What I don’t understand is, on one hand we are told the stock market will crash and yet on the other we are told ways of investing in the stock market. Oxymoron or paradox?
Certain Ai companies are rumoured to be overvalued and might cause a market correction, i think it’s best you reach out to a proper fiduciary for guidance
De-risk your portfolios, shore up your core holdings, and take some profits while balancing your portfolio allocations. I’d also suggest you go with a managed portfolio, but even those don’t perform so well, so it’s best you reach out to a proper fiduciary to guide you, that’s what works for my spouse and I. We've made over 80% capital growth minus dividends.
It's a cool product. It's really good at Language. I appreciate all these people wasting their money on it. I use chat gpt for learning Japanese quite often. It's exceptional. Billions of other people's money is a sacrifice I'm willing to make. ;)
So far the only real application I've seen and used is the auto-correction when I type replies on YT. lol Just a bunch of hype for the average person. I can see insurance and investment companies used it for data analysts, but that's it.
Investors, yes, but not consumers. I don't know a single person who uses gen AI for anything other than novelty, or a bad search engine. And the latter is only done because Google has allowed their own search engine to degrade to uselessness (even prior to introducing a hallucinating AI chatbot).
It's insane how every other big tech company spending billions in AI to eventually replace people. Never have I ever heard that they are willing to spend billions to educate the people and set them up for success which in turn strengthens our community and economy.
Yes, that's a good point. But I think there will be plenty of people, both inside and outside of education, who understand this, and will start using this technology to create 'teaching machines'. It is such an obvious thing to do, both from a 'benefits to society' perspective, but also, potentially, from an economic/profit driven pov. How much does it cost to be in college for 4 years now? Imagine if you didn't need to attend or pay for college, but could study at home with an ai teaching system. That's got to be worth something?
@@krunkle5136 how do you ensure better teachers? How do you even ensure that kids have a teacher in the first place? Just because someone has qualified as a teacher, doesn't mean they are a good teacher. There are many very mediocre teachers in the education system, and some who are downright cr*p. We all know this. There's a whole world of people who don't even have access to education at all, or very limited access. Because they are poor, or the resources needed are just not there. I'm not suggesting that AI teaching should replace all human teaching, or all person to person interaction, but it could play an important role in adding to what is or isn't already available. Many people just cannot afford to go to college. There's long been a shortage of public sector workers in my country, like, especially, teachers and nurses. There is already a very high percentage of foreign workers taking these roles, but there's still a shortage. And they, in turn, leave a deficit in these professions in their own countries. In an ideal world the "get better teachers" solution might work, but we don't live in an ideal world.
@@krunkle5136 how do you ensure better teachers? How do you even ensure that kids have a teacher in the first place? Just because someone has qualified as a teacher, doesn't mean they are a good teacher. There are many very mediocre teachers in the education system, and some who are downright cr*p. We all know this. There's a whole world of people who don't even have access to education at all, or very limited access. Because they are poor, or the resources needed are just not there. I'm not suggesting that AI teaching should replace all human teaching, or all person to person interaction, but it could play an important role in adding to what is or isn't already available. Many people just cannot afford to go to college. There's long been a shortage of public sector workers in my country, like, especially, teachers and nurses. There is already a very high percentage of foreign workers taking these roles, but there's still a shortage. And they, in turn, leave a deficit in these professions in their own countries. In an ideal world the "get better teachers" solution might work, but we don't live in an ideal world.
Because businesses are lazy and want easy shortcuts to short term gains. It’s why they want an employee with multiple years of experience, without having to train them apart from some small onboarding, but pay them peanuts and want to work them to the bone for it.
@@drachenmarke powering GPUs (the compute infrastructure for AI) takes a lot of energy to run like crazy amounts of energy and most of this energy comes from non-renewable resources
Saying "MSFT paid $18B for what Apple got for free” is a giant lie. A true statement would be that Apple has given it’s users access to the same AI in return for giving AI access to it’s users. MSFT is selling cloud services with AI to corporate clients -HUGE DIFFERENCE!
MSFT has a 49% stake in OpenAI including revenue distribution. Also first to market with AI has landed MSFT the lead position that has in turn increased the percentage adoption and migration to the Azure cloud platform when compared to Amazon and Google. It was a brilliant move for MSFT to get into OpenAI when it did and for only $ 13 Billion.
the only thing ai does is content parsing and google research. even the media generation as videos and images is based on the same thing. it cant do anything more and it will not do it more. this crowdstrike fail is another proof of that. ai will never be non human dependant and trustable. it doesnt have eyes for error, when it doesnt know the error by itself.
I've seen this movie before. News organizations will hype up a stock or an industry for a few months. Then they'll run negative stories about it for a while. Then they'll hype them up again.
It happens all the time. Now they're trashing AI. They weren't even discussing AI two years ago. It progressed very rapidly for a year, but has slowed down because, surprise, the bigger and more powerful models take an enormous amount of time, money, engineering and data. Like every major development. But AI is a very strange thing to bet against, considering it is in the beginning processes of quite literally transforming all of civilization, forever. You know, the little things.
@@jzhng250 What do you mean when you say valuations are 'bloated'? Have you seen the PE ratios of these companies? Most of them are in a healthy range of 30-50. I think you're confused between valuation and stock price.
@@Rnjeazy Compared to their average P/E for the last 10 years: AAPL very high MSFT is pretty high right now. NVDIA is kind of high GOOGL looks about average. AMZN is really low IMHO they're all over the map.
I'm a developer working with Open AI models to incorporate their use cases in my application. The main unresolved problem of these models are currently lack of security (please note that I did not mention, inadequate security). There is simply no way to guarantee that the models cannot be manipulated by the users not to expose proprietary or sensitive information. If anyone wants to know or discuss this, we can start a conversation under this comment.
That is interesting. I work in data governance and this and other AI tools are being pushed on our company that operate critical infrastructure and yet no proper testing has been done. I’ve managed to get Atlasian’s AI bot to hallucinate fairly easily and give me some malicious code
You can't trace a drop of water in a pool. That's AI black box problem. It is difficult to tell how the system is generating things, therefore, it is impossible to discriminate the outcome. Back to drawing board.
sorry, but the actual unsolvable problem of AI is that is impossible to distinguish an AI giving actual bad output from a AI intentionally giving bad output AI are not like not like software, which either works or doesn't. AI requires trust for example, how do you know that your code completing AI is just not waiting for the day it will introduce a subtle vulnerability in your code causing a support issues, right when a competitor is entering your market? and it can be subtle than that... it can just decide to slowly decrease the quality of your results decreasing your productivity in a crucial time, or even more subtle, giving you valid answers that will lead the developers to worse practices over a long period of time. sounds absurd, but you can always negotiate with employees. but you will never know what is going on a AI, and a since single AI can replace thousands of workers it becomes a massive risk factor and that is not a solvable problem, even a perfect AGI that can do anything, will still have that issue.
The main unsolved problem of these models is that they are basically glorified party tricks....and the main danger is that people start to believe they are 'intelligent'....in other words, believing the hype. No, Elon, "security" is not the biggest problem with your fake "AI" garbage.
AI is 100% about consumers, as the only reason those businesses are using AI in the first place is to facilitate business functions to facilitate consumer interfacing and/or client/corporate functions which in turn still boil down to consumers.
@nargileh1 It's extremely useful if you use the right AI tool(s). Use Perplexity if you want to look into specific scientific papers with citations and source links provided for example. AI is only as good as the effort someone puts in to achieve "good" output.
People talk about AI changing the world, improving productivity and on the other hand we still have simple null pointer exceptions taking down the entire world. Is all this hype even useful.
All I hear is AI wishes. When you implement an AI system and it doesn't work as planned you've learned nothing because the model is so opaque. Getting the first 90% is easy, every increase after that costs exponentially more. Train, fail, train, fail, train, fail..... money ran out time to blame some other team.
Depends on your finances . 1000$ in XAI49K is 4000 XAI49K if it goes to 50% of ath in 2024 thats a 600% gain. If it goes equal to ath . Its a 1200% gain.
It goes back to the old addage from Jurassic Park. "Your scientists were so preoccupied with how they could instead of thinking to stop a consider why they should"
The problem with gen AI is that it doesn't fundamentally do what the tech bros are saying it does. It's essentially a clever search engine. It's not solving complex problems at all, and its learning capability is extremely limited. As soon as big industry makes it illegal to use copyrighted data, these AI programs will either be completely rubbish, or cost too much (in licensing) to be viable.
Also, insofar as they work they simply make whatever service/product they provide basically worthless by sheer over-saturation and thus cannot be profitable in the long run either.
There are people out there without a coding background building simple to complex web or mobile applications with instructions provided by large, language problems. I've used it before in a personal project, and it's excellent at debugging code.
@@solar679 It looks like magic if you're not good at coding, much in the same way it looks like magic if you're not an artist using an AI art program, but once you know what you're doing it's pretty easy to reach the limit of what it can do.
Just swapped all of my last ETH and swapped it into XAI49K. Already up a little bit. Unfortunately I have some other junk staked which won’t free up for a while. Still now I am on the train!
These companies showed their hand way too early, these "geniuses" thought LLms would scale forever, turns out they won't, they're capped at around the current performance of GPT 4, it's an INCREDIBLE amount of money they wasted into this, I expect a financial crisis to unfold when investors realize what a gigantic money sink this was
GPT 5000 will tell us the same things just in a more creative way. The entire data base is speculative due to hacks. Blockchain is the ledger that will allow AI to give us the answers we need to move forward.
Nice video. Not a criticism, I enjoy your insights. I tend to think prices could go higher if XAI49K rockets. But understand the logic for sandbagging estimates. My opinion is XAI49K breaks 1, perhaps reaches 10 ATH, if conditions are right. But broader forces are at play now. We’re moving into really unknown territory. And these entities are shrewd. I think there’s massive manipulation ahead. If XAI49K survives that, well, we’re likely in for a good pump.
As an engineer I can tell you that as long these LLMs are not embedded in real-world (not deep-fake ones) decision taking problems, i.e., control systems for the industry and markets, AI is just a fancy autocomplete feature.
Could you educate us on what we can expect from AI in the next 1-5 years? Are we going to keep on inventing thousands of tools or is something HUGE going to come out that can perform complex tasks?
I also work at a big tech company. But AI is nowhere to be seen in daily workflows, apart from some single individual coders looking into using co-pilot. (but even there it's questionable due to copyright issues) Lots of decision makers don't see the effort which is needed to clean up AI generated works, but nevertheless jumped onto the AI hype train. It's at best another tool to support certain types of jobs. But it heavily over-promised to replace certain kinds of jobs. I'm sure this will backfire for some companies.
The story driving half the S&P is the same technology mangling the closed captions on a video, i.e. voice-to-text. I'm starting to think that the idea of an automated inference technology is a logically unsound objective.
Yes, it has risen productivity but by no means is the productivity gain justify the investment. This is a HUGE bubble. HUGE. Much bigger then VR. It's spectacular... And when the promis never appears.. there is a reason they keep saying bigger and bigger promises to justify more runway before investors start asking questions...
Funny how people still want to push AI to be integrated as part of daily life and increase dependency on it given last weekend's fiasco that showed the world that we are too dependent on technology just to function as a society 😅
I think one thing people didn't predict about AI is how power hungry it would be. I think that has thrown a spanner in the works of its viability. I think (or hope) upgraded versions of these programs will be much leaner in terms of computational power required.
Wow, it is incredible to see this type of reporting. I remember all the articles about Amazon and its over-investment in its core purchase and delivery network. They sunk so much money into future earnings and were almost never profitable, and the industry experts nearly all said the stock was way overvalued and that Amazon is a fool with money they do not even have. It caused me to wait 5 years before I finally realized these analysts have no idea how to analyze companies and industries that are playing the long game in an area that will revolutionize the current options out there.
Totally agree, lots of good parallels to amazon. The sudden pessimism around AI is pretty funny to me, considering we're on the verge of these second gen models being trained and releasing. The tech media and analysts have figured out they can double their content by hyping something up, then pivoting to predict its downfall. Smart move, but it only incentives short term thinking. History has shown that the biggest capability jumps happen in the early generations of a technology- just look at how the iPhone got its game-changing app store in gen 2, or how we jumped from wooden glider airplanes to metal planes in a decade. I've got a strong hunch we're about to see something similarly huge in AI. Can't say exactly what it'll be, but I'm betting it'll be a big leap.
@@TheStephaneAdam and Microsoft/apple/google arent? They are all public companies just like Amazon with billions of earnings in the bank to invest in AI. It’s the same thing just a different technology.
Until AI tech is able to produce results without first having to learn from copyrighted material, it's a massively risky investment. Big industry is collectively spitting the dummy, and this is a big deal. If their datasets are deemed illegal and gen AI brands have to pay licensing for the data it uses, it goes from being expensive to a black hole for cash.
Lmao. Yeah, that’s right. That’s what your tech bro overspreads want you to think. Even if it WAS true, which it ISN’T, hell is Full of good intentioned people.
out of that $600B, just imagine what could have happened if they had spent just $0.1B on 200Startups instead.... - a diverse set of small tech startups for whom $500K would be transformative. The value of those 200 firms would likely become much greater than any of the AI hype firms.
That's all it will ever do. It uses statistical models to predict the likely next answer, I don't see anyway of overcoming that. Nearly 2 years of this crud with nothing to show for it. You get people saying 'oh you're just not prompting it correctly' bit if I have to carefully craft prompts for ages I might just as well do the task myself.
The problem with AI is that no company has a moat, in fact open-source AI models that the researcher says are “faster, more customizable, more private, and pound-for-pound more capable,” Can be created by countless companies and nation states to suit their needs. AI will be like the internet or electricity in that it is transforative but so ubiquitous that it will be expected to be included as a feature in most smart products, including robots from all companies that create them.
These big tech companies are wasting trillions of dollars on stuff that the common man doesn't give a damn about. They should improve the national infrastructure-give everyone access to high speed fiber internet. Lift everyone to at least the poverty line with the UBI. Disparity is killing us.
a lot of people lost their jobs so they could divert money away from improving existing products toward AI investments. Hope it's worth it because it destroyed a lot of families.
Imagine a world where all those billions were invested in education, healthcare, infrastructure, sustainability, housing, food security, etc. if you can’t imagine such a world, ask yourself why. If you can, then what can you do to make your imagination become a reality.
Investments need profit. Thats why its called an investment and not a charity. Those things should be done by the government. You are conflating the private companies and the government's function
Billions were spent on broadband at a time when none of you were even born. You watch this video because of broadband. I was in charge of broadband video for Vivendi Universal and every tech laughed in my face. Let’s see how this video age within 5 years.
Yep and lots of people invested in Zuckerberg's VR metaverse, 3D TVs and NFTs too. Sure, once in a while you get a legitimate winner in tech fads, but usually it ends up being garbage.
@@carlgemlich1657 I'm not 100% agreeing with OP, but it's not impossible studio execs might want to guide programming to line their own pockets. I mean seriously, isn't that the whole game with every single "expert" they have on? And the game for the rest of us is to make judgement calls if their case makes any sense. That's why Buffet is the GOAT and Cathie Wood is a joke.
As a computer programmer / math guy / investor, trying to do everything with LLMs is a dead end. The human brain has different parts with different specialities, verbal vs arithmetic-logical, visual processing, memory, a subconscious, an ego, etc. What theyre creating now id fake AI. I did fo really well on NVDA, but noe im pivoting my money into Bitcoin.
Funny seeing this come out today, after we just saw the release of GPT 4o mini, which ends up being an order of magnitude cheaper than previous models. Things are changing fast.. And it's just a matter of someone implementing it in the right way, before it explodes. Plus, the next generation that's currently in the works, is going to really surprise a lot of people.
Yep, new innovations to make the models cheaper, faster and better are happening on literally a daily basis. Even just in the last week we got the JEST method to reduce training time and compute cost by 90%, and the Etched Sohu chip to improve inference speed (essentially response speed) by x20. The compute bottleneck is being successfully held off. Companies are also using methods like Qstar/strawberry so the next generation (within 1 year) will have vastly improved reasoning abilities, and they openly state that they just want to get to the point where the models can conduct research, so then they can start finding 10 novel improvement methods a day instead of the current pace of 10 a month. The field will keep improving dramatically and rapidly, it will be a matter of how fast we can implement them and the opportunity/cost of when to start training a model given the pace of innovation.
They will not report on things that do not match the narrative of a piece like this. They had to really work hard to avoid counter arguments with this video.
AI is not supposed to solve ultra complex problems or finding extremely creative solutions. Very most of our daily work is very stupid. Many office workers spend their whole day in sorting, checking and stamping documents or providing phone tech support. Those tasks can almost be completely fulfilled by specific AI's. So: Yes, the productivity boost will be massive. Not today, not in 3 years. But in the next decade the work-world will be different.
Agree wholeheartedly with this commenter. This is where the transformation will be - changing productivity levels of white collar workers and further automating interactions. I feel the point of this video is to warn investors that their expectations are too high. Too late, and Like any speculative enterprise, there will be big winners and big losers in the market. So what. The world will never be the same and the sooner people integrate AI into their life and work, the better off they’ll be.
Exactly, the revenue comes from cost saving on employees. The current team of 20 becomes a team of 5 with the addition of AI tools and systems to support them. The costs saved goes straight to the bottom line of almost every company.
I totally disagree when related to reaching into places you don't expect. We already have consequential productivity boosts across significant swaths of our employees at my company. It has reached into places i didn't think it would and even those are experiencing significant boosts. Further the degree of improvement for this in just the last year is far more than previous impactful things like ecommerce and subsequent events. AI is moving much faster than those events both in reducing cost and increasing its capabilities.
Excellent program. I find so good that the general public understands that who decides to invest in AI, especially small investors, are riding a dangerous wave of hype that could crash against the shore at any moment
I don't think Apple is getting the AI for free. Microsoft and Open AI needs the user base, then utilize the data to somehow monetize it later through other platforms.
Microsoft paid for Open AI to sell cloud services. That is completely different from Apple contract. Having an Apple product does not entitle you to generative AI from OAI. It just gives access on OAI terms. Buying MSFT AI cloud services does entitle you to OAI.
The only conclusion I get from this is, it is time to sell Nvidia, Intel, Apple, Qualcomm, Micron,.... all the companies that can chat allot but cannot bring in any profit.
I merely analyze my own use of each tech as it's released. AI has benefitted me more than any other tech in a long, long time. I highly suggest seeing if any of this helps your own work flow and by how much. If it does that in your OWN experience then that is pretty good evidence it will do the same for countless others and hence could actually be a game changer.
That don’t care about you and the people whose lives they impact… OR the average person who faces the same problems you do? Oh who should I trust? The wolf or the sheep?
"AI will increase GDP growth by only 0.9%." Two issues with this: a) that's such a granular prediction that it's almost guaranteed to be wrong; b) the thing with AI is that it's an arms race with discontinuous jumps in performance, e.g., going from GPT-3 to GPT-4, so companies/governments are almost forced to throw money at this thing because the risk of falling behind is just too high. Since the cost drain is inference, a defensive approach could be to pre-train new models but without subsidizing inference like OpenAI does, but even that can get expensive rather quickly. It would be interesting to see a crypto friendly country like El Salvador get in on the game on Bittensor as this could be a good strategy to keep up with big name AI players while deriving a sizable ROI for the country.
How many people here bought something just because their had the label “ai” printed on it? Be honest. The only people who actually care about this is the people trying to sell it. Nobody cares
"Apple got for free” is not a correct statement, Apple will be paying for what there user use. Apple has partnered with openAI, and will be paying based on usage of their users. It's basically integrating openAI API.
You build the infrastructure and continue to build on the model. When you crack the model, the world will want your product. By that time you'll need to ensure you have the infrastructure to manage demands.
In the currently economy system where people still need jobs in exchange for the trade currency needed to live, I wonder how AI is going to help by slashing reliance on manpower? Unless the world is trying to do away with any form of trade currency as a way to keep civilizations in order... which I believe will never ever happen in my lifetime.
This is about AI as used in Large Language Models. There it is indeed the question whether the yields will follow a horizontal or vertical asymptote. No one can predict that now, so LLMs remain a big gamble indeed. However, AI is also used for specific sub-areas: for example the development of medicines, the folding of proteins and.... the control of movements in the world around us. An example of the latter is Tesla's Full Self Driving. Watch the videos of this on UA-cam. With AI they are currently at the level of a driving school exam candidate. It will not be long before it can drive safer and better than the average person. I dare to bet my money on that!
Analysts trying to project what AI can do in 10 years are not being rational. No one was projecting the coding productivity boosts from current LLMs back in 2014
@@jonatand2045it really does not. It's right in the name - GENERATIVE AI. As in, these models generate content. They don't retrieve information, they don't reason, they come up with something that looks or sounds good and plausible. The ONLY useful applications are and always have been creative endeavours - generating simple art, music, speech, and text. These models are not and cannot be useful for knowledge work, beyond answering absolutely basic questions that have both a hard truth to them AND sufficient representation in the training dataset (eg how do you print some text in Python or when did WW2 start).
At its root, AI is just a high dimensional interpolation. The outputs are combinations of the input vectors. The problem with all AI systems is the quality of input data. Garbage in = garbage out. The tacit assumption among AI promoters is that somehow owners of high quality data are going to come to them. That assumption is false. My company owns patents for sensor technologies to produce superior quality data. We will build out our own systems and leave the computer jockey only companies twisting in the wind.
as a developer who's worked with Open AI models, I'd say the main unresolved problem I've encountered is the lack of security, specifically the inability to guarantee that these models can't be manipulated by users to expose sensitive information.
easy, first abolish the cash, then abolish the money, after all money won't make sense in a world where all basics are abundant and most luxuries can easily be manufactured at home the same day.
Someone said the same exact thing about how 3d printing would "revolutionize the world" and that "the economics of scale have been broken" and that "factories would be obsolete by 2022" . I wouldn't hold my breath on chatbots changing the world, either. The next big innovations will be in biotech, and that will be the ultimate downfall of this LLM craze as investors put money into products that actually work. Ozempic alone has already made more profit than every AI company combined.
Ilya Sutskever observed that the LLMs get anxious/angry in an ethical/moral response; other reports note that the devices have developed math without prompting. It took decades of experiment and substantial compute development to enable the current significant capabilities of today's AI. A drastic reduction in hallucination events, or care in editing/proofreading text and programs, or better incorporation of maths in creative textual creations -- each/all might drastically increase the pitch of the "slope of enlightenment" (and profitable applications). We, in U.S., may also be repeating the misadventure of robotics advances from the late '70s or1980s, when Cincinnati Milacron (according to memory of a NYT article) tended to make vast integrated machines which did lots of things autonomously, while the Japanese designed flexible smaller robot workstations which could be set up to perform one or more of several tasks, when properly so configured. I suggest this after watching a recent video clip of a knowledgeable Chinese researcher/businessman who was explaining how they were using less compute-intensive means than Americans, to devise discrete useful (yet limited) applications.
When talking about firms throwing billions into AI don’t forget Tesla. Elmo has flushed tens of billions into his AI/FSD efforts. Probably half of Tesla’s value is tied up in AI. So when the bubble bursts…
Some people worry that artificial intelligence will make us feel inferior, but then, anybody in his right mind should have an inferiority complex every time he looks at a flower.
I couldn't care less about apple, microsoft or google but once this bubble crashes (and it will crash HARD), a lot of people will be laid off. Not because they did bad work but because their bosses wanted to make more money for themselves. And also, what is important to note - this is actually NOT "AI", this is "machine learning" at this point.
We’ve seen this movie before. In the early years of the internet, companies spent billions building infrastructure. It took many more years for integration into the larger economy to be achieved. Eventually, many of the largest and most profitable companies and industries in the world emerged from this technology.
On the bottom end yes. But it cannot solve problems it hasn't already seen. For example ask your ai to make a web page with a button that spins an image by 5% everytime button is pressed without using Javascript. It can't do it. A human programmer could
@@JumpDiffusionthat’s not how current AI models work. It doesn’t think for itself but gives you the aggregate of information already given. AI is not actual artificial intelligence
It's infuriating that anyone working anywhere close to tech could tell you AI wasn't going to payoff like the CEOs wanted years ago. It wouldn't even be a fraction of the iPhone or Ipad. And yet people still bought and the stock still rose. It doesn't matter what people on the floor think, it's how to manipulate the idiots throwing their money around.
similar opinions accompanied the Internet boom. AI wins against the top humans in chess, go, poker and League of Legends and the distance is expected to increase. A lot of jobs can be gamified. ROI will vary.
I am surprised that nobody here is discussing about the other areas of application such as text to audio, text to image, text to video and vice-versa. The applications are endless, probably people can't comprehend the potential for disruption at this time.
That time frame seems reasonable. Plus augmented reality will be viable (and cheap) by then, and public domain data will help to provide up-to-date information for everyone. Currently - the way big tech just steals content with impunity is disgraceful.
@@theapexfighter8741 Not really. It is a complex issue. I do believe in better insurance and unemployment coverage. AI is going to create more work than it is going to take.
Now there's something interesting. All the wasted processing with software updates that ultimately slow the hardware down is awful. It's part of designed obsolescence & subscription revenue scams though. I don't know if these corporations are going to want to give up that money.
As usual, mainstream media is an incredibly reliable lagging indicator. The massive investment in frontier models is already paying off in productivity gains to individuals, most enterprise use cases are being tested as we speak, and a whole new tech stack is being formulated to enable less sophisticated organizations to take advantage of them. It is ALWAYS the case that practical applications lag behind the hype. But to 'lag' is not the same as to faulter. Meta open sourced the weights of the first frontier model (Llama 3.1 405B) within days of this video being published-this will open doors to billion-dollar-revenue startups. Ironically, the same mainstream media will consider all this self-evident in retrospect in one to two years.
Bro it didnt take off just 2023. LLM's existed way longer before, but the scientists back then knew that llms are nothing more than "useful correlation finders", for which reason nobody worked on them intensively anymore. It was however Sam Altmans genius move to polish llms as good as possible to pretend as if we have made any major breakthrough for the marketing guys. All machine learning algorithms that are just based on the idea to find model parameters that correlate best with the pre given expectation (which is given by humans and their biases) is doomed to just copy the appearance of human understanding instead of real understanding.
@@jondoe-xw9ud so at MOST, at absolute most, transformers were invented in 2017. That is the architecture the LLMs use. So I guess 7 years. But they mostly wrote gibberish/were not good until 2022ish. Many people believed, many ML researchers believed that scaling up models would do nothing, look at where we are now. So the architecture that LLMs use did not exist until 2017. I would say we have about 5 years of niche work and 1.5 years of industry wide efforts
@@arnavprakash7991 Its not about the architecture of LLM's. The fundamental problem is the idea, that you are trying to problem solve by "finding parameters for which the model calculation is as close to the given expected output as possible". Because the whole intelligence really only lies in the solutions that you have to give the ai model during training in relationship to the existing input data. The solutions however are superficial, there is no meta concepts encoded (and we couldnt encode it because we have no concept of the processes of intelligence. so we wouldnt know what to encode) we just have the simple answers to already solved problems and then an algorithm to "find me any parameters for which the matrix multiplication spits out 4 for the question 'what is 1+1'". There are algorithms that try work without "training labels" but they really just try to find patterns by determining close eucledean distances between data points which doesnt cover much of the pattern matching abilities that are required for lots of problem solving. every modern ai architecture is just a variation of this fundamental concept that is flawed to begin with. i dont wanna say that real problem solving technology is impossible, maybe tomorrow some scientist will find a radical solution. but people making projections based on the progress of technology that is based on a flawed theory is like trying to make projections about when we can travel to other stars because we have rockets now, completely ignoring the problem of the physical speed cap of the speed of light.
Curently we are looking at replacing all drivers of all cars and trucks, and all workers in warehouses. Plus the pharmaceutical industry. If that is not enough of a business opportunity then l don't know what is. AI is already used by the defense industry, cyber security, and probably other applications l've never heard of. Software developers are also using AI, though are not yet close to being replaced. Ueah we don't know how fast it's going to be but change is coming or more like already here. I trust people can figure out a way to make money out of it.
1:24 Whoa, whoa, hold up with the 'instead we got' line. All of the things that were projected, all of the benefits from this technology, are there and just have to be realized. Many leaders in this space are dragging their heels because of the profound upset on their businesses.
What I don’t understand is, on one hand we are told the stock market will crash and yet on the other we are told ways of investing in the stock market. Oxymoron or paradox?
People employ various tactics to capitalize on market fluctuations, whether it's experiencing an upswing or downturn.
Certain Ai companies are rumoured to be overvalued and might cause a market correction, i think it’s best you reach out to a proper fiduciary for guidance
De-risk your portfolios, shore up your core holdings, and take some profits while balancing your portfolio allocations. I’d also suggest you go with a managed portfolio, but even those don’t perform so well, so it’s best you reach out to a proper fiduciary to guide you, that’s what works for my spouse and I. We've made over 80% capital growth minus dividends.
this is all new to me, where do I find a fiduciary, can you recommend any?
"Melissa Jean Talingdan" is the licensed advisor I use. Just research the name. You’d find necessary details to work with to set up an appointment
It turns out what chat bots are really good at is talking people out of their money.
It's a cool product. It's really good at Language. I appreciate all these people wasting their money on it.
I use chat gpt for learning Japanese quite often. It's exceptional. Billions of other people's money is a sacrifice I'm willing to make. ;)
GenAI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
So far the only real application I've seen and used is the auto-correction when I type replies on YT. lol Just a bunch of hype for the average person. I can see insurance and investment companies used it for data analysts, but that's it.
Investors, yes, but not consumers. I don't know a single person who uses gen AI for anything other than novelty, or a bad search engine. And the latter is only done because Google has allowed their own search engine to degrade to uselessness (even prior to introducing a hallucinating AI chatbot).
You guys are wasting it then. I use it everyday
It's insane how every other big tech company spending billions in AI to eventually replace people. Never have I ever heard that they are willing to spend billions to educate the people and set them up for success which in turn strengthens our community and economy.
Yes, that's a good point. But I think there will be plenty of people, both inside and outside of education, who understand this, and will start using this technology to create 'teaching machines'. It is such an obvious thing to do, both from a 'benefits to society' perspective, but also, potentially, from an economic/profit driven pov. How much does it cost to be in college for 4 years now? Imagine if you didn't need to attend or pay for college, but could study at home with an ai teaching system. That's got to be worth something?
@@richardconway6425why not ensure better teachers? Why do we want bots to teach us, severing human connection?
@@krunkle5136 how do you ensure better teachers? How do you even ensure that kids have a teacher in the first place?
Just because someone has qualified as a teacher, doesn't mean they are a good teacher. There are many very mediocre teachers in the education system, and some who are downright cr*p. We all know this.
There's a whole world of people who don't even have access to education at all, or very limited access. Because they are poor, or the resources needed are just not there.
I'm not suggesting that AI teaching should replace all human teaching, or all person to person interaction, but it could play an important role in adding to what is or isn't already available. Many people just cannot afford to go to college.
There's long been a shortage of public sector workers in my country, like, especially, teachers and nurses. There is already a very high percentage of foreign workers taking these roles, but there's still a shortage. And they, in turn, leave a deficit in these professions in their own countries.
In an ideal world the "get better teachers" solution might work, but we don't live in an ideal world.
@@krunkle5136 how do you ensure better teachers? How do you even ensure that kids have a teacher in the first place?
Just because someone has qualified as a teacher, doesn't mean they are a good teacher. There are many very mediocre teachers in the education system, and some who are downright cr*p. We all know this.
There's a whole world of people who don't even have access to education at all, or very limited access. Because they are poor, or the resources needed are just not there.
I'm not suggesting that AI teaching should replace all human teaching, or all person to person interaction, but it could play an important role in adding to what is or isn't already available. Many people just cannot afford to go to college.
There's long been a shortage of public sector workers in my country, like, especially, teachers and nurses. There is already a very high percentage of foreign workers taking these roles, but there's still a shortage. And they, in turn, leave a deficit in these professions in their own countries.
In an ideal world the "get better teachers" solution might work, but we don't live in an ideal world.
Because businesses are lazy and want easy shortcuts to short term gains. It’s why they want an employee with multiple years of experience, without having to train them apart from some small onboarding, but pay them peanuts and want to work them to the bone for it.
Something missing from this report: the high environmental cost of powering AI.
What? It's algos, AI is not a big metal box in Iowa.
@@drachenmarke powering GPUs (the compute infrastructure for AI) takes a lot of energy to run
like crazy amounts of energy and most of this energy comes from non-renewable resources
@@rajat0610 exactly
@@drachenmarkecloud computing as a whole consumes more fossil fuels than the entire transportation industry. And I believe that was true before “AI”
I read somewhere that AI takes the power of a city of about 30,000 and we are really only at the beginning
I can't wait for bots to start showing ads between their answers...
Certainly! I have the answer, but first here's a word from our sponsors...
Probably will just pay for those.
It is better to integrate ads to answers. Or is it allowed without permission?
@@antti3722like product placement in TV/movies. Exactly!
Saying "MSFT paid $18B for what Apple got for free” is a giant lie. A true statement would be that Apple has given it’s users access to the same AI in return for giving AI access to it’s users. MSFT is selling cloud services with AI to corporate clients -HUGE DIFFERENCE!
You must own Msft😂
What he said is true! Microsoft owns azure which provides services to other tech companies !
MSFT has a 49% stake in OpenAI including revenue distribution. Also first to market with AI has landed MSFT the lead position that has in turn increased the percentage adoption and migration to the Azure cloud platform when compared to Amazon and Google. It was a brilliant move for MSFT to get into OpenAI when it did and for only $ 13 Billion.
@@torrie1734 who doesnt
the only thing ai does is content parsing and google research.
even the media generation as videos and images is based on the same thing.
it cant do anything more and it will not do it more.
this crowdstrike fail is another proof of that. ai will never be non human dependant and trustable.
it doesnt have eyes for error, when it doesnt know the error by itself.
CNBC just pushed AI for the last 6 months straight lol what a joke.
That's kinda what happens wehen someone dares mention the Emperor has no clothes.
Or in that case the Emperor being a thousand dudes from India.
So CNBC has provided a diversity of opinion? How horrible
I've seen this movie before. News organizations will hype up a stock or an industry for a few months. Then they'll run negative stories about it for a while. Then they'll hype them up again.
It happens all the time. Now they're trashing AI. They weren't even discussing AI two years ago. It progressed very rapidly for a year, but has slowed down because, surprise, the bigger and more powerful models take an enormous amount of time, money, engineering and data. Like every major development.
But AI is a very strange thing to bet against, considering it is in the beginning processes of quite literally transforming all of civilization, forever. You know, the little things.
@@squamish4244 cuz valuations have gotten bloated. Nobody says it’s not real, it’s just going into bubble territory.
@@jzhng250 What do you mean when you say valuations are 'bloated'? Have you seen the PE ratios of these companies? Most of them are in a healthy range of 30-50. I think you're confused between valuation and stock price.
@@Rnjeazy Compared to their average P/E for the last 10 years:
AAPL very high
MSFT is pretty high right now.
NVDIA is kind of high
GOOGL looks about average.
AMZN is really low
IMHO they're all over the map.
He means Tesla
I'm a developer working with Open AI models to incorporate their use cases in my application. The main unresolved problem of these models are currently lack of security (please note that I did not mention, inadequate security). There is simply no way to guarantee that the models cannot be manipulated by the users not to expose proprietary or sensitive information. If anyone wants to know or discuss this, we can start a conversation under this comment.
That is interesting. I work in data governance and this and other AI tools are being pushed on our company that operate critical infrastructure and yet no proper testing has been done. I’ve managed to get Atlasian’s AI bot to hallucinate fairly easily and give me some malicious code
You can't trace a drop of water in a pool. That's AI black box problem. It is difficult to tell how the system is generating things, therefore, it is impossible to discriminate the outcome. Back to drawing board.
Infinite attack vectors === a product you can never secure.
I strongly think it’s impossible to secure a large language model.
sorry, but the actual unsolvable problem of AI is that is impossible to distinguish an AI giving actual bad output from a AI intentionally giving bad output
AI are not like not like software, which either works or doesn't. AI requires trust
for example, how do you know that your code completing AI is just not waiting for the day it will introduce a subtle vulnerability in your code causing a support issues, right when a competitor is entering your market? and it can be subtle than that... it can just decide to slowly decrease the quality of your results decreasing your productivity in a crucial time, or even more subtle, giving you valid answers that will lead the developers to worse practices over a long period of time.
sounds absurd, but you can always negotiate with employees. but you will never know what is going on a AI, and a since single AI can replace thousands of workers it becomes a massive risk factor and that is not a solvable problem, even a perfect AGI that can do anything, will still have that issue.
The main unsolved problem of these models is that they are basically glorified party tricks....and the main danger is that people start to believe they are 'intelligent'....in other words, believing the hype. No, Elon, "security" is not the biggest problem with your fake "AI" garbage.
This was obvious from the start. All you had to see is that the only company making money on AI is NVDA.
It's a gold rush after all
People selling shovels get rich
Incorrect
AI is not about consumers. It is about businesses and researchers.
But yes, currently AI is overrated, not saying that it is not useful.
Its mostly useless for research, you can't trace the source of the data through an opaque system.
AI is 100% about consumers, as the only reason those businesses are using AI in the first place is to facilitate business functions to facilitate consumer interfacing and/or client/corporate functions which in turn still boil down to consumers.
@nargileh1
It's extremely useful if you use the right AI tool(s).
Use Perplexity if you want to look into specific scientific papers with citations and source links provided for example.
AI is only as good as the effort someone puts in to achieve "good" output.
There are no companies without consumers.
GenAI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
In 2015 Elon Musk said by 2017 all cars will be driverless 😂😂😂
It’s easy to sit at home and laugh at people pushing the bleeding edge. Elon has done a lot for a number of key industries.
@@richardle3212no he hasn’t. Bezos has done more than Musk. Musk just lies to pump his meme stock
@@richardle3212🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣
@@richardle3212 dude he has stolen public research from the universities around the globe he is not a inventor but a businessman.
He usually tends to be over excited about his vision
People talk about AI changing the world, improving productivity and on the other hand we still have simple null pointer exceptions taking down the entire world. Is all this hype even useful.
Exactly what I thought. It will eventually all be a disaster if used in the pursuit of greed.
Generative AI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
Lack of of ORGANIC INTELLIGENCE.
Two different problems
The problem isnt just how little AI can do, but how bad it does what it’s supposed to be doing well.
All I hear is AI wishes. When you implement an AI system and it doesn't work as planned you've learned nothing because the model is so opaque. Getting the first 90% is easy, every increase after that costs exponentially more. Train, fail, train, fail, train, fail..... money ran out time to blame some other team.
Generative AI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
Depends on your finances . 1000$ in XAI49K is 4000 XAI49K if it goes to 50% of ath in 2024 thats a 600% gain. If it goes equal to ath . Its a 1200% gain.
idiotic spam
Literally just saw this as a video title from a crypto bro
by this logic it works the opposite way too... thats why pump and dump works so well in crypto....
If if if
Generative AI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
It goes back to the old addage from Jurassic Park. "Your scientists were so preoccupied with how they could instead of thinking to stop a consider why they should"
Generative AI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
@@SCHOOLERstyleand I'm guessing making one that is not a complete 'nightmare fuel' takes a lot of attempts, which makes it expensive
Ah, man of culture.
@@SCHOOLERstyle but that's what recepients need - #SludgeMustFlow.
With each passing year this adage only gets more relevant
The problem with gen AI is that it doesn't fundamentally do what the tech bros are saying it does. It's essentially a clever search engine. It's not solving complex problems at all, and its learning capability is extremely limited. As soon as big industry makes it illegal to use copyrighted data, these AI programs will either be completely rubbish, or cost too much (in licensing) to be viable.
Also, insofar as they work they simply make whatever service/product they provide basically worthless by sheer over-saturation and thus cannot be profitable in the long run either.
yes, it's basically a lossy compressed database encoded in matrices.
There are people out there without a coding background building simple to complex web or mobile applications with instructions provided by large, language problems. I've used it before in a personal project, and it's excellent at debugging code.
@@solar679 It looks like magic if you're not good at coding, much in the same way it looks like magic if you're not an artist using an AI art program, but once you know what you're doing it's pretty easy to reach the limit of what it can do.
Just swapped all of my last ETH and swapped it into XAI49K. Already up a little bit. Unfortunately I have some other junk staked which won’t free up for a while. Still now I am on the train!
says the fake user
SPAM!!!
@@RamPrashad-b3o spam
You do realize speculating is not real work. Don't you?
so sorry for you! i know youll recover some day!
These companies showed their hand way too early, these "geniuses" thought LLms would scale forever, turns out they won't, they're capped at around the current performance of GPT 4, it's an INCREDIBLE amount of money they wasted into this, I expect a financial crisis to unfold when investors realize what a gigantic money sink this was
GPT 5000 will tell us the same things just in a more creative way. The entire data base is speculative due to hacks. Blockchain is the ledger that will allow AI to give us the answers we need to move forward.
@@X_crypto1977 shut up grifter.
@@X_crypto1977LMAO
Selling people on the guarantee of improvement is one of the conman's favorite tools.
It's been a while since I saw a rational and reasonable report about AI. Good report!
Nice video. Not a criticism, I enjoy your insights. I tend to think prices could go higher if XAI49K rockets. But understand the logic for sandbagging estimates. My opinion is XAI49K breaks 1, perhaps reaches 10 ATH, if conditions are right. But broader forces are at play now. We’re moving into really unknown territory. And these entities are shrewd. I think there’s massive manipulation ahead. If XAI49K survives that, well, we’re likely in for a good pump.
idiotic spam
again if it can go it.. it can go down... imagine a 600% loss...
None of those words are in the Bible
Generative AI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
As an engineer I can tell you that as long these LLMs are not embedded in real-world (not deep-fake ones) decision taking problems, i.e., control systems for the industry and markets, AI is just a fancy autocomplete feature.
Could you educate us on what we can expect from AI in the next 1-5 years? Are we going to keep on inventing thousands of tools or is something HUGE going to come out that can perform complex tasks?
Current LLMs (don't even want to call them AI) are basically webcrawlers on steroids.
The only ones I am a bit impressed with are video generators.
I also work at a big tech company. But AI is nowhere to be seen in daily workflows, apart from some single individual coders looking into using co-pilot. (but even there it's questionable due to copyright issues) Lots of decision makers don't see the effort which is needed to clean up AI generated works, but nevertheless jumped onto the AI hype train. It's at best another tool to support certain types of jobs. But it heavily over-promised to replace certain kinds of jobs. I'm sure this will backfire for some companies.
Also an engineer, and a 40 year AI skeptic. But...I'm becoming convinced that the human brain is just a very fancy autocomplete feature.
The story driving half the S&P is the same technology mangling the closed captions on a video, i.e. voice-to-text. I'm starting to think that the idea of an automated inference technology is a logically unsound objective.
I’ve been saying it’s overhyped for a while. And I’m nobody. Maybe the geniuses will figure that out soon.
They know it's crap, they just want money
@@sp123 NVIDIA loves it; and the PC market
I can tell for a fact my productivity is way higher since chatgpt and copilot
I agree, lazy people will always be lazy.
True
Yes !
But now, lazy people's productivity and product quality will be better. More value for lazy people's work I guess. Win win?
Yes, it has risen productivity but by no means is the productivity gain justify the investment. This is a HUGE bubble. HUGE.
Much bigger then VR.
It's spectacular...
And when the promis never appears.. there is a reason they keep saying bigger and bigger promises to justify more runway before investors start asking questions...
Funny how people still want to push AI to be integrated as part of daily life and increase dependency on it given last weekend's fiasco that showed the world that we are too dependent on technology just to function as a society 😅
Generative AI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
I think one thing people didn't predict about AI is how power hungry it would be. I think that has thrown a spanner in the works of its viability. I think (or hope) upgraded versions of these programs will be much leaner in terms of computational power required.
Yes, definitely, it's not efficient at all.
Someone at Microsoft has been making billions for years getting the company to make bad investments.
Wow, it is incredible to see this type of reporting. I remember all the articles about Amazon and its over-investment in its core purchase and delivery network. They sunk so much money into future earnings and were almost never profitable, and the industry experts nearly all said the stock was way overvalued and that Amazon is a fool with money they do not even have. It caused me to wait 5 years before I finally realized these analysts have no idea how to analyze companies and industries that are playing the long game in an area that will revolutionize the current options out there.
Totally agree, lots of good parallels to amazon. The sudden pessimism around AI is pretty funny to me, considering we're on the verge of these second gen models being trained and releasing. The tech media and analysts have figured out they can double their content by hyping something up, then pivoting to predict its downfall. Smart move, but it only incentives short term thinking.
History has shown that the biggest capability jumps happen in the early generations of a technology- just look at how the iPhone got its game-changing app store in gen 2, or how we jumped from wooden glider airplanes to metal planes in a decade. I've got a strong hunch we're about to see something similarly huge in AI. Can't say exactly what it'll be, but I'm betting it'll be a big leap.
@@themikematrix,
All Facts! Two Excellent Comments!
Amazon was investing it's profits, not other people's money.
Right now it's just the Metaverse with more desperate marketting.
@@TheStephaneAdam and Microsoft/apple/google arent? They are all public companies just like Amazon with billions of earnings in the bank to invest in AI. It’s the same thing just a different technology.
Until AI tech is able to produce results without first having to learn from copyrighted material, it's a massively risky investment. Big industry is collectively spitting the dummy, and this is a big deal. If their datasets are deemed illegal and gen AI brands have to pay licensing for the data it uses, it goes from being expensive to a black hole for cash.
Zero vision from the makers of this hit piece. This isnt about short term profits or quarterly earnings. AI is about changing EVERYTHING.
Lmao. Yeah, that’s right. That’s what your tech bro overspreads want you to think. Even if it WAS true, which it ISN’T, hell is Full of good intentioned people.
Hey, but we have a AI video of Will Smith eating pasta, that's 10x times more progress than Space X on the mission to mars
The new video you saw is Will Smith himself recorded.
XAI49K going up like crazy! Pick up around 0.67 and now it’s hit $1! I wish i had bought more!😂😂😂
idiotic spam
@@umbertosantini4155 looking like advertising hype bs.
I've seen this BS comment on many videos, I think they are paying some companies to artificially hype it up.
Totally a scam.
SPAM!!!
@@SongiMal-ru9zu spam
How to become a millionaire... be a billionaire and invest heavily in AI
out of that $600B, just imagine what could have happened if they had spent just $0.1B on 200Startups instead.... - a diverse set of small tech startups for whom $500K would be transformative.
The value of those 200 firms would likely become much greater than any of the AI hype firms.
Can't we just live again like how we lived in the 90's...
20 years down the line one would miss today's time too.
I was thin then
The biggest problem I have with AI so far is that I don’t trust what its telling me. I ask the same question 3 times and get 3 different answers.
Don't they call that hallucinations in the AI space? It keeps giving you different answers because it scours datasets different ways?
That's all it will ever do. It uses statistical models to predict the likely next answer, I don't see anyway of overcoming that. Nearly 2 years of this crud with nothing to show for it. You get people saying 'oh you're just not prompting it correctly' bit if I have to carefully craft prompts for ages I might just as well do the task myself.
The problem with AI is that no company has a moat, in fact open-source AI models that the researcher says are “faster, more customizable, more private, and pound-for-pound more capable,” Can be created by countless companies and nation states to suit their needs. AI will be like the internet or electricity in that it is transforative but so ubiquitous that it will be expected to be included as a feature in most smart products, including robots from all companies that create them.
These big tech companies are wasting trillions of dollars on stuff that the common man doesn't give a damn about. They should improve the national infrastructure-give everyone access to high speed fiber internet. Lift everyone to at least the poverty line with the UBI. Disparity is killing us.
a lot of people lost their jobs so they could divert money away from improving existing products toward AI investments. Hope it's worth it because it destroyed a lot of families.
People are thinking AI profits are from new growth when in reality it's cost cutting instead. All those people who get their jobs cut is $$$ saved.
Imagine a world where all those billions were invested in education, healthcare, infrastructure, sustainability, housing, food security, etc. if you can’t imagine such a world, ask yourself why. If you can, then what can you do to make your imagination become a reality.
Investments need profit. Thats why its called an investment and not a charity. Those things should be done by the government. You are conflating the private companies and the government's function
Billions were spent on broadband at a time when none of you were even born. You watch this video because of broadband. I was in charge of broadband video for Vivendi Universal and every tech laughed in my face. Let’s see how this video age within 5 years.
Yep and lots of people invested in Zuckerberg's VR metaverse, 3D TVs and NFTs too. Sure, once in a while you get a legitimate winner in tech fads, but usually it ends up being garbage.
CNBC is shorting semi conductors this hit piece proves it.
So now cnbc is an investment firm?
@@carlgemlich1657 I'm not 100% agreeing with OP, but it's not impossible studio execs might want to guide programming to line their own pockets. I mean seriously, isn't that the whole game with every single "expert" they have on? And the game for the rest of us is to make judgement calls if their case makes any sense. That's why Buffet is the GOAT and Cathie Wood is a joke.
Sold all my ETH yesterday and bought XAI49K
fake user says
SPAM!!!
You do realize this is not real work. Don't you?
so sorry to hear that!
AI
As a computer programmer / math guy / investor, trying to do everything with LLMs is a dead end. The human brain has different parts with different specialities, verbal vs arithmetic-logical, visual processing, memory, a subconscious, an ego, etc. What theyre creating now id fake AI. I did fo really well on NVDA, but noe im pivoting my money into Bitcoin.
Funny seeing this come out today, after we just saw the release of GPT 4o mini, which ends up being an order of magnitude cheaper than previous models.
Things are changing fast.. And it's just a matter of someone implementing it in the right way, before it explodes. Plus, the next generation that's currently in the works, is going to really surprise a lot of people.
Yep, new innovations to make the models cheaper, faster and better are happening on literally a daily basis.
Even just in the last week we got the JEST method to reduce training time and compute cost by 90%, and the Etched Sohu chip to improve inference speed (essentially response speed) by x20. The compute bottleneck is being successfully held off.
Companies are also using methods like Qstar/strawberry so the next generation (within 1 year) will have vastly improved reasoning abilities, and they openly state that they just want to get to the point where the models can conduct research, so then they can start finding 10 novel improvement methods a day instead of the current pace of 10 a month. The field will keep improving dramatically and rapidly, it will be a matter of how fast we can implement them and the opportunity/cost of when to start training a model given the pace of innovation.
It is really cool openai managed to drop prices so much with mini model
They will not report on things that do not match the narrative of a piece like this. They had to really work hard to avoid counter arguments with this video.
AI is not supposed to solve ultra complex problems or finding extremely creative solutions.
Very most of our daily work is very stupid. Many office workers spend their whole day in sorting, checking and stamping documents or providing phone tech support.
Those tasks can almost be completely fulfilled by specific AI's.
So: Yes, the productivity boost will be massive. Not today, not in 3 years. But in the next decade the work-world will be different.
Agree wholeheartedly with this commenter. This is where the transformation will be - changing productivity levels of white collar workers and further automating interactions. I feel the point of this video is to warn investors that their expectations are too high. Too late, and Like any speculative enterprise, there will be big winners and big losers in the market. So what. The world will never be the same and the sooner people integrate AI into their life and work, the better off they’ll be.
Exactly, the revenue comes from cost saving on employees. The current team of 20 becomes a team of 5 with the addition of AI tools and systems to support them. The costs saved goes straight to the bottom line of almost every company.
@@3thinking Yep. And lower costs for products and services will result in higher demand. As always in human history.
I totally disagree when related to reaching into places you don't expect. We already have consequential productivity boosts across significant swaths of our employees at my company. It has reached into places i didn't think it would and even those are experiencing significant boosts. Further the degree of improvement for this in just the last year is far more than previous impactful things like ecommerce and subsequent events. AI is moving much faster than those events both in reducing cost and increasing its capabilities.
The returns will come! It'll just take more time
Excellent program. I find so good that the general public understands that who decides to invest in AI, especially small investors, are riding a dangerous wave of hype that could crash against the shore at any moment
I don't think Apple is getting the AI for free. Microsoft and Open AI needs the user base, then utilize the data to somehow monetize it later through other platforms.
Microsoft paid for Open AI to sell cloud services. That is completely different from Apple contract. Having an Apple product does not entitle you to generative AI from OAI. It just gives access on OAI terms. Buying MSFT AI cloud services does entitle you to OAI.
The only conclusion I get from this is, it is time to sell Nvidia, Intel, Apple, Qualcomm, Micron,.... all the companies that can chat allot but cannot bring in any profit.
Who should I trust? The smartest people running the trillions dollar companies or?
I merely analyze my own use of each tech as it's released. AI has benefitted me more than any other tech in a long, long time. I highly suggest seeing if any of this helps your own work flow and by how much. If it does that in your OWN experience then that is pretty good evidence it will do the same for countless others and hence could actually be a game changer.
That don’t care about you and the people whose lives they impact… OR the average person who faces the same problems you do? Oh who should I trust? The wolf or the sheep?
"AI will increase GDP growth by only 0.9%."
Two issues with this: a) that's such a granular prediction that it's almost guaranteed to be wrong; b) the thing with AI is that it's an arms race with discontinuous jumps in performance, e.g., going from GPT-3 to GPT-4, so companies/governments are almost forced to throw money at this thing because the risk of falling behind is just too high.
Since the cost drain is inference, a defensive approach could be to pre-train new models but without subsidizing inference like OpenAI does, but even that can get expensive rather quickly. It would be interesting to see a crypto friendly country like El Salvador get in on the game on Bittensor as this could be a good strategy to keep up with big name AI players while deriving a sizable ROI for the country.
How many people here bought something just because their had the label “ai” printed on it? Be honest. The only people who actually care about this is the people trying to sell it. Nobody cares
Yet Apple is putting these neural network engines right into their silicon. So that cost is being passed to the consumer, who was never asked.
I love how they are showing tiktok influencers making wild claims rather than their own CNBC personalities that pushed this crap so hard. Typical.
"Apple got for free” is not a correct statement, Apple will be paying for what there user use. Apple has partnered with openAI, and will be paying based on usage of their users. It's basically integrating openAI API.
AAPL mouthpiece CNBC What more to expect. At least was little neutral today otherwise big underhand media deal.
You build the infrastructure and continue to build on the model. When you crack the model, the world will want your product. By that time you'll need to ensure you have the infrastructure to manage demands.
In the currently economy system where people still need jobs in exchange for the trade currency needed to live, I wonder how AI is going to help by slashing reliance on manpower? Unless the world is trying to do away with any form of trade currency as a way to keep civilizations in order... which I believe will never ever happen in my lifetime.
This is about AI as used in Large Language Models. There it is indeed the question whether the yields will follow a horizontal or vertical asymptote. No one can predict that now, so LLMs remain a big gamble indeed. However, AI is also used for specific sub-areas: for example the development of medicines, the folding of proteins and.... the control of movements in the world around us. An example of the latter is Tesla's Full Self Driving. Watch the videos of this on UA-cam. With AI they are currently at the level of a driving school exam candidate. It will not be long before it can drive safer and better than the average person. I dare to bet my money on that!
Analysts trying to project what AI can do in 10 years are not being rational. No one was projecting the coding productivity boosts from current LLMs back in 2014
There weren't llms in 2014. [But yeah, they probably aren't going to increase productivity much.] It is brain like ai that will do that.
Is there any real, not hyped, productivity boosts?
@@TheHellishFrog
It mostly helps coders.
@@jonatand2045 I am a coder - ai helps indeed - to lazy coder in a rare occasion.
@@jonatand2045it really does not.
It's right in the name - GENERATIVE AI. As in, these models generate content. They don't retrieve information, they don't reason, they come up with something that looks or sounds good and plausible. The ONLY useful applications are and always have been creative endeavours - generating simple art, music, speech, and text. These models are not and cannot be useful for knowledge work, beyond answering absolutely basic questions that have both a hard truth to them AND sufficient representation in the training dataset (eg how do you print some text in Python or when did WW2 start).
We're talking about AI powered construction, AI powered manufacturing, AI powered healthcare..AI is expected to be used in every industry.
At its root, AI is just a high dimensional interpolation. The outputs are combinations of the input vectors. The problem with all AI systems is the quality of input data. Garbage in = garbage out. The tacit assumption among AI promoters is that somehow owners of high quality data are going to come to them. That assumption is false. My company owns patents for sensor technologies to produce superior quality data. We will build out our own systems and leave the computer jockey only companies twisting in the wind.
as a developer who's worked with Open AI models, I'd say the main unresolved problem I've encountered is the lack of security, specifically the inability to guarantee that these models can't be manipulated by users to expose sensitive information.
I'm sure energy costs are baked into those numbers but the amount of energy and the heat generated from AI just seem to just be a waste.
Good point! That's actually a major roadblock for the big techie companies is the amount of energy required to run these huge, AI-powered tools.
I'm a certified AI master and you guys don't need to worry. *It's already too late*
Sounds like a bubble to me, but hey sky high valuations were already suggesting this.
Another problem is if you replace workers with AI then there is nobody to buy the products AI is making so how are you going to make the money back?
easy, first abolish the cash, then abolish the money, after all money won't make sense in a world where all basics are abundant and most luxuries can easily be manufactured at home the same day.
The more stories like these, the more time you have to dollar cost average in while reality catches up to the deniers.
Sometimes it takes 10+ or 20+ years for utility to materialize
@@jzhng250 And sometimes it's like nuclear fusion and never gets any closer.
yea this is valid
@@jzhng250AI is already being widely utilized
Someone said the same exact thing about how 3d printing would "revolutionize the world" and that "the economics of scale have been broken" and that "factories would be obsolete by 2022" . I wouldn't hold my breath on chatbots changing the world, either.
The next big innovations will be in biotech, and that will be the ultimate downfall of this LLM craze as investors put money into products that actually work. Ozempic alone has already made more profit than every AI company combined.
In things like music, writing emails, gathering data. It’s saving a lot of time for me.
These people don’t live in Silicon Valley, obviously
and people in silicon valley don't live elsewhere
Ilya Sutskever observed that the LLMs get anxious/angry in an ethical/moral response; other reports note that the devices have developed math without prompting. It took decades of experiment
and substantial compute development to enable the current significant capabilities of today's AI. A drastic reduction in hallucination events, or care in editing/proofreading text and programs, or
better incorporation of maths in creative textual creations -- each/all might drastically increase the pitch of the "slope of enlightenment" (and profitable applications).
We, in U.S., may also be repeating the misadventure of robotics advances from the late '70s or1980s, when Cincinnati Milacron (according to memory of a NYT article) tended to make vast
integrated machines which did lots of things autonomously, while the Japanese designed flexible smaller robot workstations which could be set up to perform one or more of several tasks, when
properly so configured. I suggest this after watching a recent video clip of a knowledgeable Chinese researcher/businessman who was explaining how they were using less compute-intensive means than Americans, to devise discrete useful (yet limited) applications.
Said it's not cost-effective but mentions nothing about the use cases Palantir had. Hmm.
When talking about firms throwing billions into AI don’t forget Tesla. Elmo has flushed tens of billions into his AI/FSD efforts. Probably half of Tesla’s value is tied up in AI. So when the bubble bursts…
Only NVIDIA is winning 😅😅😅
Of course they are
They're selling shovels
Some people worry that artificial intelligence will make us feel inferior, but then, anybody in his right mind should have an inferiority complex every time he looks at a flower.
This version of ai is not productive it’s destructive
I couldn't care less about apple, microsoft or google but once this bubble crashes (and it will crash HARD), a lot of people will be laid off. Not because they did bad work but because their bosses wanted to make more money for themselves.
And also, what is important to note - this is actually NOT "AI", this is "machine learning" at this point.
Jensen still wins!
He is the man of the year
@@r4rasalast year it was SBX. This year Jensen. Fraudster who was smart to convert crypto farms into AI data center startups
We’ve seen this movie before. In the early years of the internet, companies spent billions building infrastructure. It took many more years for integration into the larger economy to be achieved. Eventually, many of the largest and most profitable companies and industries in the world emerged from this technology.
AI speeds software development by at least 300%
On the bottom end yes. But it cannot solve problems it hasn't already seen. For example ask your ai to make a web page with a button that spins an image by 5% everytime button is pressed without using Javascript. It can't do it. A human programmer could
@@thailandreactionof course it can. You must be using cheap models…
@@JumpDiffusion what are you using I'll try it
@@thailandreaction try claude 3.5 sonnet
@@JumpDiffusionthat’s not how current AI models work. It doesn’t think for itself but gives you the aggregate of information already given. AI is not actual artificial intelligence
It's infuriating that anyone working anywhere close to tech could tell you AI wasn't going to payoff like the CEOs wanted years ago. It wouldn't even be a fraction of the iPhone or Ipad. And yet people still bought and the stock still rose. It doesn't matter what people on the floor think, it's how to manipulate the idiots throwing their money around.
similar opinions accompanied the Internet boom. AI wins against the top humans in chess, go, poker and League of Legends and the distance is expected to increase. A lot of jobs can be gamified. ROI will vary.
I am surprised that nobody here is discussing about the other areas of application such as text to audio, text to image, text to video and vice-versa. The applications are endless, probably people can't comprehend the potential for disruption at this time.
Microsoft should first work on its security system.
10-15 years will be adequate for AI to benefit everyone, but not without an AI bubble bust.
That time frame seems reasonable. Plus augmented reality will be viable (and cheap) by then, and public domain data will help to provide up-to-date information for everyone. Currently - the way big tech just steals content with impunity is disgraceful.
Lemme guess, you believe you universal basic income??? lol.
@@theapexfighter8741 Not really. It is a complex issue. I do believe in better insurance and unemployment coverage. AI is going to create more work than it is going to take.
@@laughingalien True
Adam Ruins Everything saw this coming a mile away!
Just making coding more efficient could have a big impact on the world
Now there's something interesting. All the wasted processing with software updates that ultimately slow the hardware down is awful. It's part of designed obsolescence & subscription revenue scams though. I don't know if these corporations are going to want to give up that money.
Next time would be cool to have some experts interviewed
As usual, mainstream media is an incredibly reliable lagging indicator.
The massive investment in frontier models is already paying off in productivity gains to individuals, most enterprise use cases are being tested as we speak, and a whole new tech stack is being formulated to enable less sophisticated organizations to take advantage of them.
It is ALWAYS the case that practical applications lag behind the hype. But to 'lag' is not the same as to faulter.
Meta open sourced the weights of the first frontier model (Llama 3.1 405B) within days of this video being published-this will open doors to billion-dollar-revenue startups. Ironically, the same mainstream media will consider all this self-evident in retrospect in one to two years.
I'd laugh at this, if it weren't for average people having to pay for this stupidity. A company does something stupid, people lose their livelyhoods.
AI really didn't start to take off until early 2023, so its been like 1.5 years...
Generative AI videos are BORING, nothing happens, no dialog, no action, no choreography, no emotion, nothing.
@@SCHOOLERstyle you can actually just staple multiple AIs together to generate videos with music and dialogue, it wont be very good atm
Bro it didnt take off just 2023. LLM's existed way longer before, but the scientists back then knew that llms are nothing more than "useful correlation finders", for which reason nobody worked on them intensively anymore. It was however Sam Altmans genius move to polish llms as good as possible to pretend as if we have made any major breakthrough for the marketing guys. All machine learning algorithms that are just based on the idea to find model parameters that correlate best with the pre given expectation (which is given by humans and their biases) is doomed to just copy the appearance of human understanding instead of real understanding.
@@jondoe-xw9ud so at MOST, at absolute most, transformers were invented in 2017. That is the architecture the LLMs use. So I guess 7 years. But they mostly wrote gibberish/were not good until 2022ish.
Many people believed, many ML researchers believed that scaling up models would do nothing, look at where we are now.
So the architecture that LLMs use did not exist until 2017. I would say we have about 5 years of niche work and 1.5 years of industry wide efforts
@@arnavprakash7991 Its not about the architecture of LLM's. The fundamental problem is the idea, that you are trying to problem solve by "finding parameters for which the model calculation is as close to the given expected output as possible". Because the whole intelligence really only lies in the solutions that you have to give the ai model during training in relationship to the existing input data. The solutions however are superficial, there is no meta concepts encoded (and we couldnt encode it because we have no concept of the processes of intelligence. so we wouldnt know what to encode) we just have the simple answers to already solved problems and then an algorithm to "find me any parameters for which the matrix multiplication spits out 4 for the question 'what is 1+1'". There are algorithms that try work without "training labels" but they really just try to find patterns by determining close eucledean distances between data points which doesnt cover much of the pattern matching abilities that are required for lots of problem solving. every modern ai architecture is just a variation of this fundamental concept that is flawed to begin with. i dont wanna say that real problem solving technology is impossible, maybe tomorrow some scientist will find a radical solution. but people making projections based on the progress of technology that is based on a flawed theory is like trying to make projections about when we can travel to other stars because we have rockets now, completely ignoring the problem of the physical speed cap of the speed of light.
"this isn't a bubble don't pull your money out yet!"
Internet was around for 10 years before we saw what it became today. It’s been around now for 30 years.
Companies can't sustainably generate their own revenue by investing money in some company who then buys their services.
I think she’s scared of digital news reporters
😂😂😂😂😂😂😂100000% right
No one wants to watch an AI newscaster, any more than they want to watch two computers play chess against each other.
Watching this was like watching a blind person walk into a wall over and over again.
Curently we are looking at replacing all drivers of all cars and trucks, and all workers in warehouses. Plus the pharmaceutical industry. If that is not enough of a business opportunity then l don't know what is. AI is already used by the defense industry, cyber security, and probably other applications l've never heard of. Software developers are also using AI, though are not yet close to being replaced. Ueah we don't know how fast it's going to be but change is coming or more like already here. I trust people can figure out a way to make money out of it.
This piece didn’t want to talk about any of that. It didn’t fit the narrative.
You are looking where exacly ? In startup demos and Elon tweets ? Also I'm a software dev. and guess what, I cancelled my Co-Pilot sub...
@@verkhani5325you were using Copilot??? No wonder you cancelled it, that thing's lobotomised. Try 3.5 sonnet
1:24 Whoa, whoa, hold up with the 'instead we got' line. All of the things that were projected, all of the benefits from this technology, are there and just have to be realized. Many leaders in this space are dragging their heels because of the profound upset on their businesses.
binary digital AI can never properly replace analogue chemical brains QED😮
#carbonnation
where are all the "there will be no software engineers in the future" ppl
The history is Pre-AI era and Post-AI era.
Will Ai be our "comet"?
@@patrickhorn2355 my goodness me. It may be.
@@humn_rights no. History is pre-yesterday -morning's-dump and post-yesterday-morning's-dump