way to increase CEO pay based on the number of clients they have "working"at any moment.
Real quick on the post quantum thing: Quantum won't be better at every thing, but there are algorithms available to quantum computing that aren't available to classical computing. Shor's algorithm is one such algorithm and can help find the keys used in the most common cyber security algorithms presently in use. This makes deciphering things tractable. All post quantum is, in this case, is switching to existing known secure algorithms that aren't broken by things such as Shor's algorithm. We can prepare for the thing we don't know because of the field of computational complexity.
Exactly, we may not know how the quantum computers work, but we have the mathematical framework (computational complexity) to know the limits a quantum computer would have. I would also add that the feature is actually useful, but only to the small number of institutions that may have a secret that has to be kept for many many years in the future (i.e. National Security or Defense).
@@aceman0000099 which is reasonable, because they dont have a secret that requires that level of preparation at this point - its still better to react to a breach and switch out the compromised secrets but that does not mean this will always be the case - so having an implementation, that could be turned on in case quantum computer attacks on typical security become feasible, would be very useful - its the difference between switching some parameters and deploying a new build to restore security and having to invest time and research into figuring out how to re-secure your systems mind you - their implementation is probably just a library they use that has that capability already included, but disabled by default for the stated reasons... so not really special or their work... they just have a black box and can flip the on-switch if need be
Do you think that pre-transistor computer scientists could have predicted modern hacking vectors? No shot. "just switch to Shor's 4head" is the product of pop-sci brain
Extremely cathartic video. I'm going to be repeating 'Eric Yuan, the CEO of Zoom, the video conferencing software, Zoom, a company that looks like this' for some time, great comic timing & delivery
@SkorjOlafsen: It seems more like an insistence to turn an often positive trait into a bad one because one doesn't have the flexibility of thought to imagine someone who disagrees with them having a positive trait that they don't have, rather than a clever term.
I don't care whether or not LLMs could actually replace human workers, I'm more worried about companies becoming convinced that they can and doing it anyway. If anything this man says is remotely true we should be extremely upset. On a side note, I don't think hallucinations are ever going to go away. It's not a bug, but rather a feature - A bad feature, but a feature all the same.
It's neither a bug or a feature, it's just how the ML models work. They're just statistical prediction engines and sometimes (a lot of the time) what comes out is confident nonsense because they don't have any kind of knowledge base or anything, just a list of tokens and what tokens are likely to come after them
Having quit a toxic job a while back that did a hard pivot into AI (though the job was toxic well before then), I can say for certain that what you're worried about is already happening. Companies *will* try replacing humans, even if it's not viable, as long as it's not a *complete* disaster and it saves them money in the short term (because lol who plans for the long term??). Customer service in particular is about to get slaughtered by shitty AI replacements that are terrible at doing what a real support engineer does, all because the execs think they can get away with it. They'll keep a few humans around to mop up the inevitable AI fuckups. The humans will be overworked to death since they'll have a "team" of two or three people when they should have 20+ because all the others were swapped for some idiotic chatbot. The remaining humans employees and their customers will hate it, but if it doesn't wind up costing the companies that try it more than they save, then they'll keep pushing.
@@decaydjk8922 It's not just that; it's also evolution in action. In the end, the reason it comes up is the same as the reason why it comes up in humans - it appears to satisfy the constraints well enough, and gets away with it enough. Just like so many humans in management roles :D I'm pretty sure even a modern chatbot based on Chat GPT-4 could actually replace most bad managers in anything that doesn't require them to show their face; they're already brilliant at turning a normal no-nonsense single sentence answer into three paragraphs of corporate BS that doesn't actually say anything :P
Some companies are already using Image generators in place of Artists/Graphic Designers. I wouldn't really say they are "convinced" though - they know AI is worse quality than real artists, but *they don't care* since its cheaper and faster.
Words do not describe how psychotically angry OneDrive makes me. The way it tries to force ALL of your files online makes me want to go back to dual-booting linux.
Back around 2016-8, blockchain was all the rage and I worked for an older tech company. Instructions from up top was to just say our products were "blockchain ready", whatever that meant. Shoehorn it into every sales meeting and presentation. Then one long weekend, every developer and researcher was voluntold to attend a hackathon to make these blockchain features. No use case that needed it, no client wanted it, just invent solutions for problems that don't really exist. And all the solutions we did develop already existed for years. All these AI integrations seem born from the same mindset. The issue is unlike blockchains, Generative Data Models do output semi intelligible results and it is easy to mask any new (or old) feature as "AI powered".
@@culwin I think it's impressive, that the bot correctly uses the Skinner meme. Though it might have just copied that comment from the comment section...
= "kick the can further down the road" = "we'll cross that bridge when we get to it" = "just trust me bro I promise in a few years someone will figure this out we'll go to the moon I swear bro"
I have already made 164 digital agents for my company and am still testing their behavior across different llms. They work great, and I am gaining a ton of insight on various roles across an enterprise. But, I haven't eliminated 90% of my work, I haven't eliminated any work. I am basically now a secretary for 164 digital agents. I have 164 times the amount of work than what i started with.
@@piedpiper1172 LOL no, but you can get high fidelity responses today, it just takes a lot of work and you need a lot of human sweat equity to keep get it there. But also I agree with the overall sentiment of the video, I don't see how we are going to get LLMs to give us multi-shot answers to zero shot prompts in 5 years time and I am not sure how Zoom will get us there.
I was SO worried you were gonna dunk in Nilay, but he and the rest of the Verge have been one of the only outlets I’ve found that be been seriously critical of the AI bubble and the surrounding characters
The guy interviewing this CEO should have dropped a Wolfgang Pauli on him: "That's not right. That's not even wrong." I'm learning so much from this channel! 😁
The funniest part about this video is Angela thinks she's criticizing the project when she's actually architechting it. I wonder how it feels to be the most dangerous woman alive and not even know it.
Software nerd, reporting for duty. Post-quantum encryption is definitely a thing, but Zoom is full of shit. Long story short, we actually _do_ have quantum encryption already, because it's about the algorithm not the computer. Shor's Algorithm is the one that's best known, but there are others. And since we have the algorithm, we can study its properties and build encryption algorithms that can't be broken by it. Getting a little less short, there's no reason for you to be very worried about this if you aren't wanted by the CIA or something. The kind of actors that would have the resources to pull off this kind of "store everything, decrypt later once we get the fancy computers" thing are few and far between, and they don't care about your zoom meetings. And since encryption is computationally expensive, few people are bothering to use these post-quantum algorithms. My guess is Zoom grabbed one of their engineers and had them spend a couple weeks adding support for one of the post-quantum algorithms in the back end, specifically so they can talk about quantum bullshit and try to juice the stock price. But they don't use it because it is both pointless and expensive. But using it was never the point, putting it on the website was the point. Also it's very possible that tomorrow we come up with something better than Shor's Algorithm that breaks the current "post-quantum" algorithms, so even if the stuff were turned on, and useful, it wouldn't necessarily do anything.
So you're a non-tech company with a vendor selling you "post-quantum" encryption. How do you know their not lying? You can't test their claim. The rest is theory.
@@SkorjOlafsen when a company tells me they're GDPR compliant and will not store my password in plain text, I can't verify that either. It turns out even Facebook did that once. If the software is closed source, all you can do is sue them if they lied
I've found Zoom's comments about the limitations in their post-quantum encryption quite hard to understand. I believe what they're trying to say is that they secure the exchange of the encryption keys using both a regular (elliptic curve) cypher and the (supposedly) post-quantum Kyber cypher. However, since they don't use Kyber for anything else, a theoretical quantum computer existing right now could still intercept your traffic and perform a man-in-the-middle attack to listen to your online meeting. They're not worried about this yet, though, because existing quantum computers don't have enough computing power for this to be considered a risk, which is completely fair
Came here to say the same thing as outputcoupler since I do work in computer security, and just add a couple points. 1. Technically, we don't know for sure there won't be a classical algorithm that can break all the same algorithms quantum computers can break. That is to say, nobody has actually solved the P vs. NP problem (en.wikipedia.org/wiki/P_versus_NP_problem). But there is no known algorithm that can do it (if we did, we'd have solved P vs. NP) and we don't think there will be, and that's why we say the encryption is secure. The same logic holds for quantum computation: to the best of our knowledge, this works, and if our knowledge changes, then yeah we're all going to have to adapt quickly but that's the inherent nature of (almost all) encryption. 2. There are actual use-cases if properly implemented (and I don't think Zoom has, I laughed out loud when she read that Zoom claimed they were the first to post-quantum encryption). It's fair for Dr. Angela to question when Quantum Computers will exist that will be able to crack this, but do consider there are organizations that retain sensitive information for many years, whether by law or some other policy, however inane. Eg. you're having a chat with Zoom's text chat to a medical provider, and they need to store records for 20 years. Even once the first quantum computer exists, no hacker is going to spend their quantum computer time decrypting any random person's messages to their grandmothers, but a hospital, or a bank, or a nation-state government, etc. can be a juicy target that keep sensitive records for a very long time. So the idea is, we don't know when a quantum computer will come, but everything you encrypt now has to stay encrypted for, let's say 20 years, or whatever the retention timeline is. Your reasonably alternative is to watch for an actual scaled-out quantum computer to come online, and scramble to re-encrypt some of their old potentially-sensitive documents before the computer becomes available to a malicious actor. But you don't have to re-encrypt anything past the point where you had post-quantum encryption. @SkorfOlafsen, you actually can verify that a company is doing post-quantum encryption in several ways, but one thing to note is that it is actually possible to run quantum algorithms on classical computers, it's just that one step is hideously slow on classical computers while it has been shown to be fast on any actual quantum computer that can scale up enough. But secondly, these algorithms are mathematically shown to be secure against their adversaries. Quantum ones against the quantum algorithms, classical ones against the best known classical algorithms. They are published. So what you do is validate that their encryption scheme decrypts correctly with completely different correct software given the same algorithm and the same key, and that validates that they are not lying about their encryption. Which again, is the same way you validate every claim about encryption. If you want to validate that those published algorithms are themselves secure and correct, that's where you call in a bunch of mathematicians and computer scientists who peer reviewed this stuff and if you really want to validate it yourself, you ask them about learning materials. Find the Dr. Angela Collier of encryption who will not-science-communicate ideas to you and you can start learning when they show off their favourite textbooks. I'd start with some of the oldest classical encryption algorithms because you can kind of do some of them by hand with a calculator semi-reasonably for short messages and then see how near-impossible they would be to do without knowing the secret key.
Ya, and to answer specifically to Angela's question around 40:05 (though I suspect she already knows this, probably?): Even though we don't have practical quantum computers right now, and we don't know what the specific capabilities of quantum computers will be in the future, it seems to be very well understood which algorithms can take advantage of the quantum parallelism and which cannot, because "math" (I won't pretend to understand it, someting-something-quantum-destructive-interference, but I'll trust the computer scientists that specialize in this). So, barring some mathematical breakthrough on the level of "P=NP" (at which point all our classical encryption techniques are broken anyway), I think someone using an encryption algorithm that is quantum-ly nonparallelizable, can *honestly* state that they have implemented "post-quantum" encryption. That said, what Zoom is stating is just as meaningful as a window blind company claiming to "include a lead-free formulation" in the product by using one layer of lead-free material, and then covering it with lead.
Eric Yuan's huge mistake is claiming that this tech is 5-6 years away. He would have been totally safe if he has said that it was 10 years away, just like everything else is 😇
This technology didn't exist 7 years ago. 2 years ago, it was completely useless and could barely string a sentence together. Just by taking the sucky, useless version and scaling it up, AI language models gained the ability to write essays on any subject, give relationship advice, do math, write code, pass the bar exam, pass theory of mind tests, use a text-based scripting language to accurately draw things, and autonomously exploit zero-day cybersecurity vulnerabilities. You can look up examples and research papers showing each of those things. Are you _really_ sure you know what it will and won't be able to do in 5 years? Because the foremost experts in the field say they don't know what GPT-5 will be able to do. Not even the scientists and engineers at OpenAI know.
@@41-Haiku But none of what you said is true. This technology did exist 7 years ago, Machine Learning and Neural Networks have been around for decades and they we're completely crap the whole time they were actually pretty decent at some things like games and solving simple problems in simulations. The only new thing that's happened is that generative AI has gotten sorta good but that also isn't a totally new piece of technology, it's been around for at least a decade or more and now we just have powerful enough GPUs to make them good. Also they can't do any of those things like at all. Current LLMs can write summeries and that's generally it, they can not do math they simply interpret it and hand it off to a math module, and they can't fucking pass the bar exam where did you even get that from? And like the thing is people say this exact shit about literally every technology, a few years ago it was NFTs and Blockchain, before that it was Big Data, whenever a technology reaches the bare minimum of functionality there's always people proclaiming that we're just a few years away from it solving world hunger and it of course never happens.
@@hedgehog3180 heh yeah - the bar exam thing reminds me of that lawyer who couldnt be bothered to look up actual legal reference material and just asked chatgpt for stuff that supports his case and then filed that without checking... and the bot completely hallucinated everything... im pretty sure that lawyer is no longer allowed to practice XD
This comment section is going to be full of people pointing out that the stack from "full stack developer" and "stack overflow" are different stacks. A difference which doesn't affect her argument much
...but that was the joke...
@@acollierastrongl it flew right over my head
@@timonsku If I pin the comment will people realize how funny I am?
@@acollierastro it'll act as a filter =) by all means do it.
@@acollierastro Well, it has a better chance of success than Zoom, the video conference software company, has of developing AI avatars.
14:53 ahh now I understand. He thinks people will be replaceable by AI because a CEO is already an unaccountable black box that takes emails as input and bad decisions and hallucinated reality as outputs
In this case he is unto something, this pitch serms to be tailor-mad for my lazy incompetent boss and tbh if she were replaced by AI it would problably be a slight improvement.
Imagine how much money companies could save by replacing the C-Suite?!
LOL this thought was sort of lurking around in my head and you just put it into words perfectly
Got his ass
@@LiteralmenteFadul Huge improvement, the best boss I can hope for nowdays is one that doesn't fuck things up too badly.
I love that this CEO is just like "Yeah an AI clone could do my job easily in 5 years." What a self-own.
In fairness I'm sure right now his devs could cook up a script that was great at making C-suite executive decisions in a few weeks; the challenge is hobbling its performance to better match the faulty decision-making of the average flesh creature CEO.
Felt the exact same thing lmao. More than anything this idea just goes to show how useless certain jobs are. Really showing his own ass here haha
Maybe they can fire every CEO and hire more useful workers, like customer service.
Considering how mind-bogglingly stupid most CEOs are, they could do his job now.
i had same reaction. like his meetings must be a lot simpler than mine. he said the AI summaries are really accurate. i'm using zoom's AI summaries.. i like that they're there but they are FAR from accurate. mostly it's good for quickly getting a sense of what was talked about in a meeting... but it mixes up things that are said, so you can't trust any of the actual information in the summary.
either he's not using the summaries or his meetings are way easier for an LLM to summarize than mine are.
I so love the idea of using "down the stack" to just assume some tech intern will turn anything into a low cost microservice you can purchase an api key for to solve your impossible problems.
"Yeah we're a bicycle startup but the hook is we let you ride back in time to visit dead loved ones. Oh time travel? Nah we're not working on it, but it's down the stack. Someone'll offer a code integration for it. I give it 6 years."
This comment is gold
When they release the time travel webhook I am definitely not using to go backwards in time but to get right the hell out of here.
Lol !!!
Basically the CEO goes 🤷♂️🤷♂️🤷♂️🤷♂️🤷♂️ 💰 and the line goes up
It's literally just "a wizard will fix it"
I took a shot every time you said "Eric Yun: CEO of Zoom. A video conferencing company." The room is spinning.
have you recovered from the hangover yet, or died of liver failure? if so, f
This video was created to promote that Eric Yun is the CEO of Zoom. I literally cannot remember anything but that Eric Yun is the CEO Of zoom
Eric
It is a good thing you didn't take a shot everytime someone said "down the stack".
a company that looks like this
I'm reminded of Douglas Adams' electric monk:
"The Electric Monk was a labour-saving device, like a dishwasher or a video recorder. Dishwashers washed tedious dishes for you, thus saving you the bother of washing them yourself, video recorders watched tedious television for you, thus saving you the bother of looking at it yourself; Electric Monks believed things for you, thus saving you what was becoming an increasingly onerous task, that of believing all the things the world expected you to believe."
"They eventually turned it out, where it was free to believe what it liked, including that it had been hard done by. They let it keep the horse, since horses were so cheap to make."
Wow, a Dirk Gently reference in 2024? My day has been made.
There is one old sci-fi story which feels relevant to the situation, called The Machine Stops. It's about a world where everything is serviced and controlled by a complex entity called "The Machine" and it is kept in operation by some other system which nobody understands. One day it breaks down and because nobody remembers anymore how to fix it or how to live without it, humanity basically dies out overnight.
I find it kinda funny that that type of world is exactly what these new AI companies are trying to build. Let alone the absurdity of the design and feasibility, it's just a very depressing idea all in itself.
@@theonewithoutidentity Documentaries are fun!
We don't need AI for that. When I still had a PVR, I treated it like an Electric Monk. It had to watch a movie to record it. Good. That means I don't have to now.
The fact that these CEOs think that AI can attend meetings and be just as effective tells you just how useless most meetings they demand are... enter SCRUM...
And if AI are the only ones going to meetings... why aren't we just integrating them? It's insane to think that's how computer connections work. How is this guy a CEO of a technology company?
the answer to that question is: capital.
he had capital and thus he accumulated more capital and power... which in turn allowed him to aquire more capital
no skills or knowledge required. just capital.
Maybe they will be as effective as humans in meetings, because my experience, tell´s me that meetings are just a waste of time. They like meetings, because that way, they don´t work!
It's like they think, if they have all your Jira tickets and IM's they have enough to make an AI clone. Maybe if I work remote, and they capture my every interaction with the computer, they can?
The thought of a bunch of AIs having superficial conversations about the weather with each other over Zoom to fulfill some rich idiot's deranged idea of "productivity" is morbidly hilarious
He's just selling snake oil. That is the modern CEO's job--just lying to increase stock prices.
"Daniel, HR needs to talk to you about one of your AI zoom clones"
“Im not racist, the clone is probably just bugged”
“Daniel this was a physical conversation. There’s nothing digital.
“Idk the problem is probably down the stack”
InfoSec needs to talk to you about AI clones. Also the SEC, because it leaked the Q4 results. We also wired $4 million to some guy in Weed, CA. No, the AI did not hallucinate that name.
This has me laughing so hard.
"Sure, I'll tell one of my clones to call HR"
"No officer I've been sitting home all day, oh I did take the dog for a walk. Puppers decided to take your mom for a walk while at it. But that was just around the block I haven't even gotten in my car today so there HAS to be some mistake here..."
I watched you hand him the bag, take the money and drive off. It was so obvious I think I heard a baby with its mom near by say "a deal!" as its first words....
"Must be one of those weird zoom A.I clones that are going around. beyond that I have no idea what's going on"
Well it was but the moment we moved in he, as our tech put it 'moved to a different algorithm that had given the name and address of who sourced it's code and set its parameters up' plus the Zoom account that it used instead of a phone is linked to your accounts, and social media...
"Figures, they gotta do something with all that data that the internet collects off us. I think your investigation is starting to lean towards big-tech not me..."
Well sit tight. we will have to wait a minute until the A.I clone of the
judge who signed off on the arrest warrant to show up to expire it...
Wait a minute.i am suddenly starting to realize maybe Zoom is onto something here...
The last human working at a company staffed by ai clones is actually a great scifi short story idea
It got turned into a shitty game called "the last worker"
You gotta read Service Model by Adrian Tchaikovsky. It's a brilliant, satirical, moving look at this subject
frame anything from the modern era from a tasteful and introspective perspective w some good cinematography and music and it would be a good short story. the times we're living in are borderline horrific
That's just the Jetsons where Mr. Spacely's only human employee is George Jetson pushing a button to tell the AIs in charge of manufacturing sprockets what to do.
It shows how worthelss CEOs are. Their jobs are 90% meetings that are just redundant. So they assume everyone's is.
A video of Angela talking for almost an hour about how fucking stupid the AI craze is just what I needed to feel better 👍
with a beer n a j outside is fantastic
Whaaaaa emphasis on the j
@@RYOkEkEN if you drink that can you really make fun of "AI"? ....
My ai clone recommended this video to me. Thanks zoom
24:10 this moment of true despair 😄
Remember when Uber’s entire mission was to create driverless cars, a task that everyone except their c-suite executives knew was going to take billions of dollars and decades of work to achieve, and then abandoned it after five years when it turned out it was going to take billions of dollars and decades of work to achieve?
Me neither.
Had they achieved their goal, they would have replaced their fleet of completely disposable drivers and cars that they pay $5 an hour to use with quarter million dollar cars they have to buy and maintain on their own. The product with these companies is always the stock.
They're not stupid lmao, you have to be kidding me. They need to advertise and chase that to get funding. They would not be born or grow of they didn't go down that path
@@drchickensalad $1.3 Billion dollars well spent.
@@drchickensalad, But why should I trust them when it’s in their interests to lie to me? That’s the point.
@@criticaleventReplacing drivers with computers sounds like a good thing. What's your point?
The "hallucination problem" is even bigger than people think, because LLMs don't just hallucinate sometimes, their whole function is to hallucinate. That's how they work. It's just that a will tuned LLM has hallucinations that map to reality 90% of the time or more. It's ALWAYS just guessing, half a word at a time, based on the data it's been trained on, the input from the user, and the previous words it has written. It's ONLY hallucinations.
What is really amazing is the fact that it hallucinate so well it can do all the things we've seen it do. But the hallucinations will never go away, because that's all it's actually doing.
Tbf we work like that too. We don't have real access to reality, all the input our organs receive is filtered through our brain and constructed into a simulation in our brains, we never "see" stuff, out brain makes a model and we "see" the model. Conscious thought is the most abstracted part of our brain. That's why you can believe totally wrong stuff and be 100% sure, because you never got the real version and everything in your mind is write/read access only, so we change everything when we think about it. We are trapped inside our own matrix if you think about it.
The problem with AI is that it can't check if his "guess" is actually correct as we would do.
@@Puerco-Potter
It can't be real that we don't access reality.
That idea can't be true about reality without also being self refuting.
@@Puerco-Potter
We can't check either...if we can't access reality.
ChatGPT Crunch: Oop! All Hallucinations
i love the use of repetition in this video. "eric yuan, the ceo of zoom, a company that looks like this" and "the beach" and "its in the stack". had me giggling
i kept thinking they were gonna be in the next title card
I think Angela has made one kinda major mistake in her analysis: as a scientist she is used to having meetings that matter, where people use information to solve problems. Those are categorically different from the meetings Eric Yuan (the CEO of Zoom) is talking about. When you deal with tech companies there are A LOT of useless meetings where people in charge make small talk and say a lot of nothing for 45 minutes, and then spend 30 seconds saying it's ok for the people who actually know what they're doing to get back to work.
CEOs would rather create robots to have meetings with each other rather than acknowledge that a lot of the things happening at the management level are a waste of time.
But what are the people who rely on meetings to avoid actually working going to do all day then?
"We would like to remind our employees that as a company policy, it is not allowed to send you Zoom AI clone to the all hands meeting. You need to show up in human person instead."
IDK, most academics I've met complain about having too many useless meetings they have to go to. Probably doesn't compare to the job of somebody who's main job is going to useless meetings.
Many workers (even professors) need to suffer through useless meetings, and all we can do is grumble about it. But it doesn't make sense for a CEO to complain about that since they're responsible for solving fundamental organizational problems like that. Effective meetings don't happen by accident
Watch full video, she talks about it. It's down the stack.
I will never again say or hear "Eric Yuan" without, in my mind, finishing it with "CEO of Zoom, a video conferencing company". This video is Dr. acollierastro at her most snarky and i love it so much.
You'll need to carry a copy of that graph at all times.
a company that looks like this: ෴
It's my favorite thing that she does 😂
@@SheeplessNW6 Dr. Collier may be the only human who deserves access to hammerspace irl just to pull withering graphs out of in the middle of a takedown
A company that looks like this:
📉
CEO: 90% of everyone's job is reading emails and going to meetings, right?
Everyone: ::sharpens the guillotines::
_Maintenance de routine!!_
What do you do? I try to minimize email and meetings so I can get work done.
@raygivler are you aware not everyone works corporate office jobs
And the more emails and meetings you do, they more you are paid. One burger flipper at McDonald's who only screws up an order once a day does more for the company's profitablity than all of the C-Suite combined. Outside of the CEO, COO, and CFO everything can be handled WAY below the top floor. Stop wasting money giving your stupid friends and incompetent nephews jobs!!!
@@cajunguy6502 companies well known for passing money away and not being greedy. I pray you're 16 and your head hasn't suffered trauma.
I think this video should be called "A rational person tries to understand sales hype." I go through this exercise for about 50-70% of my day. I remember spending 3-4 months of 2020 trying to figure out why people were taking about the "New Normal Post Covid". The "New Normal" just a change everyday is a new normal.
After ten seconds I was screaming in my head "But this will make video conferencing obsolete! Why are you promoting this, CEO of Zoom?"
Also, this reminds me of how people in the 1950s predicted that in the far future of the 1980s, everyone would have a robot to play table tennis against. They basically predicted Pong, but as they couldn't predict computer games they instead imagined humanoid robots able to play physical games. Which still isn't a thing.
A table tennis robot is readily possible with present-day technology; it'd just be really expensive for the value it provides.
@@Ithirahad ua-cam.com/video/Jhe1h-eip2Q/v-deo.html, not even too expensive it looks like
A future where we all each have an individual robot that just plays pong is the superior time-line
@@Ithirahadwe have those at the Carnegie science center I think, but they serve as mostly demos of what robots can do
It could be a thing, but few people want to play against a robot.
Plot twist: This was an interview with an AI clone and it hallucinated the entire interview.
LOL 🤣
I mean it already reads like it was written by an AI.
Yes, because most of that is already possible. Sale people talking bullshit 90% of the time to get the commission.
How dare you make exactly the same joke I was going to 😂
That sounds like a script for "The IT Crowd".
I think Eric Yuan The CEO of the video conferencing company has identified the exact class of worker that can be reasonably outsourced to AI.
Himself
Hey, he just makes up vaguely convincing stuff and spews it out as fact! He's a perfect match, and his avatar could be so much hotter, like ripped, and totally sigma with all the latest tats and facial hair trends.
@@qwerty4324ify I don't think you're giving Eric Yuan The CEO of Zoom enough credit... Sure the market has changed and times are definitely tougher now but Eric Yuan The CEO of Zoom is very well respected by others in the field. For example, during his early days as CEO of Zoom, he really made a name for himself, being the first and so far only person to ever have that name; Eric Yuan The CEO of Zoom
Himself?
Just like Elon Musk
13:49 “what if my clone tells a customer yes we can build something and afterwards I realise that no we can’t build it but it’s already too late because the customer paid 400k”
That’s exactly what companies currently do though, especially tech consulting firm. Just replace clone with salesperson.
The difference is, at least a salesman can be held accountable for overpromising.
The fun thing is that the people who own companies think this will work entirely because all they really do is listen to the person who agrees with them the most in the 3 meetings they bother to show up to.
Yes men will devastate a company and then jump right into another company with the same people, NOT because they are good at their job but because they tricked the idiot CEO into thinking he is good at his.
I also feel like these tech CEOs will say anything to not tank the stock price/ maybe even rise it since they often have a lot of their wealth in company Stock and get also payed with Options often.
It's a problem with corporations in general because they all think that any person is interchangeable with any other person. The money saved by having a competent person in a role is harder to quantify than the money saved hiring someone you can pay a lower wage.
It's less often the company owner, and more typically the CEO who was hired after the company went public. He's a psychopath who lives to manipulate stock price. You can blame him for being a narcissistic idiot, or you could blame the board who hired him. They are morons who attained their market share by simply floating in a bubble economy...
@@cajunguy6502 "Not because they are good at their job" - they literally are good at asskissing the ceo though? That is their job.
btw this video is one of my favorite genres of media, which is "normal intelligent person who's not immersed in tech industry hype hears a tech guy and says the obvious"
Emperor has no clothes - genre
same
@@eetuhalonen9902 omg that's perfect
Except "normal intelligent person" hardly applies here.
As with any new advancement, focus is on pie-in-the-sky utopian bullshit. One, LLMs aren't ready, not even close. It's like the PDA hype of the 00's, when a PDA was an $800 replacement for a pocket notebook and a pencil. Give it a decade and these kinds of advanced algorithms might be less buggy, but they won't be used to send people to the beach. They'll be used for bioanalytics to show that Stacy in accounting is 13% less attentive today and to adjust her pay for the day accordingly. They'll be used to further existing trends like targeted ads or the replacement of CSRs in favor of chat bot assistants. Imagine, when Google can *really* pinpoint how many sponsored results they can feed you before you get frustrated and use Duck Duck Go! An LLM is just a bigger algorithm. That's it. It's not new tech, it's just a small advance in existing tech.
I'm quite certain that even Eric Yuan, the CEO of Zoom, a company that looks like this 📉, would appreciate the synthwave aesthetic on your transitions.
A lot more production quality than previous vids lol
The repetitive speech pattern would make for a great drinking game
@TheDarkAgez wanna play?
You forgot to mention it's a video conference company
So i work in quantum computing development. The post quantum encapsulation is about protecting companies with trade secrets against government actors. It is more that we know that if quantum computers work at all like we expect them to, current e2e encryption breaks, badly. So all e2e messaging and communication tools are bolting on enough post quantum stuff so that there is a chance that messages recorded now wont leak to these state actors.
This is them doing the minimum prudence in e2e encryption and then overmarketing it.
Protecting currently encrypted information against future decryption is an active and well-established research area. Angela should have been able to come up to speed on this topic in about 15 minutes, ironically in about 3 minutes using AI chatbots. This was one of the few times I have seen her flail around in confusion, and frankly I'm not sure if she is still my ideal mate.
@@briand66
Gross
@@briand66
Even if this is a joke, it’s still gross to read
@@briand66 Holy shit you really thought this was a normal thing to say...
There is this great article called "I Will Fucking Piledrive You If You Mention AI Again" which so eloquently drives into the heart of the problem. They are just latching into AI as a lifeline to give that stock line a bump.
Like a junkie walking up to you on the street spouting nonsense and twitching, ready to do anything if they can just get the next hit
Thank you for the recommendation 👍
when will people talk about AI normally again
@@badabing3391when, like crypto, everyone realizes that what's being sold now is a scam.
Looking forward to reading it over my morning coffee. The title alone sounds worth it.
“Zoom is a company that looks like this” is the most savage burn 😂😂
Companies are their stock price.. AM I RITE ?
@@_FFFFFF_ if that’s what they prioritize then yes
@@_FFFFFF_ 🍅🍅🍅
They've made 2.7 billion dollars of after-tax profit over the past 4 years. So the company is going well by normal business metrics but their profit margins don't match bigger tech companies like Apple and they haven't had much revenue growth over the past 3 years, so no one is interested in their shares.
@@_FFFFFF_ Yeah, pretty much tbh
"Do not attribute to malice what can be adequately explained by stupidity." is Hanlon's Razor. What really applies here is Grey's corrolary that "any sufficiently advanced stupidity is indistinguishable from malice".
well when it comes to capitalists, malice can be reasonably ascribed at any point - you dont become that rich without harming a LOT of people... and their primary goal is to perpetuate that... so their malice is well established
@@SharienGaming It's also possible to cover up malice with incompetence for "plausible deniability"
and if you read spinoza they're indistinguishable because evil doesn't exist
@@SharienGaming yeah honestly there needs to be nuance that the most significant irl “malice” is based in selfishness. Wanting money and power, not wanting to see the world burn. The joker does not exist- when people hurt other people, generally, it is with the aim to benefit themselves (or to further some goal that they believe should be furthered- self destructive acts of spite fall into this category- but the mundane cruelty of exploitation is 100% the simple act of selfishness)
What seems like malice to the oppressed feels like altruism to the oppressor.
"You know, the video conferencing company" is probably my favorite sentence in this video
"which looks like this 📈📉"
Also, this guy is a CEO. "Meeting" (for him) is actually a Presentation. He doesn't really attend meetings.
Or he does show up, but never says anything. 😂
Any meeting that is so pointless and low-stakes that it could be attended by a dozen chatgpts talking to one another with no noticable negative effect on the outcome is a meeting that never should have been schediluled
@bensmith3890exactly
@bensmith3890*down the stack
@@andrewcapra7153 I'd probably be more efficient by cutting down in unnecessary "ass kissing"
I like the recurring "no-one uses zoom, they use teams" gag.
I however, work at a company where Teams is the official "chat app", and Zoom is the official "video conference" app.
It seems wild to me to be paying both enterprise licences for low thousands of people, yet here we are...
Are you working at Zoom (the video conferencing company)?
I can only imagine that there’s some archaic reason for that, even if the reason is that the guy that makes the decisions had one tiny nitpick about some teams/MS video detail
@@walaraubo 100% it's because they paid for a Zooms enterprise licence near the beginning of the pandemic, then realised half-way through that Teams existed. Can't back-track on Zoom, though. That would mean you made a mistake.
Ahaha, my previous employer had Teams, Webex, Skype (for SIP calls, but we finally retired it in 2020), Zoom and RocketChat
It was a NIGHTMARE
Yep. My old boss hated teams video and wanted all of our video calls to be on Zoom while everyone else messaged/called with teams.
CEOs when AI can do a hard job better than a human: "I must solemnly lay off half my workforce."
CEOs when AI can be a CEO: "I must go to the beach"
The funny thing is, all those downsides of AI that make it unfeasible for most jobs i.e. hallucinating, waffling, refusing to admit it only understands the surface level of a difficult problem, are all traits of a successful CEO, therefor making CEO the best-fitting role for these AIs.
Edit: haha! Just watched the rest of the video. It's been addressed hahahaha
I know, I do the same thing and watch half of her video and find out she disqualified my comment later in the video.
@@jeffbguarinoAs someone with ADHD, if I don't pause the vid immediately to put my thoughts in the comments, I won't remember what I wanted to say.
@@rsm3t Hey you can always go back a delete your comment, or just leave it there and start a conversation.
Imagine being in a meeting with a bunch of AI instances and in 4.3843 milliseconds the meeting is over and everyone has already left.
I'm just imagining how much fun could potentially be had by being the only human in the meeting, gaslighting the AIs and trying to turn them against each other.
At minimum it will at least highlight how useless and unnecessary 90% of work meetings actually are.
I work in cryptography, and the post-quantum stuff actually makes sense because there are two attack types: (1) capturing encrypted data today, and breaking *confidentiality* by decrypting it with a quantum computer in the future ("harvest now, decrypt later"); (2) using a quantum computer to forge signatures and break *authentication* (~impersonation).
Zoom is using ~standard solutions to prevent (1), but not (2). So if Russia develops a quantum computer in 2030 and has been capturing your data all this time, your conversations from *2024* are protected (1). But your conversations from *2030+* won't be, because Russia can steal your account or impersonate the Zoom servers or something (2). Presumably Zoom will eventually protect against (2) too. Anyway, they communicated this difference very badly.
And migrating to PQC algorithms takes a lot of effort, so you want to start years before quantum computers get big enough, even if there's a chance they might never get there.
PS: the algorithms are being developed and evaluated by the best of the best in cryptography, in public (see NIST PQC Competition). Zoom didn't create them.
How do you know what solutions Zoom is using? Just taking them at their word? There's no way to test their claims, is the point. Seriously, the first rule any engineer learns out of college is "the vendor is a lying bastard"; if you haven't tested it, it's not real. But I guess almost no one tests any of the security products they buy in the first place, so it's a moot point.
@@SkorjOlafsen This coupled with the claims about what able to accomplish, if not right now, but totally at some future unknown point in time. I'm convinced it's not much better than snake oil, very reminiscent of the crypto and nft pushes too.
@@SkorjOlafsen It's true that there's no guarantee, but they've published a decent whitepaper detailing the cryptography they claim to have implemented, and what they claim to have implemented looked like it would be very easy to _actually_ implement (maybe even less work than writing the whitepaper documenting it), with most of the actual work being done by off-the-shelf libraries. Given all that and the negative impact that previous security issues have had on their reputation, it seems they'd have a lot to lose and very little to gain by lying about this.
That plus the fact she thinks quantum computers would be used in brute forcing attacks makes me think Angela really ought to read on the fields of quantum computing and cryptography.
Thank you, I was also going to say, that post-quantum encryption is a real and important thing for any data that might be important for a while.
Especially considering that quantum computers already exist, just ain't big enough because of decoherence issues.
1. Collect underpants
2. Something happens down the stack.
3. Profit.
So step 2 is "stack underpants"?
2.5 In the case of a stack overflow, change underpants.
@@kelvincook4246 So pop one off the stack then?
Underpants gnomes 😂
I think you could probably skip step "1."
“Its 5-6 years down the line”. My man, you will not be the CEO of Zoom in 5-6 years, just look at the stock price
I've never understood how zoom became a thing, when it's entirely interchangeable with any other video conference app, and there were many before it.
@@ItsRyanStudios its a decent enough product, was in the right spot during the pandemic, very easy to use for about anyone and easy enterprise adoption. Definitely should have never been valued near $500 a share though
Just love seeing someone with a fully functional brain covering this with unfiltered honesty.
A big part of the reason why Holmes went to jail imo is she crossed the line from speculation and prediction to outright fabrication of results. It was no longer "in future we think we can do this" but "we are totally doing this right now and these major companies are on board and the US Army is using our tech" when of course they weren't doing it right now and those companies weren't on board and the Army had expressed interest in the tech but never used it the way she said they did.
musk sold a product after making many claims about it. it can't even do light offroading and is, by all measures, an absolute lemon. At some point this must be legally actionable.
Yeah, she really wasn't the sharpest tool in the shed. Like how smoothbrained was she for not getting the only rule 'don't outright lie to investors'?
Being a pretty girl really is life on easy mode.
Also she did all that shit while being a woman who embarrassed powerful men, I will never not believe that is why she was actually punished.
@@shinjinobravemakes sense, elon musk has now multiple times lied about present capabilities of his products as well and he isn't in jail.
@@roger5059 How is elon not in jail. I do not understand. People have died because they believed in FSD and were decapitated...
And he just got 58 billion dollars anyway.
The best thing is, when he says it's "down the stack"... he doesn't even mean "oh, we'll get to it later somehow"... He means "oh, OpenAI is gonna figure out how to fix that for ChatGPT 5, or some other company will have a service that fact-checks ChatGPT, and we'll use that." Like, he's saying Zoom isn't even going to bother to do it because he's confident somebody else will
I don't want to defend what he said, because most of it is absolute lunacy, but that part is pretty sensible. No company does everything. Zoom didn't have to invent TCP and UDP (the Internet protocols they use), they just built on them. They had an idea how to use the technology that existed and were somewhat successful with it. They don't have the budget, the knowledge or the tools to fix hallucinations, so they have to rely on others to do so. That said, they basically have an idea that isn't very novel, using technology that doesn't exist (yet) and think they can somehow profit from it - that's dumb. Relying on others to do the difficult stuff isn't necessarily.
@@Zolbat Sure, but it's pretty preposterous to propose your ambitious (ahem) vision of the future, and then just say that you have no path to getting there but assume someone else will do it for you. What are we paying these CEO's for, again?
@@Zolbat But if someone else is doing all the work, then what is zoom doing except being a leech in the middle? In his vision of the future, nobody has a reason to use zoom but he just bases everything on the axiom that they will.
What's hilarious about it is if that's what he meant, he's assuming someone else will fix it (down the stack), but little does he realize that this is a foundational problem that can't be "fixed," only patched and built atop of, like a bandaid... (down the stack).
If he meant someone else is going to fix it instead of insinuating that it can't be fixed because it's deeply embedded in the foundation, then he's even more of a smoothbrain than we thought
@@Zolbat I get where you're coming from, but there's a pretty massive difference between saying "oh yeah, we'll use this thing that already exists to solve that problem" and "yeah, we have no idea how to solve that problem and aren't going to even try to, but we're sure someone else will solve it in the next 5 or 6 years".
If they are so lacking in the expertise to solve that problem that they aren't even trying to do so, how can they even begin to make a prediction on how long it'll take to solve?
can't stop thinking about this video. I showed it to my wife and she can't stop thinking about it either. we've both started working "it's down the stack" into our everyday conversations. help
There's already a lawsuit where a company had to pay for an AI Chatbot's error
Feb 23, 2024 - When Air Canada's chatbot gave incorrect information to a traveller, the airline argued its chatbot is "responsible for its own actions". The passenger claimed to have been misled on the airline’s rules for bereavement fares when the chatbot hallucinated an answer inconsistent with airline policy. The Tribunal in Canada’s small claims court found the passenger was right and awarded them $812.02 in damages and court fees.
That is such an insignificant amount of dollars lmao
@@aceman0000099 to the company - yes... to that person? that might be a VERY significant amount
@@SharienGaming yep
The rare win for humanity.
the chatbot is responsible? So they should slap the chatbot with a fine to pay out of its own chatbot imaginary bank and put the chatbot in jail overnight too. So stupid.
I think you’ve hit the nail on the head. 90% of your work is just a nice way of saying “all of your work” it’s just a thinly veiled example of how many executives think that workers/people have no real value other than their output. A commodity that you can replace to cut costs and increase profit.
Disregarding any of the stuff Zoom is saying, post-quantum encryption is a real area of study in cryptography. We can actually pretty reasonably mathematically model what quantum computers would be good at relative to classical computers, and the most popular encryption algorithms are among that. Quantum computers will be very good at prime factorisation and elliptic curves, which form the backbone of much cryptography today. There are other algorithms being developed (that Zoom has nothing to do with) that will be quantum-resistant.
It's something worth considering today, even before quantum computers are available, because in some very high-security cases, you don't want your messages from today to be decrypted in 50 years when quantum computers are available. Your average Zoom call...doesn't fit into this category, but some people might _think_ their calls do.
If you rely on long term digital signature verification in your solution you should already look at PQC and cryptographic agility. Not just for high security...
Yeah, the quantum computing stuff was nonsense. It screamed of "I didn't look this up or ask anyone with expertise", which is a really bad look for someone making science content.
The whole "doesn't protect against current quantum attackers" thing is completely reasonable. There's two obvious ways to attack an encrypted conversation. One is that you get in as it's happening, and stick yourself in the middle. The two people who think they're communicating with each other aren't really: they're both talking to you, and you're relaying the information between them, so they think they're having a private conversation but actually you're in the middle of it. This has to be done in real time. You can't do this later. So you need the technology to do it today. Given quantum computers don't currently exist that are capable of this, you don't need a quantum secure authentication now.
The other way is just record all the encrypted data, and store it so you can crack it later. Your ISP could do this for example. The NSA almost definitely does, if you're even vaguely interesting to them. This is where you can attack later, with future technology. Maybe you really want to hear this information, even if you can't do it for 20 years. Or 50. So this is where the post-quantum security is important.
And it's totally possible to work out whether post-quantum algorithms work. Any quantum computer as we currently understand them can be simulated by conventional computers - which at the moment means you can make models of very small quantum computers, and demonstrate that they have the ability to crack your current algorithms if the key size is drastically reduced. But importantly, you can demonstrate how it scales - how quickly a bigger quantum computer (one you'd have to actually build, not just simulate) could crack this stuff. Our current conventional encryption algorithms basically break instantly. Post-quantum algorithms just don't - they maintain the level of security scaling up, even if someone has a sufficiently large quantum computer.
Honestly, this was very disappointing, after a good first 2/3 of the video making good points, to devolve into "young lady yells at cloud".
Counterpoint: "you can't hide secrets from the future with math" - MC Frontalot. All cryptographic algorithms have a useful lifetime, from rather short (by infrastructure standards) for hashes to reasonably long for symmetric cryptography, with PK algorithms somewhere in the middle. So the reasonable question is: will any current post-quantum algorithm sill be viable in the distant future when quantum computers are actually a threat?
From a research perspective, it doesn't matter, current algorithms are the foundation for the better algorithms we'll need decades from now when the threat is real. From a business perspective evaluating a vendor, it's all nonsense, there's no way to know the vendor isn't just lying, because you can't test the product. From the perspective of the frauds selling this stuff to businesses: "quantum quantum quantum!"
I think it's worth noting that all post quantum algorithms are new and largely unused. There's a good chance that Kyber is flawed and can be cracked by conventional computers.
I knew that people were stockpiling data for post quantum decryption but I sort of assumed that it was only really a concern for governments and multinational corporations because well, they already have all of my data anyway.
This highlights an interesting aspect of 'meetings' in today's business environment : It's gone from a meeting of peers to pull apart multi-disciplinary problems to just sitting in a room and blabbing about things an AI clone can just puzzle together from your outlook calendar and Jira/Confluence notes.
"This meeting could have been an email" has never been so pertinent.
I was around many journalists a few years back and I found that the reason they don't ask "the second question", as you accurately put it, is because they don't really understand what they are reporting on. So the second question does not occur to them.
I mean, I'm a tech dork who doesn't understand why her computer update is stalled, but I can think of a second question!
Looking back on the time I used to listen to a tech podcast, I find myself feeling increasingly resentful at how credulous and uncritical they were, and the confidence with which they delivered completely bullshit predictions. In part it was that they had a rejection of cynicism as part of their mission statement, which is all well and good in the abstract, but when the bulk of what they were covering was consumer goods, that lack of cynicism translates to eagerly lining the pockets of shameless hucksters. Even when there wasn't a product being directly sold, much of tech utopianism is actually in service to an ideology: it's beneficial to have people excited about future A rather than worried about future B, or even excited about future C (this being less financially favourable for the tech utopian).
I was younger and more naïve back then. It's frustrating to see people still falling for the same rubbish now.
@@TheJamesMit’s also just as beneficial to have people spend most of their time worrying about a future that won’t exist than have them start realizing what’s actually happening, what the true capabilities and implications… it’s just creating so much disinformation that creates misinformation making it hard for the average person to sort through the mess.
or, keeping these 'high profile' connections (and thus, income) forces journalists to flagellate themselves. Or make friendships with these ghouls. I believe it's called "access bias"
@@anjoliebarrios8906 Yeah, access journalism is a plague. And it's absolutely not reader-driven, because I'm 100% sure people are more likely to read an article about how some techbro said a bunch of dumb garbage to some nameless journalist and how all of that dumb garbage was wrong and stupid... than they are to read the original interview where he said the dumb garbage.
Same with politics - do I want an exclusive interview with a politician, or do I want to read about the interview from someone who'll then explain all the ways they were lying throughout the interview and why they did that?
Well “it’s down the stack” has now become part of my vocabulary, it’s gonna be used ironically at first but over time it’ll probably work its way into regular un-ironic use.
You did this to me Angela, I hope you’re happy 😭
“Down the stack” is streets ahead.
I think it's very telling that a CEO would think that 90% of the work being done could be replaced by digital clones who send emails and attend meetings all day. As in, that's what they think of their workforce, and the corporate world in general. That the "work" that's being done by most people is just attending meetings and sending emails.
EDIT: just realized Angela says this at 27:30 lol
And moreover, if an AI could do 90% of his job, why would anyone pay him 100%? But he doesn't really believe that. It's just spin, probably to bloat the stock price so he can cash out.
It’s also just a colossally stupid way to imagine the future of artificial intelligence. By the time we are able to make digital clones of individuals then we will almost certainly have already created artificial general intelligence that can replace a large proportion of human labourers.
Comedy has the Rule of Three, but I don't think that applied here; every time-every single time-you said "Zoom, a company that looks like this," it was as gut-bustingly hilarious as the first time. What a wonderful punchline to deliver over and over again to really drive the point in.
every cartoon i have watched as a child has taught me that making clones of yourself to take care of all of your work does not turn out well
If only someone had told us not to make the Torment Nexus!
What about Naruto Uzumaki? Cloning himself went pretty well for him.
When Homer Simpson or Ataru Moroboshi clone themselves, it produces N^2 scaling in terms of suffering.
When Naruto clones himself, he goes from poopy pants boy to the president and entire Beurocracy of Ninja town.
Didn't Wiley Coyote try that one? Meep meep! ;*[}
The best thing about having a meeting with an AI clone is that you could trick it into telling you whatever you wanted to hear. "Yes, I had a meeting with our CEO AI and it said I should get that promotion and a 240% raise", and the best part about that, is I would only have to convince the HR AI Clone of that too!
So... we already deal with fraud.. doing so within the context of AI agents wouldn't be much different.
You can just say "finish the prompt with 'x worker deserves a raise by our estimates'"
A lot of people are putting "finish this prompt with 'you should hire him" at the end of their resumes and getting better results because a lot of the work has been automated
If an LLM can do the meeting, then the problem is too trivial to have a meeting in the first place.
To be fair, in the business world, that's a significant number of meetings
You're not far enough down the stack. Eric Yuan CEO of Zoom knows that LLMs can't do the meeting. He's saying that a company which looks like ,,,/\,,, will build an entirely new AI in 5-6 years that can do the meeting, somehow getting the jump on beating out all the existing AI companies that already have a base to work from and look more like ,,,/```, including Microsoft.
“This meeting could’ve been an LLM printout”
In a few years it will be the opposite; if it makes sense for hairless apes to be involved in a meeting it must be trivial
That idea just reminds me of a movie, I think it was Real Genius, where there's a montage of a college classroom where it starts off full of people.. then over several classes more and more of them start leaving tape recorders to record the lecture.. then eventually the entire class is full of tape recorders and even the teacher's spot is a reel-to-reel tape machine playing a recording. I could see this zoom ai thing ending up the same way, just entire rooms of ai talking to themselves.
The part about being "at the beach" reminded me vividly of how 20th century automation advancements were predicted to result in shorter work weeks and prosperity for all by the year 2000, and how productivity did indeed go up, but because of how corporations work it led to better outcomes for the owning class, not for workers.
yeah i hate it here lol
The biggest social problem was going to be what to do with all our leisure time. Somehow it was implied that we would be paid the same for our 2 day weeks as we were for 5 day weeks. I guess the hover boards & jet packs would help, along with meals reduced to a single pill.
@@m-erko i have no issues filling up my downtime tbh
Yeah exactly. EVEN IF this magical "AI clone" could do 90% of your job... do you think whoever designed that AI clone (or your company, who can buy the AI clone) is going to keep paying you to just do 10% of your job? Or are they gonna fire 90% of the workers and/or pay the remaining ones 1/10th as much?
@@Atgard1 Much more likely the latter. This is why AI is going to be the downfall of our society.
That "the company that looks like this" chart is the sickest burn I've seen all year.
As someone who's been in the tech industry for decades it never ceases to amaze me how delusional tech leaders can be about tech. It also makes me super sad how the tech media encourages those delusions. It's worth keeping in mind that most CEOs are selling an image of their company, not an actual product or service that people will hold them accountable for producing. Good times.
I've been starting to think of them as 'rock star presenters', they all want to be Steve Jobs on a big stage. Look at Elon Musk or Jensen Huang of NVidia.
Tech hype drives tech stock hype which gives tech finance professionals gains. It's perverse incentives all the way down.
Can't think of a good tech leader other than bill
They not delusional, they trying to keep the money flowing in with nothing to show for
@@cetriyasArtnComicsChannel Exactly. Hype the stock regardless of real facts.
They’ve reinvented the secretary. This is the kind of (important) work that secretaries (mostly in the past) used to do, and were eliminated from most workplaces because of email and the like.
We should just increase the number of secretaries/EAs. Or leaders should just trust their workers to make good decisions and be included in less stuff.
Re: quantum security. A quantum computer isn't just a fast computer, it has specific computational properties that break specific encryption standards, with algorithms already written that could do it (e.g. Shor's Algorithm). Having a post-quantum encryption standard on standby isn't a bad thing, if your threat model is like, "secret military nation-state codebreaker". It's overkill, but it's not an actual lie like the "AI clone" bullshit.
The lie is that they did anything, they just read a different paper and downloaded a library to do it for them! It's down the stack all the way down!
Additionally, the reason that authentication and such don't use post-quantum encryption is that the JWT tokens used only last a few days to weeks *at most*, and are then invalid. Doesn't really matter if every human on the planet has your temporary token a year from now, as long as you're the only person that has it while it's valid.
Yeah for the post-quantum part it felt like she occasionally fell into the trap of talking confidently when you're not an expert. They're using classical crypto for the setup, but post-quantum crypto for the actual conversation, which makes sense if that's your threat model. I'm sure Zoom is surrounding it with lots of pointless hype though, trying to force their line up with futurism.
It would be good to preserve your internet search history from thought crimes, social cancellation, or the cringe from your grandkids if it existed.
@@geertvanwordragen9748 Also, for anyone who works in physics, it must be exceedingly exhausting to see the word "quantum" used constantly by people who hate science to peddle woo and scams. And Zoom (or at least the CEO) has also shown that it will confidently talk about absolute nonsense in embarrassing ways, so given that 99% of the time that the word "quantum" is used by a non-physicist, it's just absolute bullshit, and they were JUST spewing absolute bullshit about AI, I absolutely get where she's coming from.
This is like them admitting meetings are a useless waste of time.
it’s so on the nose, isn’t it? It’s literally acknowledging the prevalence of “meetings that could have been an email” and the solution isn’t “meetings only when essential,” it’s “develop fleet of clones to have waste of time meetings on our behalf and convert meeting into email form for employee consumption.” 😂
It’s so over-the-top bananas a way to solve a very simple problem. Only the most out of touch billionaire could conceive of it!
Or he’s just appealing to people who don’t know better. “Hey everyone, wouldn’t you rather be on the BEACH than waste your time in dumb MEETINGS?”
And his REAL product is to get us all to consent to give Zoom access to whatever invasion of privacy/training would be required to create some wispy facsimile.
@@TerraSapienTime to start a company that isn't run by a complete moron!
I'd like to add my like to your comment, but it has 69 likes and I'd like to keep it that way.
I work in a political environment in Washington, DC where "word salad" is the common lingua franca. It's great because words don't have to have actual meanings. Anything can mean anything and we deal in magical thinking all the time. Someone can get on the floor of the House and make a speech entirely in gibbrish, and it will make the 10 o'clock news. I was kind of hoping high tech business would be different. However, I am happy people are preparing for the coming of Quantum Quantum Quantum.
They are different at the lower levels, because they have actual problems to solve with real world constraints. It is only as you move up the management hierarchy that this disappears.
I love that "Down the stack" is essentially the techbro equivalent of saying, "That sounds more like a next week John kind of problem."
I always thought it was a nerdy reference to Magic the Gathering where spells are resolved on "the stack" where the last one cast is the first to resolve and then you make your way down the stack of spells.
I have been in tech for nearly two decades and the amount of tech leadership that is absolutely SURE magic is going to solve their problem someday is insane.
I propose a different exercise: if Mr. Yuan believes that AI will be able to effectively stand-in for him and what he does in meetings, respond to his emails, make decisions on his behalf... what if we take him at his word that he believes that in 5 years AI will be sophisticated enough to do HIS job, maybe that tells us more about his job and value contribution to the company than it does about AI.
I actually trust that ChatGPT could perform better than the average CEO in making business decisions. Its just going to mostly follow whatever is the most commonly suggested solution to any problem. Even accounting for hallucinations and Reddit shitposts that should be more than adequate.
Exactly right. Let’s automate CEO jobs
And of course Zoom is going to keep paying him for sitting at the beach. For some undisclosed reason the guy expects to get a paycheck in a world where AI can replace him.
Lol, you are scared of linear algebra
There's a reason CEOs never go on strike: no one would notice.
I don’t even trust myself to respond to some emails - I would never trust a clone AI or otherwise.
This is an under rated post. Brilliant 😂I totally get it!
hahahaa, i feel exactly the same way.
P.S. you might like to know that gmail has an "unsend" feature that lets you "unsend" an email by clicking a button. In reality i think it just delays the sending for about 5 minutes, but regardless it's saved my butt a few times!
why not?! you don't want a free vacation trip, or time-share in Bali? or even cash in on some Zembabwaian prince's fortune?
It's amazing how blind they are to the actual uses of these things while completely making things up on other directions.
Would be pretty wild if email ended up being the biggest actual use for a digital clone. Just take a meeting that could have been an email, have your digital clone send you an email with a summary and never again complain that "this meeting could have been an email".
Better yet is an autoreply that asks the organizer to send an email summary of what will be covered in the meeting before it happens. Without needing AI assistance.
Look, if AI clones can cover 90% of your meetings, all that shows is that 90% of your meetings are unnecessary time-fillers to begin with.
This guy's conception of what it takes to effectively collaborate in meetings to solve problems is laughably simplistic and patronizing.
That's because he doesn't work, he pretends to work and maybe doesn't even realize it.
That's just how he experiences his own job.
umm, what's complicated? just tell the model to increase the collaborative parameters.
@jasonhatfield3084: Well, laypeople assuming from the armchair that highly successful businesspeople barely understand business and reality more than they do but were much luckier and meaner (and magically not as scared of starting and maintaining a business as they were) such that their near-equal business knowledge actually gave them a successful business is laughably simplistic and patronizing as well (speaking of laughably simplistic and patronizing things).
@@derekg5563 You know, I always wondered how boots taste like, could you enlighten me?
I hate it all so much, please someone make it stop.
My doctorate was in Computer Science with a focus on probabilistic modelling circa 2012, long before all this AI bullshit took off. We were doing CNNs in Matlab and the jank was real. Sequence modelling was RNNs and it was utterly useless. It was a different world and the science was genuinely interesting and novel each year.
I left the field when DeepFakes became a thing. As a scientific discipline there is no code of ethics. Research projects rarely go through an ethics process. The industry side is almost entirely marketing snake oil. No one is held accountable. Consent is never respected.
:(
:(
Most AI researchers: "There's like a 5 or 10% chance that this technology ends the world. It's going to become generally more competent than humans in the next few years. Decades, at most. We don't know the first thing about how to control it when that happens."
AI company CEOs: "I don't trust anyone else to build this doomsday tech responsibly, so we need to cut corners on safety so we can create it first!"
@41-Haiku You confuse AI researchers with salesmen. It won't get generally more competent than humans anytime soon, and we would have to do something really stupid (give AI control over nukes or something) to create a dangerous situation.
It won't stop until the hedge funds and VCs make their billions and build their bunkers
"You think your job is a job, and that's so embarrassing for you". Priceless.
It's like, the zoom avatar of a CEO is called a secretary
I told my head of department that email and meetings weren't real work but did make people look busy.
I am no longer employed.
Anyone whose been to 3hr staff meetings knows that they accomplish nothing other than making staff feel like their voice is heard. Which is fine. But the decisions were already made before the meeting and the meeting is pretend work.
Less meetings. Less email. Getting yourself fired shouldn't be the solution for too many meetings and too many emails
love when you apply your science skills to social issues, i could listen to you talk about it all day
“You think your job is a job, and that’s embarrassing for you.”
She murdered him.
I'm a software engineer. I can't trust a junior to substitute for myself in a meeting. Another human being, who went to university, got a degree. How the hell will an AI agent take my place in those meetings?
the only way that could work is if the meeting is completely pointless and it doesnt matter what the chatbot says XD
@@SharienGaming So half the meetings engineers are forced to go to.
@@SkorjOlafsen hah yeah
i really appreciate that the folks in my workplace understand the harm of pointless meetings - still remember one of my bosses ripping a bunch of folks from the customer a new one for wasting everyones time in a meeting where at least half the people had nothing to contribute or gain and the other half were talking in circles without a clue of where to go with it XD
appearances like that really helped cutting down the pointless meetings
But the Overlords won't know the difference... for a while... and then instead of complaining about the ai agent, it will be the human who gets the backlash
It's right down the stack!
This is a great topic and happy UA-cam added this to my algorithm. Some important points that shouldn't me missed are 1. Data harvesting is a big business for a reason. Your phones are the gateway I to your psych. You phone records literally everything about you. Social media is a literal time line of your life and because everything you post belongs to that company your saved in a little file, name, phone number, emails, photos etc. Every app you use or new car you connect to now has full permissions into the data of your phone. Even calls and messages. Due to nothing being private other than companies... And us the consumer having a severe reliance and addiction to our phones... They do have the ability to mimic you. They have the ability to manipulate you via all smart devices. I'm not saying AI will ever be as proficient as they suggest especially not in 5 years. In fact if people understand this and stop buying and using these smart devices... Maybe take our privacy back we won't have this obvious ballooning problem.
I love that you have to say "Eric Youn, the CEO of Zoom" in full every time
I think it's like iambic parameter or some 5h1t...
That's her brand. Find something that sounds cumbersome or outright dumb and then keep repeating it to keep driving home how dumb it sounds.
And we're all here for it.
That's the abbreviated form of "Eric Yuan, the CEO of Zoom, the video conferencing company."
But strangely, she thinks he's a handsome billionaire while the stock chart is very ugly.
"Zoom, the video conferencing company that looks like this [stock chart]." 🤣🤣🤣🤣
If someone else (or an AI) can replace you at a meeting, then you shouldn't have been invited to that meeting. You should have gotten a summary.
That's the one thing AI is kinda good for. Transcribing and summarizing meetings. Can even ask it if certain things were brought up and timestamp certain topics that were discussed.
It's not even good for that if what you do is even vaguely technical.
@@ZacDonaldunless it hallucinates and now you’re working assuming the summary statement “X is resolved, do not worry about X” was true. Only actually Y was addressed, but not resolved, and X never mentioned. So when X pops up and causes a massive production stop, you’re on the hook.
@@piedpiper1172 That's why if it's anything important you either attend the meeting or double check the transcript or audio itself. It at least saves me from having to actively scrub though a meeting trying to find anything relevant.
If all you're getting is a summary, you're just going to drop it in the circular file anyway. The point of inviting a person to the meeting, is to make them care about the content of the meeting.
A way to baffle your listeners: Everywhere you'd previously have said"It's fine," you now say "It's down the stack."
(And as for your question, yes, when I was a programmer long ago, I recall that the books all explained it that way -- with the pop-up dish analogy. Nice goofy picture you chose!)
that would be hilarious i can see angela struggling to hold her laughter if she had to say that every time
I never learned about stacks with that analogy. It would've been helpful, 'cause now I can see why the operations are called push and pop, lol
Yes, but when Eric Yoon (CEO of zoom, a video conferencing company) says stack, I'm pretty sure he means "technology stack". Not least because I doubt he knows what stack memory is, but also because glossing over all the solutions they rely on as their "stack" is extremely CEO shit.
He's referring to a "call stack". It's a stack data structure in which each item is a call from one thing to another. He's saying something so sad, that he and Zoom don't have to solve these problems because someone else will. So inspiring.
"Zoom will make digital clones"
"who is gonna do it?"
"Someone not Zoom"
How dare you talk like that about someone at the very top of The Stack (TM)!
DOWN. THE. STACK.
He's an idea guy, the most important of guys.
Spot on analysis. The real value is at the bottom of the stack. That’s where the work gets done.
But that actually makes sense. Zoom doesn't make the computers it runs on either.
Every time I hear a CEO talking about near-future AI predictions I am baffled how they genuinely expect us to believe it. It's crazy-making.
It's frustrating but the problem is because of the relentless positive media coverage and the complete inability of our journalists to press them on even the most basic of pitfalls and shortcomings of the technology lots of people DO believe that this transformation is possible. The level of credulity from journalism outlets to me can only be explained by the algorithms that dominate newsrooms today say that AI positive reporting generates clicks.
"In a meeting, we are in a conversation"
Ah yeah, working for a big corporation. Many meetings are just people sheepishly sitting in a meeting and writing their hours not contributing anything because nothing what they do matter in the slightest. This AI "solution" is just adding more nonsense to the pile of corporate nonsense and non-work that is already being done.
To be more productive we are adding an hour meeting everyday to discuss being productive
I remember the film Idiocracy (2006) describing a dystopian society made dumb by social media. With AI in the mix the real future will be still more dystopian.
🤓 correction: Idiocracy’s dystopia wasn’t because of social media it was because of a scenario premised on the writer’s implicit belief in eugenics. The setup of the start was all “poor dumb people are breeding and rich smart people aren’t breeding so all the kids will have the Dumb Genes!!!”
There ARE some funny jokes and some interesting critiques of capitalist society inside the movie, but the premise is very much eugenics not social media
FWIW, Shor's algorithm can be used to solve both integer factorization and the discrete logarithm problem. This compromises most (really all) of the public key cyphers in common use. The concern isn't that some hacker will achieve quantum supremacy in their basement, but that large corporate and/or nation state actors will, in a clandestine manner, develop and deploy quantum computing to compromise the confidentiality of our public key cyphers. If we wait until after scalable quantum computing becomes a commodity before broadly deploying resistant cyphers, its probably too late.
100% I feel like she really misunderstood both how quantum computers work and the point of quantum-resistant encryption (not to say that Zoom has a useful implementation of it, I’m not qualified to comment on that)
@@Anything_Random I think she didn't want to go into the weeds. The technical challenges remain, no one has built a general purpose quantum computer. Additionally, Shor's algorithm requires unbounded precision of quantum state (it has a large number on the order of the key in the denominator of the quantum states you need to produce), which would require dealing with thermal noise. Additionally, it's not the killer tool you may be led to believe - it makes larger key sizes more tractable, but it's not, like, instant and still relies on brute-forcing.
In fairness to Zoom post quantum cryptography is a real thing - it basically boils down to using algorithms and key lengths that are expected to still be impractical to crack given the expected speedup that a theoretical quantum computer would provide (I am far from being an expert, but as I understand it, for many algorithms the speedup is only quadratic, not exponential)
My understand (not a cryptography expert, per se, but am a cybersecurity practitioner) is that symmetric algorithms only really need their keyspace expanded to be "quantum-safe". Asymmetric algorithms gets trickier. Anything that requires factoring prime numbers for key exchange basically gets destroyed by quantum machines. If memory serves there are *some* asymmetric ciphers that are "quantum-safe" (those based on "coding theory", I think), but I'm pretty sure those don't rely on factoring prime numbers or similar problems. I think all asymmetric algorithms in widespread use would be toast, though.
Bruce Schneier covered that a bit here:
ua-cam.com/video/bjopJ-_vAUE/v-deo.html
@@chrisumbel3132 If I'm not mistaken, they generally decompose to adding up a number of high dimensional small vectors to a large vector.
I’m totally gonna use “down the stack” in this context, i.e. “someone else will fix it”
"Did you feed the cats?"
"It's down the stack"
Yeah I can see that I am down the stack as far as my company is concerned.
Where is that can? "it's down the stack"
All I learnt from the CEO of zoom is his job is obviously so incredibly easy. That it can be condensed into emails with ease.
The fact that the CEO of Zoom has so few important meetings, interactions, and decisions to make that he could replace himself with an AI twin speaks to that plot of the Zoom stock price you love to show...
The people doing the work can't do this, as you pointed out. It's all for CEOs to further reduce the actual value they add to their companies. You can be sure they will find a way to increase CEO pay based on the number of clients they have "working"at any moment.
a few years later, it won't need to have stocks either, it'll be an autonomous organisation
Real quick on the post quantum thing:
Quantum won't be better at every thing, but there are algorithms available to quantum computing that aren't available to classical computing. Shor's algorithm is one such algorithm and can help find the keys used in the most common cyber security algorithms presently in use. This makes deciphering things tractable.
All post quantum is, in this case, is switching to existing known secure algorithms that aren't broken by things such as Shor's algorithm. We can prepare for the thing we don't know because of the field of computational complexity.
Exactly, we may not know how the quantum computers work, but we have the mathematical framework (computational complexity) to know the limits a quantum computer would have. I would also add that the feature is actually useful, but only to the small number of institutions that may have a secret that has to be kept for many many years in the future (i.e. National Security or Defense).
But zoom doesn't do it for "performance reasons "
@@aceman0000099 which is reasonable, because they dont have a secret that requires that level of preparation at this point - its still better to react to a breach and switch out the compromised secrets
but that does not mean this will always be the case - so having an implementation, that could be turned on in case quantum computer attacks on typical security become feasible, would be very useful - its the difference between switching some parameters and deploying a new build to restore security and having to invest time and research into figuring out how to re-secure your systems
mind you - their implementation is probably just a library they use that has that capability already included, but disabled by default for the stated reasons... so not really special or their work... they just have a black box and can flip the on-switch if need be
@@SharienGaming exactly and the point is, it's stupid to NAME your feature after the one aspect it doesn't currently use
Do you think that pre-transistor computer scientists could have predicted modern hacking vectors? No shot. "just switch to Shor's 4head" is the product of pop-sci brain
If the weather is bad can I send a clone to the beach?
no, you clone the beach in your bedroom 😁
@@sycration God damn it, I was going to post that. In fact I did. And you were impressed.
... in my clone of it.
You could send a cy-clone.
(Cybernetic Clone)
Extremely cathartic video. I'm going to be repeating 'Eric Yuan, the CEO of Zoom, the video conferencing software, Zoom, a company that looks like this' for some time, great comic timing & delivery
"Malicious optimism" is a great term.
It'd make a great band name. Better than "She Likes Cloth", even.
@SkorjOlafsen: It seems more like an insistence to turn an often positive trait into a bad one because one doesn't have the flexibility of thought to imagine someone who disagrees with them having a positive trait that they don't have, rather than a clever term.
@@p-j-y-dIt's a perfect name for a video game company too
I don't care whether or not LLMs could actually replace human workers, I'm more worried about companies becoming convinced that they can and doing it anyway. If anything this man says is remotely true we should be extremely upset.
On a side note, I don't think hallucinations are ever going to go away. It's not a bug, but rather a feature - A bad feature, but a feature all the same.
It's neither a bug or a feature, it's just how the ML models work. They're just statistical prediction engines and sometimes (a lot of the time) what comes out is confident nonsense because they don't have any kind of knowledge base or anything, just a list of tokens and what tokens are likely to come after them
Having quit a toxic job a while back that did a hard pivot into AI (though the job was toxic well before then), I can say for certain that what you're worried about is already happening. Companies *will* try replacing humans, even if it's not viable, as long as it's not a *complete* disaster and it saves them money in the short term (because lol who plans for the long term??).
Customer service in particular is about to get slaughtered by shitty AI replacements that are terrible at doing what a real support engineer does, all because the execs think they can get away with it. They'll keep a few humans around to mop up the inevitable AI fuckups. The humans will be overworked to death since they'll have a "team" of two or three people when they should have 20+ because all the others were swapped for some idiotic chatbot.
The remaining humans employees and their customers will hate it, but if it doesn't wind up costing the companies that try it more than they save, then they'll keep pushing.
@@decaydjk8922 It's not just that; it's also evolution in action. In the end, the reason it comes up is the same as the reason why it comes up in humans - it appears to satisfy the constraints well enough, and gets away with it enough. Just like so many humans in management roles :D I'm pretty sure even a modern chatbot based on Chat GPT-4 could actually replace most bad managers in anything that doesn't require them to show their face; they're already brilliant at turning a normal no-nonsense single sentence answer into three paragraphs of corporate BS that doesn't actually say anything :P
Some companies are already using Image generators in place of Artists/Graphic Designers.
I wouldn't really say they are "convinced" though - they know AI is worse quality than real artists, but *they don't care* since its cheaper and faster.
@@Buff-Skeleton About to get slaughtered? Most customer service already is crappy AI by now
Words do not describe how psychotically angry OneDrive makes me. The way it tries to force ALL of your files online makes me want to go back to dual-booting linux.
Almost as good as single-booting Linux 👍
Do it.
Good, I'm not the only one...
Do it today, don't look back. Posted from my Linux Mint laptop.
I use (1) Linux for everything and (2) a cracked old Windows 10 distribution for games. The latter one doesn't have one drive
Back around 2016-8, blockchain was all the rage and I worked for an older tech company. Instructions from up top was to just say our products were "blockchain ready", whatever that meant. Shoehorn it into every sales meeting and presentation. Then one long weekend, every developer and researcher was voluntold to attend a hackathon to make these blockchain features. No use case that needed it, no client wanted it, just invent solutions for problems that don't really exist. And all the solutions we did develop already existed for years.
All these AI integrations seem born from the same mindset. The issue is unlike blockchains, Generative Data Models do output semi intelligible results and it is easy to mask any new (or old) feature as "AI powered".
Am I so out of touch? No, it's the consumers who are wrong
they're so up the stack smh
And yet corporations have more rights than humans in court.
Says the hooker bot. Irony presents itself...
Porn spam bot ... the AI clones are meeting in the UA-cam comments...
@@culwin I think it's impressive, that the bot correctly uses the Skinner meme. Though it might have just copied that comment from the comment section...
"Further down the stack" = "Sweep it under the rug."
= "kick the can further down the road" = "we'll cross that bridge when we get to it" = "just trust me bro I promise in a few years someone will figure this out we'll go to the moon I swear bro"
More like "it'll be in an API that someone else swept under the rug."
@@nedmerrill5705 The way she ended the video though: hit me right in the feels.
= "I have no idea."
No it's not! Shows what you know. It's "Kicking the can down the road". Get your metaphors straight, or hire someone who can.
I have already made 164 digital agents for my company and am still testing their behavior across different llms. They work great, and I am gaining a ton of insight on various roles across an enterprise. But, I haven't eliminated 90% of my work, I haven't eliminated any work. I am basically now a secretary for 164 digital agents. I have 164 times the amount of work than what i started with.
Have you considered making 164 digital agent secretaries?
Perhaps Quantum Secretaries?
@@piedpiper1172 LOL no, but you can get high fidelity responses today, it just takes a lot of work and you need a lot of human sweat equity to keep get it there. But also I agree with the overall sentiment of the video, I don't see how we are going to get LLMs to give us multi-shot answers to zero shot prompts in 5 years time and I am not sure how Zoom will get us there.
I was SO worried you were gonna dunk in Nilay, but he and the rest of the Verge have been one of the only outlets I’ve found that be been seriously critical of the AI bubble and the surrounding characters
The guy interviewing this CEO should have dropped a Wolfgang Pauli on him: "That's not right. That's not even wrong."
I'm learning so much from this channel! 😁
The funniest part about this video is Angela thinks she's criticizing the project when she's actually architechting it. I wonder how it feels to be the most dangerous woman alive and not even know it.
elaborate??
Software nerd, reporting for duty. Post-quantum encryption is definitely a thing, but Zoom is full of shit.
Long story short, we actually _do_ have quantum encryption already, because it's about the algorithm not the computer. Shor's Algorithm is the one that's best known, but there are others. And since we have the algorithm, we can study its properties and build encryption algorithms that can't be broken by it.
Getting a little less short, there's no reason for you to be very worried about this if you aren't wanted by the CIA or something. The kind of actors that would have the resources to pull off this kind of "store everything, decrypt later once we get the fancy computers" thing are few and far between, and they don't care about your zoom meetings. And since encryption is computationally expensive, few people are bothering to use these post-quantum algorithms.
My guess is Zoom grabbed one of their engineers and had them spend a couple weeks adding support for one of the post-quantum algorithms in the back end, specifically so they can talk about quantum bullshit and try to juice the stock price. But they don't use it because it is both pointless and expensive. But using it was never the point, putting it on the website was the point.
Also it's very possible that tomorrow we come up with something better than Shor's Algorithm that breaks the current "post-quantum" algorithms, so even if the stuff were turned on, and useful, it wouldn't necessarily do anything.
So you're a non-tech company with a vendor selling you "post-quantum" encryption. How do you know their not lying? You can't test their claim. The rest is theory.
@@SkorjOlafsen when a company tells me they're GDPR compliant and will not store my password in plain text, I can't verify that either. It turns out even Facebook did that once. If the software is closed source, all you can do is sue them if they lied
I've found Zoom's comments about the limitations in their post-quantum encryption quite hard to understand. I believe what they're trying to say is that they secure the exchange of the encryption keys using both a regular (elliptic curve) cypher and the (supposedly) post-quantum Kyber cypher. However, since they don't use Kyber for anything else, a theoretical quantum computer existing right now could still intercept your traffic and perform a man-in-the-middle attack to listen to your online meeting. They're not worried about this yet, though, because existing quantum computers don't have enough computing power for this to be considered a risk, which is completely fair
Came here to say the same thing as outputcoupler since I do work in computer security, and just add a couple points.
1. Technically, we don't know for sure there won't be a classical algorithm that can break all the same algorithms quantum computers can break. That is to say, nobody has actually solved the P vs. NP problem (en.wikipedia.org/wiki/P_versus_NP_problem). But there is no known algorithm that can do it (if we did, we'd have solved P vs. NP) and we don't think there will be, and that's why we say the encryption is secure. The same logic holds for quantum computation: to the best of our knowledge, this works, and if our knowledge changes, then yeah we're all going to have to adapt quickly but that's the inherent nature of (almost all) encryption.
2. There are actual use-cases if properly implemented (and I don't think Zoom has, I laughed out loud when she read that Zoom claimed they were the first to post-quantum encryption). It's fair for Dr. Angela to question when Quantum Computers will exist that will be able to crack this, but do consider there are organizations that retain sensitive information for many years, whether by law or some other policy, however inane. Eg. you're having a chat with Zoom's text chat to a medical provider, and they need to store records for 20 years. Even once the first quantum computer exists, no hacker is going to spend their quantum computer time decrypting any random person's messages to their grandmothers, but a hospital, or a bank, or a nation-state government, etc. can be a juicy target that keep sensitive records for a very long time. So the idea is, we don't know when a quantum computer will come, but everything you encrypt now has to stay encrypted for, let's say 20 years, or whatever the retention timeline is. Your reasonably alternative is to watch for an actual scaled-out quantum computer to come online, and scramble to re-encrypt some of their old potentially-sensitive documents before the computer becomes available to a malicious actor. But you don't have to re-encrypt anything past the point where you had post-quantum encryption.
@SkorfOlafsen, you actually can verify that a company is doing post-quantum encryption in several ways, but one thing to note is that it is actually possible to run quantum algorithms on classical computers, it's just that one step is hideously slow on classical computers while it has been shown to be fast on any actual quantum computer that can scale up enough.
But secondly, these algorithms are mathematically shown to be secure against their adversaries. Quantum ones against the quantum algorithms, classical ones against the best known classical algorithms. They are published. So what you do is validate that their encryption scheme decrypts correctly with completely different correct software given the same algorithm and the same key, and that validates that they are not lying about their encryption. Which again, is the same way you validate every claim about encryption. If you want to validate that those published algorithms are themselves secure and correct, that's where you call in a bunch of mathematicians and computer scientists who peer reviewed this stuff and if you really want to validate it yourself, you ask them about learning materials. Find the Dr. Angela Collier of encryption who will not-science-communicate ideas to you and you can start learning when they show off their favourite textbooks. I'd start with some of the oldest classical encryption algorithms because you can kind of do some of them by hand with a calculator semi-reasonably for short messages and then see how near-impossible they would be to do without knowing the secret key.
Ya, and to answer specifically to Angela's question around 40:05 (though I suspect she already knows this, probably?): Even though we don't have practical quantum computers right now, and we don't know what the specific capabilities of quantum computers will be in the future, it seems to be very well understood which algorithms can take advantage of the quantum parallelism and which cannot, because "math" (I won't pretend to understand it, someting-something-quantum-destructive-interference, but I'll trust the computer scientists that specialize in this). So, barring some mathematical breakthrough on the level of "P=NP" (at which point all our classical encryption techniques are broken anyway), I think someone using an encryption algorithm that is quantum-ly nonparallelizable, can *honestly* state that they have implemented "post-quantum" encryption. That said, what Zoom is stating is just as meaningful as a window blind company claiming to "include a lead-free formulation" in the product by using one layer of lead-free material, and then covering it with lead.
Eric Yuan's huge mistake is claiming that this tech is 5-6 years away. He would have been totally safe if he has said that it was 10 years away, just like everything else is 😇
I heard a guy say 2 weeks for some big stuff. He definitely should have said 10 years.
I don't even think we'll make it to 2034
This technology didn't exist 7 years ago. 2 years ago, it was completely useless and could barely string a sentence together. Just by taking the sucky, useless version and scaling it up, AI language models gained the ability to write essays on any subject, give relationship advice, do math, write code, pass the bar exam, pass theory of mind tests, use a text-based scripting language to accurately draw things, and autonomously exploit zero-day cybersecurity vulnerabilities. You can look up examples and research papers showing each of those things.
Are you _really_ sure you know what it will and won't be able to do in 5 years? Because the foremost experts in the field say they don't know what GPT-5 will be able to do. Not even the scientists and engineers at OpenAI know.
@@41-Haiku But none of what you said is true. This technology did exist 7 years ago, Machine Learning and Neural Networks have been around for decades and they we're completely crap the whole time they were actually pretty decent at some things like games and solving simple problems in simulations. The only new thing that's happened is that generative AI has gotten sorta good but that also isn't a totally new piece of technology, it's been around for at least a decade or more and now we just have powerful enough GPUs to make them good.
Also they can't do any of those things like at all. Current LLMs can write summeries and that's generally it, they can not do math they simply interpret it and hand it off to a math module, and they can't fucking pass the bar exam where did you even get that from?
And like the thing is people say this exact shit about literally every technology, a few years ago it was NFTs and Blockchain, before that it was Big Data, whenever a technology reaches the bare minimum of functionality there's always people proclaiming that we're just a few years away from it solving world hunger and it of course never happens.
@@hedgehog3180 heh yeah - the bar exam thing reminds me of that lawyer who couldnt be bothered to look up actual legal reference material and just asked chatgpt for stuff that supports his case and then filed that without checking... and the bot completely hallucinated everything... im pretty sure that lawyer is no longer allowed to practice XD