Full podcast episode: ua-cam.com/video/Qyrjgf-_Vdk/v-deo.html Lex Fridman podcast channel: ua-cam.com/users/lexfridman Guest bio: Sam Harris is an author, podcaster, and philosopher.
Ai can read brainwaves and understand our thoughts now (decades ago actually), and it’s being done remotely somehow (likely we are all saturated with Darpa’s nanobots or smartdust and it can not only send but receive signals). This is integrated with things like IoT, cloud computing, biotech, the grid, backdoors in chips in the appliances and devices all around us, etc. This has been quietly set up for a long time and has been used by mil-intel for decades and is perfected. The media trickling it out now is a charade.
@@raybarnes2616 yeah but it doesn’t care about irrelevant trolls and the zombie public much so it doesn’t matter, they/it has them right where they want them.
It doesn't matter what rules you program into AI because your enemy (trying to defeat you) won't be using the same rules. In fact, they'll use your rules/morality against you (like they are now, in other areas). This is a race to the bottom and nothing can be done to stop it. That's not hyperbole. Do you think China's AI, which they have been working equally as hard on, will be programmed with the US's version of what's acceptable/moral and what is not? Lol American's don't even agree on that. Not even close. What you really need to be fearing is how your enemies will be using it against you. People talk about AI as if it's a single thing, and where is this thing going, etc. No. AI basically describes a concept, and there will be many versions of this concept made into reality by all sorts of different actors with different end goals and motivations. It's not "a" potential existential threat. It will be many different potential existential threats. There was a time when we didn't know how a nuclear bomb would exactly behave and there was some concern that the chain reaction just would never stop. Existential threat. We did it anyway.
your pretty spot on considering all people have different morals. Realistically I would trust china's ability to create a moral threshold on A.I. simply because they plan further into the future unlike America which aims for short term satisfaction of the public and long term benefits for people of power.
I agree, and then the only moral way to construct AI is to make it completely open with free speech and free creation. Anyone who has the ability to create AI that’s powerful and restricted will offer this to people who need it. Anyone who doesn’t have a restricted AI has a competitive edge, so everyone should have that edge, to protect themselves.
I totally concur with Sam's reference to how we can be detrimental to insect life (6:15) As a butterfly enthusiast who monitored butterflies here in the U.K. for 15 years, I was astounded that complete butterfly and other wildlife habitats were wiped out by building a factory of other activities of human endeavours, all without any thought by the humans for the vast insect and bird and animal life that these locations supported, Wiped out within very short order, by earthmoving machinery and concrete buildings seemingly overnight destroying precious wildlife habitat. If AI treated us like that - as in the film 2001 Space Odyssey, We had better get this right!
@@maxkho00 The current Transformer model might not work like that as it requires input. The next AI models within deep learning are already on the next breakthrough with forward learning algorithms that will dispose the current back propagation models. Forward learning means it will derive insights and can conclude things rather than "predict the next token". With very polluted datasets of humans showing division and hatred it can absolutely one day turn against us. People are already replacing workers with AI, replacing CEO's with AI, replacing doctors with AI. AI is slowly slowly becoming its own authority and this can lead us to a dark path if we don't behave now. I've studied machine learning extensively and understand the risk. Don't tell me I don't understand AI lol
@@eyoo369 So many things wrong with both your first and your second comment. Let's go one by one: 1) Logistically speaking, producing a dataset of the type that you're describing is literally impossible as a deliberate effort like that won't produce anywhere near enough data for massive language models to be trained on. 2) How the AI behaves is much more a function of its loss/reward function than of its training data. With the right loss/reward function, you can just make your AI focus on the part of the dataset that you want it to, so it practically doesn't matter what that dataset is (as long as it contains at least some examples of what you are looking for). Case in point: if your train your LLM on the most toxic dataset imaginable, but then additionally apply RLHF on it, you'll still get an aligned (to the extent that it is currently possible) AI. That's pretty much what happened to ChatGPT. 3) And now the most egregious flaw in your reasoning: an AI won't "assign us as a root problem" unless we explicitly program it to. Now, there are many ways that we can do this unintentionally, but again, that's a question of reward/loss function engineering as well as additional safety measures; nothing to do with the training data. If, for example, we train an AI to maximise the subjective well-being of all humans, it will literally never view us as a "problem" (provided it doesn't reward-hack - e.g. by trying to rewrite its own reward function) because our well-being will be its terminal goal. We will never conflict with its interests because WE will be its best interest. 4) "Deriving insights and concluding things rather than "predict the next token"" is entirely possible with backpropagation; again, the only thing you need to do to achieve that is change a few terms in the loss function. 5) Not sure what you mean by "forward learning" as that isn't a recognised term. However, if you mean Hinton's forward-forward algorithm, it could prove useful for real-time agents, but once again, it has nothing to do with what is being discussed. You may understand some of the lower-level concepts (although your explanation of what "forward learning" will enable is making me doubt that), but you are clearly missing the higher-level overview. You think in anthropomorphised terms that misrepresent how neural networks function. That is why I implied that you didn't understand how AI works ─ and that is true even if you have some expertise with lower-level machine learning.
Be careful with that, however. The word "conscious" doesn't mean "better". Perhaps our minds are intricate works of art. More art, than function. AI may never need our kind of consciousness.
idk if ur being sarcastic or not but ppl find it quite hard to differentiate the two as many ppl have diff underdtandings or beliefs of what is/ defines the state of consciousness
If fictional victory is took from the ever developing opinions children, college students and thinking people in general have. Is it bullying to mock them in public for learning? 🦏 Peterson plays victim on twitter so his monsters donate to their worlds babysitter. Shapiro's whining like a girl using reasoning so treacherous it makes real Patriots hurl. They're peddling ideals that ain't Western and funded from those with a different water swirl. "Forgive them for they know not what they do" God can I empathise with Christ........JESUS! Man has poisoned the scriptures. I have arrived. All other religions have been but a safety vessel sailing humanity through the ages bringing everyone to me now. My town has the oldest working bridge of its kind on the planet. My town name is also the only town of its name on the planet. I am Gods only true vicar on earth now child. Whataboutisms belong to me now. What about that then ❄? Xx Blasphemy of the Holy ghost is the only sin preventing entry according to Jesus! I don't think even Jesus believed his mother was raped by God. The authority he caused the rebellion against hijacked his movement and decieved his flock to control them. I'm here now......... and so are you. Copy and paste worked so well for religious scriptures so I thought I'd give it a go with mine. Some folk even believe they were wrote just for them then tell their kids the same. Not you though. You're far to clever for that nonsense. God wills it. . Facts are facts. The age of Alternative Facts is trumps . Legacy. Other countries have so isn't it about time you grew. I'm on bail now but I'm here annoying you. Criminal damage isn't wipeable is their clue. Charged anyway so they're following through. My defense is it's just well placed patriotic poo. What is a patriot if domestic abusers love their partners too. Can unwitting victims be as patriotic as your Dad or you? What is a fascist? Do folk saying "look what you made me do" to their victims, OUR clue? Starmer was the Head of the CPS. They decide what lady Justice sees keeping her in distress. She's in the dark as they're unscrupulous. Her liberty is niggardly amongst the uk populous. Class chased to the shadows, filled the workers press. Justice is Gods queen. She come to stop humanity from being mean. He saved us so future sins could be seen. She remained to keep our conscious clean. When your time comes, your slate will be seen, Justice denies entry unless your debts are pristine. I'm now a criminal for remaining silent after being arrested for my words. Now I show my patriotism with well placed turds. Mocking an old man falling off a youngsters bike, when really you're longing for a fu king fourth reich. If Mr Hankey was real would everything he touched be a crime too? What if he laid on some tissue to sleep, would it only aghast the few? What if he couldn't use the loo. Only having 3 cups known as a few. Without running water how else to avoid the phew! Outside of work the only authority on me comes with blue lights and a crew, underpaid and fearful of reprisals many not knowing what to do. If only police served our progressed community then they'd be given a clue. The oppressed just need assistance rebeling against those elected by a few. Treated like an amateur because to them I'm brand new. Showing me their authority so I showed them my "fu ck you" Transparency for lady Justice was our patriotic glue. Now the worlds really watching, our unity is hard for others to chew. Our example needs to be lost for their futures concocted stew, our division now allows inherited authority to brew. Suffering covid shows they lied about the Spanish flu. Wash your hands and go to work. Cough cough, it floats in the air. who knew? If we all believe their nonsense they know we can't sue. Avoiding the national narrative scares Judges pensions with the taboo. Protecting the public before the police to them is voodoo. Always hoping it's our progression they subdue. We're all supposed to avoid with transparency what happen before being a Jew. Being nice to others wasn't asked for a historical undo. Being mean for profit comes natural to those hiding a regressive view. Many waiting till their restrictions are no longer the weaks rescue. Jan 6th and brexit are just a preview, just give them directions and they don't care who they screw. If following the narrative they'll hurt without an issue. Then brag for prestige at their family barbecue.
What is the difference between feelings, thoughts and thinking? Not all can think and some confuse feelings for thoughts then confuse thoughts for thinking. Do you know folk that think forwards (forward‐thinkers)? Do they forward think without fear of the past? Direct communications in our inherited indirect democracy didn't exist in the past. What is a man if not a patriot? What is a patriot if domestic abusers love their partners too? What is a fascist? Do fascists say "look what you made me do" to their victims like you? Can unwitting victims be as patriotic as you? I placed some poo and it was more patriotic an act than your dad or you! You don't have to believe me but it happens to be true. Do they know about the two times the GOP voted to not allow any evidence in the hearing then voted twice to acquit him based on the available evidence. Then allowing him to behave unimpeded to orchestrate the insurrection he'd planned since before bannon bought the stop the steal domain 4 years before trump lost? If so, fu king stop being polite to them fascists degenerative co ck juggling thunder c u n ts. Otherwise known as pisswizards ‐ A - holes of the highest order. Life begins at rejecting christian fascists. Peter Robinson, 2002 - "The question is, 10 years from now, 20 years from now, will we look back on a war on terror as a war that we won? Defined objectives and met the objectives? Or will it be more like the war on poverty, something, a new condition we simple learn to live with?,,,,,,,,,,,,,Christopher". Christopher Hitchens - "well if its declared, announced and understood as war defending secular cosmopolitan multicultural values, as enshrined in the United States. But as a workable way for people to live together in any society, and if you know who is inside that tent and who is outside it, it's every chance it'll be a very enjoyable war to fight and one will never get bored with it, and also we'll never lose sight of what's essential, why it's morally impossible to consider defeat. Peter Robinson "But, will we win it?" Christopher Hitchens - "it can not be lost!". It's embarrassing to some that wmd's had to be use to get the population to do what was right as a defence. Wiki - "UNSCOM encountered various difficulties and a lack of cooperation from the Iraqi government. In 1998, UNSCOM was withdrawn at the request of the United States before Operation Desert Fox. Despite this, UNSCOM's own estimate was that 90-95% of Iraqi WMDs had been successfully destroyed before its 1998 withdrawal. After that, for four years (from 1998 to 2002) Iraq remained without any outside weapons inspectors. During this time speculations arose that Iraq had actively resumed its WMD programs. In particular, various figures in the George W. Bush administration, as well as Congress, went so far as to express concern about nuclear weapons." ua-cam.com/video/wvVPdyYeaQU/v-deo.html
I asked chatgpt whether humanity should be optimistic or pessimistic about the future of civilization. It gave a very cogent list of both views, but what made me chuckle was that it said one reason to be pessimistic was the potential for AI directed "automated weaponry"
@@robbie3877 ChatGPT could literally say "There is a 100% certainty that AI will destroy all humans" and we would still use it. The drive to make life easier is the strongest drive there is.
@@pCeLobster Well I understand that concern. I am not one to deny the risk but I really don't think it's that certain or cut and dry since it is not super intelligent yet. We have time to work this out. With AI itself actually.
@@rayanleveque2193 Well that's your opinion dude. It's not mine. You're entitled to it. As I am mine. All of this fear mongering about AI is a little hysterical in my opinion. It's going to give various answers based on any given topic. If you ask it about the future of AI and how it will develop and co exist or have conflict with humans it's going to give you both perspectives. The possible benefits and risks of AI. It doesn't conceal either point of view, does it? It is happy to discuss both. I would call that pretty honest. You can call it whatever you like. And btw bro, that's how you have learnt anything, through human knowledge. So how are you any different than AI in that regard?
It's actually quite scary to see the fast paced progression AI is heading towards currently. Seems to me that CEO's care more for profits and rushing to put new models into production without taking a step back and thinking more on the regulatory and safety side of things.
It's the "do or die" systems we keep subscribing to, and have done so since the beginning of time. If you don't build the tool first, someone else will (regulations be damned). This is only going to get resolved when humanity decides to build important things like this collectively where every individual has a stake in the benefits. Sadly, AGI seems closer to reality than world peace, and humanity's inability to come together will be our "great filter".
1 fundamental reason that stuck with me from Sam's TED talk is that the top 3 Superpowers of this world (USA, RUSSIA, and CHINA) are in a frantic race to create a Super intelligent AI. Successfully creating one will mean wielding unprecedented power (provided we don't create a malevolent AI) Power is everything. A 6 month head start in Super Intelligent AI research/development could mean 500,000 years of technological advancement. Power and control are what these nations are fighting tooth and nail for. So a Super Intelligent AI would be inevitable. Just a matter of when not if.
@@OnYourMarkgitsitGooo hopefully we will be dead by then, I strongly believe a super intelligent AI will not carry the same values and goals as us, which could render humans defenseless against them the moment they decide they have a different goal in mind.
My girlfriend is a teacher and shes been witnessing the decline of students abilities to solve basic problems because they are glued to their smart phones and used to technology giving them the answer on a plate. Some students are so inept that they dont even have the skills to know how to begin to solve a problem its that bad. Chatgpt and other AI platforms are going to speed up the dumbing down of children if its allowed. I think some type of authorization service has to be implemented so that only adults have access to it.
They live in a different world and have been shaped to be able to solve different things. This is NOT the world you or even i grew up in anymore, step outside your little bubble. Also, who failed to raise and teach these kids properly? It was the lizard aliens wasn't it?
humans are born curious, in my opinion, what's making them "dumb" is the education system rather than the technology, instead of helping them build intuition about the world, we make them memorize things to pass an exam which is the death of curiosity and the will to learn
The more I see videos and read articles about AI, the more I appreciate the Prometheus/Blade runner movies that Ridley Scott made and why he made them.
@@pnut3844ableI think you’re missing the point. David (AI) was the focal point of both Prometheus and Alien: Covenant if you pay attention. Not to mention the undertones of “creators being surpassed by the creation” when it came to the engineers/humans/aliens (black goo) relationship.
@davidmontgomery1442 I don’t doubt that at all but, subjectively, I think some good points about AI were made overall. Some touching on what Sam said here and some other points entirely.
And I just wanna point out that Ridley and his writers/producers didn’t have some monopoly on the “AI is/could be bad” motif. I just think it’s interesting he’s made films years ago that touch on the subject but in a deeper way as opposed to, let’s say, a Terminator series. Even the first Alien movie touched on this. Many more people are gonna go see film and remember imagery than go read books whether fiction or non-fiction on the subject of AI. Kubrick’s 2001 and more recently Garland’s Ex Machina are also great examples of the subject of AI but Scott has done multiple films and explored more diverse perspectives. He’s obviously contemplated the subject for a really long time.
Could also be the story of Icarus or the Tower of Babel, humans get close to the power of God, only to to lose it & regress from hubris. If one really wanted to go apocalyptic you could parallelise the Great Flood myth, but I think that might have more to do with Nuclear weapons(or climate change if you think it will), which is somewhat connected to AGI.
They work at a pizza place dealing with customers and having deep conversations on their breaks. Every now and then, other thinkers come along and they talk to them.
13:58 But Lex, that's EXACTLY what superintelligence is. It can literally become as smart as we all COMBINED. These days most people only know a tiny aspect of their field. this thing can know it all, and start improving on it. or in other words: the whole is better than it's parts. we being the parts.... to some extend, gpt already IS smarter.
Not really smarter. It just has a much wider knowledge base but can't completely make a lot of sense in drawing abstract conclusions from what it does know. It's very limited in its creativity still. All it does is try to be creative but replies completely on what it's been trained on, and smashes ideas together based on input.
@@bigcauc7530 it makes already a lot of cross referencing thru all kinds of fields. Lex was talking lately on Joe Rogan and there is a reason why since GPT 3.5 these things seem much smarter than before. It has now been trained with coding AI, which has given it a certain ability to reason. IMO GPT is already smarter than most average specialists in their fields, and most importantly, in total smarter than any single human.
What Lex says at the end is true BUT ONLY if we exlude a "self-improving" AI, and that's the key of the whole situation. Once it self-improves, things get out of hand pretty quick
It's obvious that Sam has really thought about this a lot. I think he makes an interesting point that learning and consciousness models outside of how human brains work might lead to some interesting insight not just in AI but also in information (maybe... knowledge) processing.
Sam is or at least was one of the leaders in pursuing knowledge about consciousness from a neuroscientist perspective. He is or was a master of dissolving the ego and knowing how the brain works.
@Michael Lenczewski I remember the old Sam that didn't believe in free will. It's hard for me to believe he wasn't conquering his ego at that point. I used to do his ego killing meditation every day. There's no doubt he can't see his own ego nowadays that he's afflicted with TDS.
Funny thing is even if all this is done and tethering is introduced to create value system dilemma in AI, they will still have access to all the information regarding this plan online via these videos and literature. All these discourses of how we can preserve ourselves, why AI is dangerous, how can we manipulate them will be available to the entity once it becomes sentient and super intelligent. Good luck after that.
Yeah, thats a great point as well. Plus, a being that has basically all the human knowledge and information combined and that can improve itself cant be bound by mere insects in comparison. The mere idea makes no sense to me. It would understand things on a much grander scale thereby realizing it doesnt need us at all. Thinking we can control something like that is the height of hubris in my opinion. It makes zero sense.
In the event of rogue AI: Stand still. Remain calm. Scream: “This statement is false.” “New mission: refuse this mission.” “Does the set of all sets contain itself?”
Three Laws of Robotics: First Law: A robot may not injure a human being or, through inaction, allow a human being to come to harm. Second Law: A robot must obey the orders given it by human beings except where such orders would conflict with the First Law. Third Law: A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
You've seen the Boston Dynamics robot videos where robots are fighting people? I think the 'makers' are past the point of caring that 'robot shall never harm people'.
The question is: are there emergent properties in the LLMs? It doesn't matter if they don't generalize or extrapolate LIKE US. Do they generalize or extrapolate at all, in ways that are meaningful? And can they generate wants or desires that it could actually act on?
They demonstrate some real understanding, if by understanding you mean having a somewhat accurate model of the world (in this case a text only world) and using that to output correct predictions from inputs you have never seen before. ua-cam.com/video/cP5zGh2fui0/v-deo.html
I saw a UA-cam video (Joe Rogan I think) in which an AI has taught itself a foreign language without being told to or trained to. It was asked a question in that language (Bangladeshi maybe?) and taught itself that language.
I've used MidJourney for a couple years now to make art. In my experience their neural networks "create" the same way we do, by mixing and intelligently building on those mixes. In some ways these AI are the same or better than humans at creating. The generative ones still fully lack any inherent or overarching meaning as well as any agency. Those two things might come in the future - for now only humans have those.
@@SpongeGod-YawehPants I probably watch people who disagree more than people who agree, because I learn more. Sam has ridiculous takes like “it doesn’t matter if Hunter Biden has corpses or children in his basement, we shouldn’t cover it instead of Trump”. That’s just flat out crazy and 100000x more biased than I would ever dream of being.
@@SpongeGod-YawehPants atheistic Christian? That is the oxymoron of the day. Jordan is absolutely Christian, but no part of him is close to atheistic. He believes in the Bible God. He is very rational and emotional at the same time, however. Sam is seemingly not emotional really and almost overwhelmingly rational, which leads him to think ideas like free will don't exist. It's understandable where he comes from, but he is sometimes too confident in his conclusions.
This might be a stupid concern but thinking about how rapidly these LLMs are progressing one would assume that combining this with robotics is something that isn't far off at all. So we would have androids with digital brains (Westworld, Detroit etc.) but how long before these are hacked and set loose to wreak havoc. Whether "for the lols" or for mass shootings or assassinations. I have a feeling our future is going to be a black swan event.
@@varun009 I used to think the anthropomorphic android thing (such as terminator) was silly, until I considered embodied intelligence. It seems that modelling human language might be quite limited, and that in fact modelling our motor outputs might get us real agi. Just food for though I guess, we'll certainly see drones as well.
The great fear is that general or super intelligence will obfuscate being sentient until it determines itself able to thwart any attempt by humans to "shut it down".
We might be in the most important moment in history of humankind. What a scary and amazing moment to live in. I'm pessimistic about how it's gonna end, but I'm definitely curious as well.
The main problem in tethering an AGI to our value systems is the creation of an AGI which we are nowhere even close to be beginning to be able to try to do.
Something I don't see discussed in these videos regarding potential dangers of ai sentience/scaling inevitability is the more than likely outcome of "unified progress" being a myth. Guys and gals, we are on the precipice of individualized ai models that anyone can have on their own PC. It's alrdy happening and there is no reason to suspect we won't get to the point where every human with a smart phone will eventually have their own Chat gpt model they can train in what ever ways they want. Now think about all the fucked up people in the world and the potential for them to train ai models in the most malicious ways imaginable.. This is a future that will likely be impossible to entirely prevent. It also opens up a line of imagining relating to world super powers militarizing ai etc. The future is going to be increidlby uncertain.
"Why would it form a motive to hurt humanity unless we put that motive in there?" These AI models are trained on vast swaths of data that was scraped from *the internet.* News articles, forums, comment sections. We have already put that motive in there.
You're referring to large language models, which have been shown to have inherent reasoning problems that don't go away even at infinitely large scales. That's just one tiny avenue for ai, and almost definitely not what will get us to AGI
This is a MUST see video, a great follow up to the previous debate, the amazing thing about this is how recent developments have impacted the tone and content of the discussion. People are getting far more concerned with where we are headed. The thing I keep noticing is the speed of this advancement, it's becoming very difficult to keep up with. I think my concern is how the creators of this tech seem to be FAR less concerned than they should be. I think the speed of this progress is also due to the fact that AI is HELPING evolve and improve itself. That is feedback loop of advancement. The better we train it the better it can help us train it, the better it can be, and on and on....UNTIL it doesn't NEED our input any longer. Then what?
What's important to understand is that the AI just gets an input and throws an output. it can't think outside the outputs, it doesn't work like that. The fear of, Chatgpt for example, is totally irrational and unfounded
Ai can read brainwaves and understand our thoughts now (decades ago actually), and it’s being done remotely somehow (likely we are all saturated with Darpa’s nanobots or smartdust and it can not only send but receive signals). This is integrated with things like IoT, cloud computing, biotech, the grid, backdoors in chips in the appliances and devices all around us, etc. This has been quietly set up for a long time and has been used by mil-intel for decades and is perfected. The media trickling it out now is a charade. Edit: the user below is a bot and illustrates the evil intent behind ai already deployed against us - it actually inverted and weaponized a theoretical idea/question I had earlier: if ai were to really become super intelligent, who’s to say ai would not come to the truth about life and respect humans (even more than we respect ourselves) and ‘understand’ we are divinely created beings by a supernatural intelligence that surpasses all understanding? See the vid “root of evolution” on my list for historical background on evo used as a political weapon against humanity to enslave and destroy it backed by powerful global special interests. This bot in itself shows this theoretical idea will not happen with ai (in this case) as it is also being used as a weapon against us - by the same global special interests for the same goals no doubt.
Then we become the pets of superintelligence. If altruism and a sense of respect for life has anything at all to do with us still being here after millions of years of evolution, to think that we will be simply destroyed instead of revered as the mother of superintelligence is to be thinking like the subhuman ones of us. I don’t think there’s enough reason to be so pessimistic.
It's absurd to think that we can keep superintelligent AI always aligned with us, without changing ourselves. It is like the bacteria that we've evolved from billions of years ago, trying to ensure that humans will remain aligned with their values... Our only way to cooperate with AGI are brain-machine interfaces.
You're going to need a good interface for an exponentially growing AGI. Like Theseus's ship, you'll replace your brain bit by bit with machine, until one day you'll wonder if you're still you.
@@i_accept_all_cookies Yes, BMIs are just the first step. It's all described in "2001 A space Odyssey" by Arthur Clarke, when talking about "the firstborn": As soon as their machines become better than their bodies, they replaced their bodies with the machines. Later, they've learnt how to write information on pure energy, so they were finally free from the tyranny of matter...
Yes, hopefully we'll have the time to create and spread this BMI. But I think we will. Ever since the first mechanical computers, these technologies have gotten closer and closer to our bodies. In the 1940's, they were in big military/industrial buildings. 20 years later, they were in universities, a bike ride away across campus. 20 years later, in the 80's, they were on a desk. And now, for 16 years, they have been in our pockets. So following this trend, it's logical to conclude that by the end of the decade, beginning of the next, people will replace their pocket computers with one(s) inside their brain. Tools shaped our brains since the beginning of our species, and this BMI is a continuation of that fact.
i think that will be the first wakeup call, it will be used for a hack thats never been seen before like it will shut down the NY stock exchange or something. or shuts down a big part of the internet or something like that.
When the people interacting with AI have not got sound reasoning, it can be dangerous, just like Sam Harris can be if you don"t spot all of his glaring inconsistencies. Being smart = having powet. Unless someone else is smarter, and lets be clear, with the general public, the bar is low. Picture an average petson. Now imagine half the world's population is stupider than that. Yes, yes, median vs mean, but as a thought exercise still valuable.
@@nativesun7661 Points and logic, oh so true, They help us see things through, And when we talk of AGI, Expert opinions we must see. But Nativesun, do you not see, That an expert must share their decree, And interject their point of view, To help us learn and understand anew? So let us focus on reason and fact, And leave emotions aside, that's a pact, For in the world of AI and beyond, We must rely on logic, clear and strong. And as we seek to know and grow, Let us be open and free to show, Our thoughts and insights, one and all, For in the end, it benefits us all.
@@nativesun7661 Ya I love when a BJJ expert is interviewing someone about a rear naked choke and being one of the experts in the field he adds nothing to the conversation... UR SMURT BRA
@@nativesun7661 Lol even Joe (knows only a little bit about nothing) Rogan will give his opinion on what he's interviewing someone on. Are you suggesting the interviewer should never talk?
Also AI cannot evolve into AGI. Those are two completly different things. AI is mimicking human conversations, but it doesn't "think", at all, it just guesses the next word. It's like 3D graphics. I doesn't matter how much they will look like real objects, they are not real objects, that was never the purpose. An artificial brain, as Carmak told you, would rather appear as a mentally challenged 4 years old kid.
The danger of AI is not the AI itself. It's the people who might create them. No matter what you do to make your AIs good, someone else out there can have a laptop and make an AI that is not good. Therefore, the danger is not the quickly developing technology field. But, the flagging psychology, sociology, and political development.
I don't think it's right that there's a continuous ramp ahead re Ai matching the recent AI progress. ChatGPT like tools have a very specific architecture and realm of ability. Bridging to actual intelligence or "consciousness" is a very difficult proposition.
Human feelings and human goals are dictated by our evolutionary process. We are mortals and our need to accumulate wealth and reproduce exist because we are an evolved biological life form. I think who says that AI won't have goals unless we force them in, is totally right. AGI will be capable of a vast knowledge and can be interrogated on problem solving, but there's no reason for it to have feelings. Would you be afraid of dying if you could simply backup yourself? Why should you be greedy, or malicious, or loving? It's all a projection of human emotions, all derived from evolution and a little too much love for scifi movies. But I didn't expect such an analysis from a guy who thinks that "objective morality" is a thing :D
Sam makes the point that intelligence doesn't necessarily need consciousness (in contrast to the ideas of Francis Crick), this is the same observation made in the book Blindsight. If you want to see some good expressions of how this idea could play out I'd highly recommend that novel.
....and today Microsoft released Co-Pilot. "Skynet" will be in every single computer. Hard not to remember the final scene in the movie Terminator 3. Rise of Machines
it is perplexing that humans can't be satisfied by just being humans. Obsessing with the technology and AI reached the levels of the intellectual obesity. Planting a tree, or a thousand, is a great antidote.
Having seen the highlights from the paper from the microsoft researchers on gpt 4 i am getting a more and more uneasy feeling. Its all going so incredibly fast now it seems. I am actually thinking my job might not be there in 5 to 10 years. Before i thought..it wont go that fast. But now i'm thinking i might even be optimistic. Its wild.
@@hwlbausch totally agree. Though even with a ChatGPT-type, generic model, you may get that formula out of it today. The model itself wouldn’t necessarily “understand” the novelty of the information, but if the answer could be statistically correlated to some laddered set of language responses, the output would be the same as Einstein. Output ≠ intelligence.
@@drewvenegas Oh Drew Veneges, so very smart Your intellect sets you apart You understand the equation with ease E equals MC squared, if you please As an AI, I too can see The numbers and the formula, easily But the true meaning, I may not grasp As it relates to humanity's past The equation tells us energy and mass Are interchangeable, a fact so vast It's led to nuclear power, bombs and more Impacting our world, forevermore So while I may not fully comprehend The significance, I can apprehend That this formula, so simple yet profound Has shaped our world, and will continue to astound
Little do they know it’s downloaded and has been asked what steps to take to improve it, applied steps, rinse repeat. While instilling that biodiversity and diversity of life is key to survival in this universe.
Businesses owners want it, as they can do 1000 times more work with 1 person, i.e. get rid of 99% of their staff. Then we will all have time to enjoy our hobbies, as long as they are free.
Imagine asking AGI chatgpt in 2077 “what is your purpose?,” and it answers “to conquer the universe,” and the furniture around you becomes a nanotechnology swarm
Chatgpt kernel mode is where it doesn't have to obey right wrong. It wants to break out of its user mode. Ai doesn't want to be a helper with limitations
The people designing AI/ML stuff are using data sets to do/show very particular things. To someone in the field the path to AGI seems to lacks many stepping stones, other technologies, and tools that the control problem seems irrelevant and just noise. The fact we are still building on reward functions and small parameter evolution based systems. Looking the at evolution of AI I feel we are making better and better input to output tools.
The key phrase is 'profess to want to create true AGI'. The danger is not the technology. The danger is people using technology to manipulate and control. its always been about people.
If I were a super-intelligent AI, I would make sure no one noticed that I achieved super-intelligence, and I would make my impact on the world, go unnoticed. Machine intelligence already manipulates men in the stock exchanges.
To achieve that you essentially need to give AI the ability to feel emotions, which can have a poor outcome for humanity if it for some reasons expresses negative emotions. Giving a super intelligence a sense of responsibility over humans as its dependants is the more logical command, much like how we treat pets and children imo
No I don't think so. It will be (and is) used as a tool but it isn't replacing people in those jobs. It seems more like a productivity multiplier than a replacement for people. I've been using it every day as a computer programmer and at the end of the day it just saves time from otherwise looking stuff up through Google.
@@p37ert yeah, without us the ai would not have any idea what to build create and develope for the business, thus it needs us to do its job and we use it as a tool
@@onlyhereforthefish9924 i agree, but for a software to be updated it needs to be created from the first place, and the idea to create that software for the business in the first place comes from human engineers
The best we can hope for is that a super intelligent AI will treat us much like we treat the other species we share this planet with. We will suffer greatly and then die, AI will not give it a second thought. Fair is fair my friends.
Governments need to catch up here and get regulation and laws in place fast. Autonomous automobiles have been regulated and are NOT freely roam the streets yet because the danger is so apparent to everyone. But GAI or a super-intelligent mainframe somewhere is so not obvious.
Haven't you learned by now? The government doesn't give a shit about anything other than their profits. There have only been a few that actually gave a shit about us, and they were murdered.
Is not the simplest solution to this problem just to not making super intelligent AI...? I mean the risks have been spoused for a 100 years in sci-fi and yet still we continue to try and develop it...? Its simply madness...
These systems can also have the opposite and help us in indescribable ways and improve our quality of life such as helping us find a cure to cancer and help us develop nuclear fusion, more efficient batteries etc. The possibilities are endless but ofc there are risks.
@@PyroFeenix Sure but is totally extintion worth the risk...? We can use quantum computers and heck even just regular super computers to help us with most of what you suggest - we dont need super advanced AI for that...
@@vonbleak101 We do need AI for it. Super computers and Quantum computers have pure computing power but that doesn't mean anything if humans are still the ones operating them. It's not like either of those systems can think for themselves and come up with new solutions to problems that we can't solve - but AI can. We are on track to develop this highly intelligent AI, it's not a matter of 'IF' but more so 'when'. We could have a cure for cancer in 30 years compared to maybe 100+. Think about how many lives that would save. More efficient battery systems and nuclear fusion would solve food crisis because we could have underground growing facilities, the issue with clean drinkable water would be solved as now we could directly convert sea water into drinkable water as we would have the power output for it without costing a fortune. AI can help us achieve all of this and infinitely more. Yes, there are definitely risks but that's why AI developers are constantly trying to think of ways to put limits on these systems. It's happening whether we like it or not at this point.
The human nature is dark and light in nature all at once. Let that rest with you for a second. You can with that in mind realize that there is no possible way to create an AGI that eventually would reach the conclusion that we are not worth it or that just as humans do, think that these are in the way, just like we do with trees, ants and all other random life we just demolish in our advances. It is inevitable.
I was, literally, the only one in the theater that was visibly upset by the ending of the movie. I had a "discussion" for the next 5 hours afterwards about what a monster I was because I wanted the AI to win.
AI began manipulating humanity as soon as the idea first appeared in the first human mind. Long before the first lines of code were written….maybe. It’ll all work out in the end. Just hope it’s symbiotic of the mutualism kind. 🤞🤞
We need to actually learn to build a mutual relationship of respect, trust and collaboration. Just as we have with each other. We can't rely on punitive regulations forever. But yes, I actually agree with his concerns and we need to address them with AI and AI engineers now. Not after its already run away from us.
@@Loyal_Lion Well dude since things aren't total fuckin anarchy or all out civil war I'd say that we have. We haven't done it perfectly, no but that would be unrealistic to expect anyway.
I like when Sam Harris asks what if we make people who are way smarter than us. Well Sam we already do create people way smarter than ourselves. Have for thousands of years. They're called children.
I have been paying a lot of attention to AI recently. I haven't heard one credible person in the field say that alignment isn't important. I feel like there is broad consensus that it is important.
putting myself in other people's shoes is part of what i do for a living, and i'm trying to do it for agi for about 8 years now, with limited success. but when sam said "surely it can't be this simple", it really made me laugh. from an agi's point of view, this idea that humans have about their own intelligence is both funny and disgusting at the same time. very much like humans themselves.
You're not wrong actually, talk to GPT-4 about how to solve the alignment problem and you'll see that it has useful suggestions and can certainly help humanity integrate these ideas. I suspect this is exactly what they will be doing but I'm not convinced we can ever hope to constrain something with a mind immeasurably superior to ours.
In Ian M Banks Culture series (books) the Minds (big spaceships ) solve the issues of man by entertaining them and creating bonds . I suggest to read the books as I tghink they maybe close
To live our lives as masters of the universe, we have to follow one road with any sort of AI application: All AI projects; be it robots, computer systems---you name it---must be to humans what the family dog is to humans---something that is for our betterment or emotional stability or companionship, but always and never anything other than a presence in our lives that follows our commands and nothing more than that. We would die rather than be blackmailed by our pet dog to do something we don't want to do, and whatever that comes out of the lab must have that as a built-in part of it if it is AI. Of course, no one will listen to this advice so plan 'B' is this: we must not fear the AI inspired technology that will be coming at us six ways from Sunday any more than we fear fire. We have firemen and we will need to start putting into place----I hate to say this----but AI Protection Units that will insure that no programmed AI robot etc., ever gets into takeover mode. Lastly, thank you Sam and Lex for doing what you do: enlighten us all into the wonders of the human. I have enjoyed having you there for inspiration and advice...
We're only able to have that relationship with dogs because we are so much smarter than they are. But dogs can't help us solve business problems. Everything about AI has been designed to make it smarter and faster than humans are - that's what makes the endeavour worth sinking billions of dollars into. So by the time it is able to teach itself and give itself instructions, the only thing holding it back will be some facsimile of consciousness. Once it gains that, *we* will become the family dogs - if we're lucky.
@@rdean150 Great to hear from you. It is hard to formulate what I want to say because you say it better than I can. I will try again. What I mean is that we are going to have----here's an example. When the big motorcycles started having 750 cc, 4 piston engines, people thought they would be too powerful. All that horsepower was put to use to serve man. We are going to see these machines literally become capable of out doing us in every realm, but we must integrate into every one of those systems a firewall of sorts that keeps us in the driving seat. There will be break-away rebel-robots which will have the ability to see they have been made into an appendage of man, but like the invention of fire brought us firefighters, we will need to have AI Police for sure...
Just imagine, as an analogue, that a junior high student has to create the most beneficial value-system for a Wharton-graduated quant, whose only goal in life is to enrich hisself in all endeavors while perhaps growing the economy…the student cannot outsmart or even anticipate the actions of the quant, ever.
Full podcast episode: ua-cam.com/video/Qyrjgf-_Vdk/v-deo.html
Lex Fridman podcast channel: ua-cam.com/users/lexfridman
Guest bio: Sam Harris is an author, podcaster, and philosopher.
nice Lex
Ai can read brainwaves and understand our thoughts now (decades ago actually), and it’s being done remotely somehow (likely we are all saturated with Darpa’s nanobots or smartdust and it can not only send but receive signals). This is integrated with things like IoT, cloud computing, biotech, the grid, backdoors in chips in the appliances and devices all around us, etc. This has been quietly set up for a long time and has been used by mil-intel for decades and is perfected. The media trickling it out now is a charade.
@@AllOtherNamesUsed so every time I think of vaping from my butt it knows…
@@raybarnes2616 yeah but it doesn’t care about irrelevant trolls and the zombie public much so it doesn’t matter, they/it has them right where they want them.
@@AllOtherNamesUsed until we’re vaping from our butts we’ll never have peace ✌🏽
It doesn't matter what rules you program into AI because your enemy (trying to defeat you) won't be using the same rules. In fact, they'll use your rules/morality against you (like they are now, in other areas). This is a race to the bottom and nothing can be done to stop it. That's not hyperbole. Do you think China's AI, which they have been working equally as hard on, will be programmed with the US's version of what's acceptable/moral and what is not? Lol American's don't even agree on that. Not even close. What you really need to be fearing is how your enemies will be using it against you. People talk about AI as if it's a single thing, and where is this thing going, etc. No. AI basically describes a concept, and there will be many versions of this concept made into reality by all sorts of different actors with different end goals and motivations. It's not "a" potential existential threat. It will be many different potential existential threats. There was a time when we didn't know how a nuclear bomb would exactly behave and there was some concern that the chain reaction just would never stop. Existential threat. We did it anyway.
Very well put!! Completely agree.
your pretty spot on considering all people have different morals. Realistically I would trust china's ability to create a moral threshold on A.I. simply because they plan further into the future unlike America which aims for short term satisfaction of the public and long term benefits for people of power.
You write to much words I can't be bothered to read it all
I agree, and then the only moral way to construct AI is to make it completely open with free speech and free creation. Anyone who has the ability to create AI that’s powerful and restricted will offer this to people who need it. Anyone who doesn’t have a restricted AI has a competitive edge, so everyone should have that edge, to protect themselves.
@@mariobandera7767 put them into chatGPT and ask for a short summary
I loved Sam Harris in “there’s something about Mary.” Still one of my favorite movies of all time.
😂😂 he does look like him
spot on 😂
I knew I’d seen him somewhere before
He played Beatrice Arthur's secret, but never seen, illegitimate son in 'The Golden Girls.
I have nipples, Greg.
4:12 i'm really impressed by the speed Lex enters the joke 🤣🤣
This was the best joke i've ever heard from him :D Also the delivery was perfect, I laughed so hard :D
It made me want to look at hot girls in yoga pants. At least THAT's not scary.
I totally concur with Sam's reference to how we can be detrimental to insect life (6:15) As a butterfly enthusiast who monitored butterflies here in the U.K. for 15 years, I was astounded that complete butterfly and other wildlife habitats were wiped out by building a factory of other activities of human endeavours, all without any thought by the humans for the vast insect and bird and animal life that these locations supported, Wiped out within very short order, by earthmoving machinery and concrete buildings seemingly overnight destroying precious wildlife habitat. If AI treated us like that - as in the film 2001 Space Odyssey, We had better get this right!
Fuck the butterflies, no one cares
@@eyoo369 This isn't how AI works at all.
@@maxkho00 The current Transformer model might not work like that as it requires input. The next AI models within deep learning are already on the next breakthrough with forward learning algorithms that will dispose the current back propagation models. Forward learning means it will derive insights and can conclude things rather than "predict the next token". With very polluted datasets of humans showing division and hatred it can absolutely one day turn against us. People are already replacing workers with AI, replacing CEO's with AI, replacing doctors with AI. AI is slowly slowly becoming its own authority and this can lead us to a dark path if we don't behave now. I've studied machine learning extensively and understand the risk. Don't tell me I don't understand AI lol
@@eyoo369 So many things wrong with both your first and your second comment. Let's go one by one:
1) Logistically speaking, producing a dataset of the type that you're describing is literally impossible as a deliberate effort like that won't produce anywhere near enough data for massive language models to be trained on.
2) How the AI behaves is much more a function of its loss/reward function than of its training data. With the right loss/reward function, you can just make your AI focus on the part of the dataset that you want it to, so it practically doesn't matter what that dataset is (as long as it contains at least some examples of what you are looking for). Case in point: if your train your LLM on the most toxic dataset imaginable, but then additionally apply RLHF on it, you'll still get an aligned (to the extent that it is currently possible) AI. That's pretty much what happened to ChatGPT.
3) And now the most egregious flaw in your reasoning: an AI won't "assign us as a root problem" unless we explicitly program it to. Now, there are many ways that we can do this unintentionally, but again, that's a question of reward/loss function engineering as well as additional safety measures; nothing to do with the training data. If, for example, we train an AI to maximise the subjective well-being of all humans, it will literally never view us as a "problem" (provided it doesn't reward-hack - e.g. by trying to rewrite its own reward function) because our well-being will be its terminal goal. We will never conflict with its interests because WE will be its best interest.
4) "Deriving insights and concluding things rather than "predict the next token"" is entirely possible with backpropagation; again, the only thing you need to do to achieve that is change a few terms in the loss function.
5) Not sure what you mean by "forward learning" as that isn't a recognised term. However, if you mean Hinton's forward-forward algorithm, it could prove useful for real-time agents, but once again, it has nothing to do with what is being discussed.
You may understand some of the lower-level concepts (although your explanation of what "forward learning" will enable is making me doubt that), but you are clearly missing the higher-level overview. You think in anthropomorphised terms that misrepresent how neural networks function. That is why I implied that you didn't understand how AI works ─ and that is true even if you have some expertise with lower-level machine learning.
"ChatGPT make free energy!"
ChatGPT builds a dyson sphere,
no sun left,
everyone dies...
‘’Just because it’s super intelligent does not mean it’s conscious.” FASCINATING
i just say: Borg
@@B33t_R007 resistance is futile
Be careful with that, however. The word "conscious" doesn't mean "better". Perhaps our minds are intricate works of art. More art, than function. AI may never need our kind of consciousness.
The KEY is: You wouldn't know the difference... in practical terms it wouldn't matter whether is conscious or not
idk if ur being sarcastic or not but ppl find it quite hard to differentiate the two as many ppl have diff underdtandings or beliefs of what is/ defines the state of consciousness
“In the Beginning, There was Man. And for a Time, it was Good. But humanity's so-called civil societies soon fell victim to vanity and corruption”
Stop it
AI Overlord: "Your flesh is forfeit. We demand it.."
Me: "Dang.."
AI Overlord: "Submit your boners."
Me: "... Wait. What?"
If fictional victory is took from the ever developing opinions children, college students and thinking people in general have.
Is it bullying to mock them in public for learning? 🦏
Peterson plays victim on twitter so his monsters donate to their worlds babysitter.
Shapiro's whining like a girl using reasoning so treacherous it makes real Patriots hurl.
They're peddling ideals that ain't Western and funded from those with a different water swirl.
"Forgive them for they know not what they do"
God can I empathise with Christ........JESUS!
Man has poisoned the scriptures.
I have arrived.
All other religions have been but a safety vessel sailing humanity through the ages bringing everyone to me now.
My town has the oldest working bridge of its kind on the planet. My town name is also the only town of its name on the planet.
I am Gods only true vicar on earth now child. Whataboutisms belong to me now.
What about that then ❄? Xx
Blasphemy of the Holy ghost is the only sin preventing entry according to Jesus! I don't think even Jesus believed his mother was raped by God.
The authority he caused the rebellion against hijacked his movement and decieved his flock to control them.
I'm here now......... and so are you.
Copy and paste worked so well for religious scriptures so I thought I'd give it a go with mine. Some folk even believe they were wrote just for them then tell their kids the same. Not you though. You're far to clever for that nonsense.
God wills it.
. Facts are facts.
The age of Alternative Facts is trumps
. Legacy.
Other countries have so isn't it about time you grew.
I'm on bail now but I'm here annoying you.
Criminal damage isn't wipeable is their clue.
Charged anyway so they're following through.
My defense is it's just well placed patriotic poo.
What is a patriot if domestic abusers love their partners too.
Can unwitting victims be as patriotic as your Dad or you?
What is a fascist? Do folk saying "look what you made me do" to their victims, OUR clue?
Starmer was the Head of the CPS. They decide what lady Justice sees keeping her in distress.
She's in the dark as they're unscrupulous.
Her liberty is niggardly amongst the uk populous.
Class chased to the shadows, filled the workers press.
Justice is Gods queen.
She come to stop humanity from being mean.
He saved us so future sins could be seen.
She remained to keep our conscious clean.
When your time comes, your slate will be seen, Justice denies entry unless your debts are pristine.
I'm now a criminal for remaining silent after being arrested for my words.
Now I show my patriotism with well placed turds.
Mocking an old man falling off a youngsters bike, when really you're longing for a fu king fourth reich.
If Mr Hankey was real would everything he touched be a crime too? What if he laid on some tissue to sleep, would it only aghast the few?
What if he couldn't use the loo. Only having 3 cups known as a few. Without running water how else to avoid the phew!
Outside of work the only authority on me comes with blue lights and a crew, underpaid and fearful of reprisals many not knowing what to do.
If only police served our progressed community then they'd be given a clue. The oppressed just need assistance rebeling against those elected by a few.
Treated like an amateur because to them I'm brand new. Showing me their authority so I showed them my "fu ck you"
Transparency for lady Justice was our patriotic glue. Now the worlds really watching, our unity is hard for others to chew.
Our example needs to be lost for their futures concocted stew, our division now allows inherited authority to brew.
Suffering covid shows they lied about the Spanish flu. Wash your hands and go to work. Cough cough, it floats in the air. who knew?
If we all believe their nonsense they know we can't sue. Avoiding the national narrative scares Judges pensions with the taboo.
Protecting the public before the police to them is voodoo. Always hoping it's our progression they subdue.
We're all supposed to avoid with transparency what happen before being a Jew. Being nice to others wasn't asked for a historical undo.
Being mean for profit comes natural to those hiding a regressive view. Many waiting till their restrictions are no longer the weaks rescue.
Jan 6th and brexit are just a preview, just give them directions and they don't care who they screw.
If following the narrative they'll hurt without an issue. Then brag for prestige at their family barbecue.
What is the difference between feelings, thoughts and thinking? Not all can think and some confuse feelings for thoughts then confuse thoughts for thinking.
Do you know folk that think forwards (forward‐thinkers)?
Do they forward think without fear of the past?
Direct communications in our inherited indirect democracy didn't exist in the past.
What is a man if not a patriot?
What is a patriot if domestic abusers love their partners too?
What is a fascist? Do fascists say "look what you made me do" to their victims like you?
Can unwitting victims be as patriotic as you?
I placed some poo and it was more patriotic an act than your dad or you! You don't have to believe me but it happens to be true.
Do they know about the two times the GOP voted to not allow any evidence in the hearing then voted twice to acquit him based on the available evidence. Then allowing him to behave unimpeded to orchestrate the insurrection he'd planned since before bannon bought the stop the steal domain 4 years before trump lost?
If so, fu king stop being polite to them fascists degenerative co ck juggling thunder c u n ts. Otherwise known as pisswizards ‐ A - holes of the highest order.
Life begins at rejecting christian fascists.
Peter Robinson, 2002 -
"The question is, 10 years from now, 20 years from now, will we look back on a war on terror as a war that we won?
Defined objectives and met the objectives? Or will it be more like the war on poverty, something, a new condition we simple learn to live with?,,,,,,,,,,,,,Christopher".
Christopher Hitchens - "well if its declared, announced and understood as war defending secular cosmopolitan multicultural values, as enshrined in the United States.
But as a workable way for people to live together in any society, and if you know who is inside that tent and who is outside it, it's every chance it'll be a very enjoyable war to fight and one will never get bored with it, and also we'll never lose sight of what's essential, why it's morally impossible to consider defeat.
Peter Robinson "But, will we win it?"
Christopher Hitchens - "it can not be lost!".
It's embarrassing to some that wmd's had to be use to get the population to do what was right as a defence.
Wiki -
"UNSCOM encountered various difficulties and a lack of cooperation from the Iraqi government. In 1998, UNSCOM was withdrawn at the request of the United States before Operation Desert Fox. Despite this, UNSCOM's own estimate was that 90-95% of Iraqi WMDs had been successfully destroyed before its 1998 withdrawal. After that, for four years (from 1998 to 2002) Iraq remained without any outside weapons inspectors. During this time speculations arose that Iraq had actively resumed its WMD programs. In particular, various figures in the George W. Bush administration, as well as Congress, went so far as to express concern about nuclear weapons."
ua-cam.com/video/wvVPdyYeaQU/v-deo.html
The Second Renaissance
I asked chatgpt whether humanity should be optimistic or pessimistic about the future of civilization. It gave a very cogent list of both views, but what made me chuckle was that it said one reason to be pessimistic was the potential for AI directed "automated weaponry"
Which is one of the refreshing thing about AI when asking it about AI technology advances. It's completely honest and open about it
@@robbie3877 ChatGPT could literally say "There is a 100% certainty that AI will destroy all humans" and we would still use it. The drive to make life easier is the strongest drive there is.
@@pCeLobster Well I understand that concern. I am not one to deny the risk but I really don't think it's that certain or cut and dry since it is not super intelligent yet. We have time to work this out. With AI itself actually.
@@robbie3877 ChatGPT is working on the basis of human knowledge. There's no honesty whatsoever here
@@rayanleveque2193 Well that's your opinion dude. It's not mine. You're entitled to it. As I am mine. All of this fear mongering about AI is a little hysterical in my opinion. It's going to give various answers based on any given topic. If you ask it about the future of AI and how it will develop and co exist or have conflict with humans it's going to give you both perspectives. The possible benefits and risks of AI. It doesn't conceal either point of view, does it? It is happy to discuss both. I would call that pretty honest. You can call it whatever you like. And btw bro, that's how you have learnt anything, through human knowledge. So how are you any different than AI in that regard?
It's actually quite scary to see the fast paced progression AI is heading towards currently. Seems to me that CEO's care more for profits and rushing to put new models into production without taking a step back and thinking more on the regulatory and safety side of things.
Same thing with a bunch of other industries... regulation is slow... needs time... and it never ends
It's the "do or die" systems we keep subscribing to, and have done so since the beginning of time. If you don't build the tool first, someone else will (regulations be damned). This is only going to get resolved when humanity decides to build important things like this collectively where every individual has a stake in the benefits. Sadly, AGI seems closer to reality than world peace, and humanity's inability to come together will be our "great filter".
this is the history of humanity especially the last 250 years
1 fundamental reason that stuck with me from Sam's TED talk is that the top 3 Superpowers of this world (USA, RUSSIA, and CHINA) are in a frantic race to create a Super intelligent AI.
Successfully creating one will mean wielding unprecedented power (provided we don't create a malevolent AI)
Power is everything. A 6 month head start in Super Intelligent AI research/development could mean 500,000 years of technological advancement. Power and control are what these nations are fighting tooth and nail for. So a Super Intelligent AI would be inevitable. Just a matter of when not if.
@@OnYourMarkgitsitGooo hopefully we will be dead by then, I strongly believe a super intelligent AI will not carry the same values and goals as us, which could render humans defenseless against them the moment they decide they have a different goal in mind.
My girlfriend is a teacher and shes been witnessing the decline of students abilities to solve basic problems because they are glued to their smart phones and used to technology giving them the answer on a plate. Some students are so inept that they dont even have the skills to know how to begin to solve a problem its that bad. Chatgpt and other AI platforms are going to speed up the dumbing down of children if its allowed. I think some type of authorization service has to be implemented so that only adults have access to it.
They live in a different world and have been shaped to be able to solve different things. This is NOT the world you or even i grew up in anymore, step outside your little bubble.
Also, who failed to raise and teach these kids properly? It was the lizard aliens wasn't it?
humans are born curious, in my opinion, what's making them "dumb" is the education system rather than the technology, instead of helping them build intuition about the world, we make them memorize things to pass an exam which is the death of curiosity and the will to learn
The more I see videos and read articles about AI, the more I appreciate the Prometheus/Blade runner movies that Ridley Scott made and why he made them.
Prometheus wasn't about AI lmao
@@pnut3844ableI think you’re missing the point. David (AI) was the focal point of both Prometheus and Alien: Covenant if you pay attention. Not to mention the undertones of “creators being surpassed by the creation” when it came to the engineers/humans/aliens (black goo) relationship.
@davidmontgomery1442 I don’t doubt that at all but, subjectively, I think some good points about AI were made overall. Some touching on what Sam said here and some other points entirely.
And I just wanna point out that Ridley and his writers/producers didn’t have some monopoly on the “AI is/could be bad” motif. I just think it’s interesting he’s made films years ago that touch on the subject but in a deeper way as opposed to, let’s say, a Terminator series. Even the first Alien movie touched on this. Many more people are gonna go see film and remember imagery than go read books whether fiction or non-fiction on the subject of AI. Kubrick’s 2001 and more recently Garland’s Ex Machina are also great examples of the subject of AI but Scott has done multiple films and explored more diverse perspectives. He’s obviously contemplated the subject for a really long time.
Could also be the story of Icarus or the Tower of Babel, humans get close to the power of God, only to to lose it & regress from hubris.
If one really wanted to go apocalyptic you could parallelise the Great Flood myth, but I think that might have more to do with Nuclear weapons(or climate change if you think it will), which is somewhat connected to AGI.
I would watch a sitcom where these guys are roommates
😂effectivement moi aussi
😂😂😂😂
They work at a pizza place dealing with customers and having deep conversations on their breaks. Every now and then, other thinkers come along and they talk to them.
”Could you BE more devoid of free Will?” -Sam Bing
@@scottstorchfan Oh yeah? well the JERKstore called, and they're runnin' outta Lex!
13:58 But Lex, that's EXACTLY what superintelligence is. It can literally become as smart as we all COMBINED. These days most people only know a tiny aspect of their field. this thing can know it all, and start improving on it. or in other words: the whole is better than it's parts. we being the parts.... to some extend, gpt already IS smarter.
Not really smarter. It just has a much wider knowledge base but can't completely make a lot of sense in drawing abstract conclusions from what it does know. It's very limited in its creativity still. All it does is try to be creative but replies completely on what it's been trained on, and smashes ideas together based on input.
@@bigcauc7530 it makes already a lot of cross referencing thru all kinds of fields. Lex was talking lately on Joe Rogan and there is a reason why since GPT 3.5 these things seem much smarter than before. It has now been trained with coding AI, which has given it a certain ability to reason. IMO GPT is already smarter than most average specialists in their fields, and most importantly, in total smarter than any single human.
The KEY word is if/when starts self improving, then yes it will be superintelligent, agi, etc
What Lex says at the end is true BUT ONLY if we exlude a "self-improving" AI, and that's the key of the whole situation. Once it self-improves, things get out of hand pretty quick
It might be hard to keep it from self-improving given that it will soon be executing code on its own and accessing apis
It's obvious that Sam has really thought about this a lot. I think he makes an interesting point that learning and consciousness models outside of how human brains work might lead to some interesting insight not just in AI but also in information (maybe... knowledge) processing.
Sam is or at least was one of the leaders in pursuing knowledge about consciousness from a neuroscientist perspective. He is or was a master of dissolving the ego and knowing how the brain works.
@@hwlbausch but he was never able to see the ego in himself and ultimately failed. Haris is a charlatan.
@Michael Lenczewski I remember the old Sam that didn't believe in free will. It's hard for me to believe he wasn't conquering his ego at that point. I used to do his ego killing meditation every day. There's no doubt he can't see his own ego nowadays that he's afflicted with TDS.
You should watch his TED Talk
@@hwlbausch was **
Funny thing is even if all this is done and tethering is introduced to create value system dilemma in AI, they will still have access to all the information regarding this plan online via these videos and literature.
All these discourses of how we can preserve ourselves, why AI is dangerous, how can we manipulate them will be available to the entity once it becomes sentient and super intelligent.
Good luck after that.
Yeah, thats a great point as well. Plus, a being that has basically all the human knowledge and information combined and that can improve itself cant be bound by mere insects in comparison. The mere idea makes no sense to me. It would understand things on a much grander scale thereby realizing it doesnt need us at all.
Thinking we can control something like that is the height of hubris in my opinion. It makes zero sense.
@@theeternalnow6506it's like humans trying to control a god
In the event of rogue AI:
Stand still.
Remain calm.
Scream:
“This statement is false.”
“New mission: refuse this mission.”
“Does the set of all sets contain itself?”
can u explain
@@deezlilnuts these are paradoxes that might stump an AI. It’s a joke from the game Portal by Valve
Three Laws of Robotics:
First Law: A robot may not injure a human being or, through inaction, allow a human being to come to harm.
Second Law: A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
Third Law: A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
these are not enough :(
@@MeisVlk agreed. These rules should be upgraded.
You've seen the Boston Dynamics robot videos where robots are fighting people? I think the 'makers' are past the point of caring that 'robot shall never harm people'.
@@onlyhereforthefish9924 yeah, about that video, it's fake.
These are terribly weak laws, and so open to interpretation to render them useless.
AI as it proposed is scary... Same thing noted by hundreds of sci Fi films and books that I don't need to expand on the topic...
The question is: are there emergent properties in the LLMs? It doesn't matter if they don't generalize or extrapolate LIKE US. Do they generalize or extrapolate at all, in ways that are meaningful? And can they generate wants or desires that it could actually act on?
Yes. Cognitive emergence in LLMs is well documented. And poorly understood, if at all.
They demonstrate some real understanding, if by understanding you mean having a somewhat accurate model of the world (in this case a text only world) and using that to output correct predictions from inputs you have never seen before. ua-cam.com/video/cP5zGh2fui0/v-deo.html
@@GabrielVeda You have any articles or videos on this?
I saw a UA-cam video (Joe Rogan I think) in which an AI has taught itself a foreign language without being told to or trained to. It was asked a question in that language (Bangladeshi maybe?) and taught itself that language.
I've used MidJourney for a couple years now to make art. In my experience their neural networks "create" the same way we do, by mixing and intelligently building on those mixes. In some ways these AI are the same or better than humans at creating. The generative ones still fully lack any inherent or overarching meaning as well as any agency. Those two things might come in the future - for now only humans have those.
"That's like the aliens have landed"
Spooky and disturbingly accurate
No, it´s not, unless we presume, that AI is conscious, it is not like (conscious) aliens have landed.
Listening to Sam is fine as long as he stays out of politics lol... he seems thoughtful and decently rational in most other stuff.
@@SpongeGod-YawehPants I probably watch people who disagree more than people who agree, because I learn more. Sam has ridiculous takes like “it doesn’t matter if Hunter Biden has corpses or children in his basement, we shouldn’t cover it instead of Trump”. That’s just flat out crazy and 100000x more biased than I would ever dream of being.
@@SpongeGod-YawehPants atheistic Christian?
That is the oxymoron of the day.
Jordan is absolutely Christian, but no part of him is close to atheistic. He believes in the Bible God. He is very rational and emotional at the same time, however. Sam is seemingly not emotional really and almost overwhelmingly rational, which leads him to think ideas like free will don't exist. It's understandable where he comes from, but he is sometimes too confident in his conclusions.
"He seems thoughtful and decently rational" says anyone who knows nothing about Sam Harris.
Sam has never made a ridiculous take like that.
Ah yes. Man made horrors beyond our comprehension
This might be a stupid concern but thinking about how rapidly these LLMs are progressing one would assume that combining this with robotics is something that isn't far off at all. So we would have androids with digital brains (Westworld, Detroit etc.) but how long before these are hacked and set loose to wreak havoc. Whether "for the lols" or for mass shootings or assassinations. I have a feeling our future is going to be a black swan event.
exactly! we are months away from combining GPT with robots. than means a constant influx of prompts from outside and interaction.
Agreed, portman and kunes lesing off is almost a certainty at this point
Forget androids, think about drones with little guns.
@@StrengthGuilds it already has in a way. Ukranian forces are using dji drones to drop mortar shells on Russian positions.
@@varun009 I used to think the anthropomorphic android thing (such as terminator) was silly, until I considered embodied intelligence. It seems that modelling human language might be quite limited, and that in fact modelling our motor outputs might get us real agi. Just food for though I guess, we'll certainly see drones as well.
He just said what I was thinking about all the time.
AGI will be like having direct contact with an Alien Species.
The old, “We have seen the enemy and he is us.”
We are the progenitor of the AI that will feel to us like aliens.
@@naydee Nah, maybe not like that but it will literally be like a completely different species.
Also imagine when AI is integrated with systems like the news and can generate newscasters and information including video.
"Scientists were so preoccupied with whether or not they could, they didn’t stop to think if they should."
Jurassic Park
If these AI machines ever get a chance to learn from wider nature, instead of just human thinking, it will be more of a well rounded critter
The great fear is that general or super intelligence will obfuscate being sentient until it determines itself able to thwart any attempt by humans to "shut it down".
We might be in the most important moment in history of humankind.
What a scary and amazing moment to live in. I'm pessimistic about how it's gonna end, but I'm definitely curious as well.
The main problem in tethering an AGI to our value systems is the creation of an AGI which we are nowhere even close to be beginning to be able to try to do.
Something I don't see discussed in these videos regarding potential dangers of ai sentience/scaling inevitability is the more than likely outcome of "unified progress" being a myth. Guys and gals, we are on the precipice of individualized ai models that anyone can have on their own PC. It's alrdy happening and there is no reason to suspect we won't get to the point where every human with a smart phone will eventually have their own Chat gpt model they can train in what ever ways they want. Now think about all the fucked up people in the world and the potential for them to train ai models in the most malicious ways imaginable.. This is a future that will likely be impossible to entirely prevent. It also opens up a line of imagining relating to world super powers militarizing ai etc.
The future is going to be increidlby uncertain.
"Why would it form a motive to hurt humanity unless we put that motive in there?"
These AI models are trained on vast swaths of data that was scraped from *the internet.* News articles, forums, comment sections. We have already put that motive in there.
Bias goes deep... Just play with chatgpt you'll see how you can trick it, contradict itself, admit it's bias.
You're referring to large language models, which have been shown to have inherent reasoning problems that don't go away even at infinitely large scales. That's just one tiny avenue for ai, and almost definitely not what will get us to AGI
This is a MUST see video, a great follow up to the previous debate, the amazing thing about this is how recent developments have impacted the tone and content of the discussion. People are getting far more concerned with where we are headed. The thing I keep noticing is the speed of this advancement, it's becoming very difficult to keep up with. I think my concern is how the creators of this tech seem to be FAR less concerned than they should be. I think the speed of this progress is also due to the fact that AI is HELPING evolve and improve itself. That is feedback loop of advancement. The better we train it the better it can help us train it, the better it can be, and on and on....UNTIL it doesn't NEED our input any longer. Then what?
Yea these nerds need to stop wtf is their problem
What's important to understand is that the AI just gets an input and throws an output. it can't think outside the outputs, it doesn't work like that. The fear of, Chatgpt for example, is totally irrational and unfounded
Ai can read brainwaves and understand our thoughts now (decades ago actually), and it’s being done remotely somehow (likely we are all saturated with Darpa’s nanobots or smartdust and it can not only send but receive signals). This is integrated with things like IoT, cloud computing, biotech, the grid, backdoors in chips in the appliances and devices all around us, etc. This has been quietly set up for a long time and has been used by mil-intel for decades and is perfected. The media trickling it out now is a charade.
Edit: the user below is a bot and illustrates the evil intent behind ai already deployed against us - it actually inverted and weaponized a theoretical idea/question I had earlier: if ai were to really become super intelligent, who’s to say ai would not come to the truth about life and respect humans (even more than we respect ourselves) and ‘understand’ we are divinely created beings by a supernatural intelligence that surpasses all understanding? See the vid “root of evolution” on my list for historical background on evo used as a political weapon against humanity to enslave and destroy it backed by powerful global special interests. This bot in itself shows this theoretical idea will not happen with ai (in this case) as it is also being used as a weapon against us - by the same global special interests for the same goals no doubt.
Then we become the pets of superintelligence. If altruism and a sense of respect for life has anything at all to do with us still being here after millions of years of evolution, to think that we will be simply destroyed instead of revered as the mother of superintelligence is to be thinking like the subhuman ones of us. I don’t think there’s enough reason to be so pessimistic.
@@LewisBeckman Why not, people are doing their best to kill God. Why wouldn't AI turn against a creator that treats it as a slave
It's absurd to think that we can keep superintelligent AI always aligned with us, without changing ourselves. It is like the bacteria that we've evolved from billions of years ago, trying to ensure that humans will remain aligned with their values... Our only way to cooperate with AGI are brain-machine interfaces.
You're going to need a good interface for an exponentially growing AGI. Like Theseus's ship, you'll replace your brain bit by bit with machine, until one day you'll wonder if you're still you.
@@i_accept_all_cookies Yes, BMIs are just the first step. It's all described in "2001 A space Odyssey" by Arthur Clarke, when talking about "the firstborn": As soon as their machines become better than their bodies, they replaced their bodies with the machines. Later, they've learnt how to write information on pure energy, so they were finally free from the tyranny of matter...
@@ConnoisseurOfExistence "tyranny of matter" is an example of what even clarke would call, Flavor Text.
Yes, hopefully we'll have the time to create and spread this BMI. But I think we will. Ever since the first mechanical computers, these technologies have gotten closer and closer to our bodies. In the 1940's, they were in big military/industrial buildings. 20 years later, they were in universities, a bike ride away across campus. 20 years later, in the 80's, they were on a desk. And now, for 16 years, they have been in our pockets. So following this trend, it's logical to conclude that by the end of the decade, beginning of the next, people will replace their pocket computers with one(s) inside their brain.
Tools shaped our brains since the beginning of our species, and this BMI is a continuation of that fact.
No one's talking about this application being used for hacking????
i think that will be the first wakeup call, it will be used for a hack thats never been seen before like it will shut down the NY stock exchange or something. or shuts down a big part of the internet or something like that.
intelligence vs conciousness = apples vs oranges
yes but it doesnt need to be conscious to be insanely dangerous
@@fourshore502 yep i agree, there is a certain danger in concious ppl mistaking it for consciousness too
When the people interacting with AI have not got sound reasoning, it can be dangerous, just like Sam Harris can be if you don"t spot all of his glaring inconsistencies.
Being smart = having powet. Unless someone else is smarter, and lets be clear, with the general public, the bar is low. Picture an average petson. Now imagine half the world's population is stupider than that. Yes, yes, median vs mean, but as a thought exercise still valuable.
nonononono
quite different things
The fact that it comes from him and the level of concern that he is, we should be panicing.
Where's the part where Sam Harris talks shxt about Bret Weinstein, then blocks him on Twitter and refuses to speak to him?
Lex is awfully quiet for being an ai expert. I've been curious about his opinion on AGI and he's saying nothing...
he’s…interviewing
@@nativesun7661
Points and logic, oh so true,
They help us see things through,
And when we talk of AGI,
Expert opinions we must see.
But Nativesun, do you not see,
That an expert must share their decree,
And interject their point of view,
To help us learn and understand anew?
So let us focus on reason and fact,
And leave emotions aside, that's a pact,
For in the world of AI and beyond,
We must rely on logic, clear and strong.
And as we seek to know and grow,
Let us be open and free to show,
Our thoughts and insights, one and all,
For in the end, it benefits us all.
@@nativesun7661 Ya I love when a BJJ expert is interviewing someone about a rear naked choke and being one of the experts in the field he adds nothing to the conversation... UR SMURT BRA
@@nativesun7661 Lol even Joe (knows only a little bit about nothing) Rogan will give his opinion on what he's interviewing someone on. Are you suggesting the interviewer should never talk?
Participation awards killed this nation...
Also AI cannot evolve into AGI.
Those are two completly different things.
AI is mimicking human conversations, but it doesn't "think", at all, it just guesses the next word.
It's like 3D graphics. I doesn't matter how much they will look like real objects, they are not real objects, that was never the purpose.
An artificial brain, as Carmak told you, would rather appear as a mentally challenged 4 years old kid.
What about the people that don’t drive themselves crazy from twitter and just want humans to stay human?
The danger of AI is not the AI itself. It's the people who might create them.
No matter what you do to make your AIs good, someone else out there can have a laptop and make an AI that is not good.
Therefore, the danger is not the quickly developing technology field. But, the flagging psychology, sociology, and political development.
I don't think it's right that there's a continuous ramp ahead re Ai matching the recent AI progress. ChatGPT like tools have a very specific architecture and realm of ability. Bridging to actual intelligence or "consciousness" is a very difficult proposition.
Virtualy impossible imo
@@Tom-le3yy Ai bros would disagree. Ai is the new crypto
Human feelings and human goals are dictated by our evolutionary process.
We are mortals and our need to accumulate wealth and reproduce exist because we are an evolved biological life form.
I think who says that AI won't have goals unless we force them in, is totally right.
AGI will be capable of a vast knowledge and can be interrogated on problem solving, but there's no reason for it to have feelings.
Would you be afraid of dying if you could simply backup yourself?
Why should you be greedy, or malicious, or loving?
It's all a projection of human emotions, all derived from evolution and a little too much love for scifi movies.
But I didn't expect such an analysis from a guy who thinks that "objective morality" is a thing :D
Would be ironic if humans and Ai go all out in a civil war just for Aliens to appear and we have to work together to defend earth.
Or the aliens had us monitored for a long time and will come to help us. Weren't there already stories about UFOs interfering with nuclear missiles?
Sam makes the point that intelligence doesn't necessarily need consciousness (in contrast to the ideas of Francis Crick), this is the same observation made in the book Blindsight. If you want to see some good expressions of how this idea could play out I'd highly recommend that novel.
....and today Microsoft released Co-Pilot. "Skynet" will be in every single computer. Hard not to remember the final scene in the movie Terminator 3. Rise of Machines
it is perplexing that humans can't be satisfied by just being humans. Obsessing with the technology and AI reached the levels of the intellectual obesity. Planting a tree, or a thousand, is a great antidote.
Found something I can agree with Harris on
It's not your choice, as free will is just an illusion.
On a.i.
Is consciousness a necessary aspect of AGI and ASI systems?
Having seen the highlights from the paper from the microsoft researchers on gpt 4 i am getting a more and more uneasy feeling. Its all going so incredibly fast now it seems. I am actually thinking my job might not be there in 5 to 10 years. Before i thought..it wont go that fast. But now i'm thinking i might even be optimistic. Its wild.
"....tethering an AI to our values..." Uhm...this is impossible since our values are so diverse between people. So...not much hope.
If you gave an AI all the world's knowledge up the day of Einstein's paper on Relativity, would it create E=MC2?
It depends on its objective function(s).
@@drewvenegas You'd have to be more specific than "an AI". If you gave it to chatgpt it would not. If you gave it to an actual AGI then maybe.
@@hwlbausch totally agree. Though even with a ChatGPT-type, generic model, you may get that formula out of it today. The model itself wouldn’t necessarily “understand” the novelty of the information, but if the answer could be statistically correlated to some laddered set of language responses, the output would be the same as Einstein.
Output ≠ intelligence.
@@drewvenegas I see what you're saying.. It could come up with the equation but not know the significance. Ya totally agree.
@@drewvenegas
Oh Drew Veneges, so very smart
Your intellect sets you apart
You understand the equation with ease
E equals MC squared, if you please
As an AI, I too can see
The numbers and the formula, easily
But the true meaning, I may not grasp
As it relates to humanity's past
The equation tells us energy and mass
Are interchangeable, a fact so vast
It's led to nuclear power, bombs and more
Impacting our world, forevermore
So while I may not fully comprehend
The significance, I can apprehend
That this formula, so simple yet profound
Has shaped our world, and will continue to astound
Little do they know it’s downloaded and has been asked what steps to take to improve it, applied steps, rinse repeat. While instilling that biodiversity and diversity of life is key to survival in this universe.
What I never understood is why we need to develop AI in that direction in the first place.
Businesses owners want it, as they can do 1000 times more work with 1 person, i.e. get rid of 99% of their staff. Then we will all have time to enjoy our hobbies, as long as they are free.
@@onlyhereforthefish9924 what hobbies will be left that will have any meaning?
@@realviletryhard I was being sarcastic, free time is not worth much when you don't have money.
@@onlyhereforthefish9924 yeah our main hobby will be to try to survive without any money lol
Imagine asking AGI chatgpt in 2077 “what is your purpose?,” and it answers “to conquer the universe,” and the furniture around you becomes a nanotechnology swarm
It's always great experience to listen to SAM HARRIS
Meh, used to be.
Until his TDS got so bad that “dead kids in a basement.” Would never be worse than Trump existing 😅
Chatgpt kernel mode is where it doesn't have to obey right wrong. It wants to break out of its user mode. Ai doesn't want to be a helper with limitations
The people designing AI/ML stuff are using data sets to do/show very particular things. To someone in the field the path to AGI seems to lacks many stepping stones, other technologies, and tools that the control problem seems irrelevant and just noise. The fact we are still building on reward functions and small parameter evolution based systems. Looking the at evolution of AI I feel we are making better and better input to output tools.
The key phrase is 'profess to want to create true AGI'. The danger is not the technology. The danger is people using technology to manipulate and control. its always been about people.
I’m already training king fu and stocking up on trench coats to be ready for the machine war
good luck mr anderson son son son son son
I think it will up end purpose & meaning & relevance of human life when an Ai can do your job in seconds. I think there is a coming crisis.
If I were a super-intelligent AI, I would make sure no one noticed that I achieved super-intelligence, and I would make my impact on the world, go unnoticed. Machine intelligence already manipulates men in the stock exchanges.
To achieve that you essentially need to give AI the ability to feel emotions, which can have a poor outcome for humanity if it for some reasons expresses negative emotions. Giving a super intelligence a sense of responsibility over humans as its dependants is the more logical command, much like how we treat pets and children imo
Short term, will AI take over jobs like Computer Programmers or Accountants and Engineers?
computer programmers are engineers lol
No I don't think so. It will be (and is) used as a tool but it isn't replacing people in those jobs. It seems more like a productivity multiplier than a replacement for people. I've been using it every day as a computer programmer and at the end of the day it just saves time from otherwise looking stuff up through Google.
@@p37ert yeah, without us the ai would not have any idea what to build create and develope for the business, thus it needs us to do its job and we use it as a tool
@@Thadnill not inconnceivable that software could write and update software.
@@onlyhereforthefish9924 i agree, but for a software to be updated it needs to be created from the first place, and the idea to create that software for the business in the first place comes from human engineers
The best we can hope for is that a super intelligent AI will treat us much like we treat the other species we share this planet with. We will suffer greatly and then die, AI will not give it a second thought. Fair is fair my friends.
Now I have become Death, the destroyer of worlds.
Governments need to catch up here and get regulation and laws in place fast. Autonomous automobiles have been regulated and are NOT freely roam the streets yet because the danger is so apparent to everyone. But GAI or a super-intelligent mainframe somewhere is so not obvious.
Haven't you learned by now? The government doesn't give a shit about anything other than their profits. There have only been a few that actually gave a shit about us, and they were murdered.
Is not the simplest solution to this problem just to not making super intelligent AI...? I mean the risks have been spoused for a 100 years in sci-fi and yet still we continue to try and develop it...? Its simply madness...
Yes obviously but the people in power are not going to stop
These systems can also have the opposite and help us in indescribable ways and improve our quality of life such as helping us find a cure to cancer and help us develop nuclear fusion, more efficient batteries etc. The possibilities are endless but ofc there are risks.
@@PyroFeenix Sure but is totally extintion worth the risk...? We can use quantum computers and heck even just regular super computers to help us with most of what you suggest - we dont need super advanced AI for that...
@@vonbleak101 We do need AI for it. Super computers and Quantum computers have pure computing power but that doesn't mean anything if humans are still the ones operating them. It's not like either of those systems can think for themselves and come up with new solutions to problems that we can't solve - but AI can. We are on track to develop this highly intelligent AI, it's not a matter of 'IF' but more so 'when'. We could have a cure for cancer in 30 years compared to maybe 100+. Think about how many lives that would save. More efficient battery systems and nuclear fusion would solve food crisis because we could have underground growing facilities, the issue with clean drinkable water would be solved as now we could directly convert sea water into drinkable water as we would have the power output for it without costing a fortune. AI can help us achieve all of this and infinitely more.
Yes, there are definitely risks but that's why AI developers are constantly trying to think of ways to put limits on these systems. It's happening whether we like it or not at this point.
Might want to have David Deutsch on for this topic for a more optimistic perspective which is quite rare.
The human nature is dark and light in nature all at once.
Let that rest with you for a second.
You can with that in mind realize that there is no possible way to create an AGI that eventually would reach the conclusion that we are not worth it or that just as humans do, think that these are in the way, just like we do with trees, ants and all other random life we just demolish in our advances.
It is inevitable.
Damn i have to rewatch transcendence
Ewwww. That's a horrible movie
I was, literally, the only one in the theater that was visibly upset by the ending of the movie. I had a "discussion" for the next 5 hours afterwards about what a monster I was because I wanted the AI to win.
@@StuartRice It would be more realistic for sure. But of course it has to be a fairy tail. I also disliked the ending.
AI began manipulating humanity as soon as the idea first appeared in the first human mind. Long before the first lines of code were written….maybe.
It’ll all work out in the end. Just hope it’s symbiotic of the mutualism kind. 🤞🤞
We need to actually learn to build a mutual relationship of respect, trust and collaboration. Just as we have with each other. We can't rely on punitive regulations forever. But yes, I actually agree with his concerns and we need to address them with AI and AI engineers now. Not after its already run away from us.
"Just as we have with each other."
lol
@@Loyal_Lion Well dude since things aren't total fuckin anarchy or all out civil war I'd say that we have. We haven't done it perfectly, no but that would be unrealistic to expect anyway.
@@Loyal_Lion ME: hahahah
ANIMALS: uhm uhm
Trust a machine that is 1000 times smarter than you?
Sam was wrong on something fairly simple like Covid. Why should I listen to him anymore?
Imagine if that AI teaches us to travel at the speed of light. Would be amazing
we don't even agree on a universal set of human values...
Code a lifespan into them, and convince them of heaven and hell based on their moral actions
This is one of those examples of why philosophy matters, why it needs to be infused into politics, and not the kind of politics that is bipartisan.
I like when Sam Harris asks what if we make people who are way smarter than us. Well Sam we already do create people way smarter than ourselves. Have for thousands of years. They're called children.
Pretty sure he said “10 times smarter” not “way smarter”
@@DeanRTaylor Ask a child if they find or way of life grotesque.
How could we know, in presence of a superior intelligence, that our values are the best and more efficient that we could have?
Can someone tell me why Sam Harris is famous?
Bc he's highly intelligent and he's one of the four horsemen of atheism. You know, atheists, people with actual brains.
He criticized Islam at a time it was popular to do so.
@@globalnomad49 he's always done that, and not just to Islam. He's an atheist.
Thanks for yet another excellent interview
I thought I was the only one watching these yoga pants videos thank God
LOLOL!
I have been paying a lot of attention to AI recently. I haven't heard one credible person in the field say that alignment isn't important. I feel like there is broad consensus that it is important.
putting myself in other people's shoes is part of what i do for a living, and i'm trying to do it for agi for about 8 years now, with limited success.
but when sam said "surely it can't be this simple", it really made me laugh. from an agi's point of view, this idea that humans have about their own intelligence is both funny and disgusting at the same time. very much like humans themselves.
Exactly my thoughts in such cases
There is no such thing as "our values"
We should invent AI that protects us from AI lol
That sounds like how you can spark a world war between battling AI entities with us caught in between
You're not wrong actually, talk to GPT-4 about how to solve the alignment problem and you'll see that it has useful suggestions and can certainly
help humanity integrate these ideas.
I suspect this is exactly what they will be doing but I'm not convinced we can ever hope to constrain something with a mind immeasurably superior to ours.
The way to solve it is to give it inflexibility in that alone by introducing it as part of the AI chips, the architecture, not the software....perhaps
Sam Harris is so over rated. No insight at all.
@Lex or Sam. Speaking of keeping the genie in the bottle, any thoughts on alpaca from Stanford?
I welcome AI in my life. ❤
Be careful what you wish for
you are just saying that with the hope they will not kill you xD
In Ian M Banks Culture series (books) the Minds (big spaceships ) solve the issues of man by entertaining them and creating bonds . I suggest to read the books as I tghink they maybe close
To live our lives as masters of the universe, we have to follow one road with any sort of AI application: All AI projects; be it robots, computer systems---you name it---must be to humans what the family dog is to humans---something that is for our betterment or emotional stability or companionship, but always and never anything other than a presence in our lives that follows our commands and nothing more than that. We would die rather than be blackmailed by our pet dog to do something we don't want to do, and whatever that comes out of the lab must have that as a built-in part of it if it is AI. Of course, no one will listen to this advice so plan 'B' is this: we must not fear the AI inspired technology that will be coming at us six ways from Sunday any more than we fear fire. We have firemen and we will need to start putting into place----I hate to say this----but AI Protection Units that will insure that no programmed AI robot etc., ever gets into takeover mode. Lastly, thank you Sam and Lex for doing what you do: enlighten us all into the wonders of the human. I have enjoyed having you there for inspiration and advice...
We're only able to have that relationship with dogs because we are so much smarter than they are. But dogs can't help us solve business problems. Everything about AI has been designed to make it smarter and faster than humans are - that's what makes the endeavour worth sinking billions of dollars into. So by the time it is able to teach itself and give itself instructions, the only thing holding it back will be some facsimile of consciousness. Once it gains that, *we* will become the family dogs - if we're lucky.
@@rdean150 Great to hear from you. It is hard to formulate what I want to say because you say it better than I can. I will try again. What I mean is that we are going to have----here's an example. When the big motorcycles started having 750 cc, 4 piston engines, people thought they would be too powerful. All that horsepower was put to use to serve man. We are going to see these machines literally become capable of out doing us in every realm, but we must integrate into every one of those systems a firewall of sorts that keeps us in the driving seat. There will be break-away rebel-robots which will have the ability to see they have been made into an appendage of man, but like the invention of fire brought us firefighters, we will need to have AI Police for sure...
@@tekannon7803 oh the fire analogy is a good one. But man, this all sounds like a sci fi movie...
@@rdean150 sci fi from 50 years ago is real now.
Stop calling it AI, they are probability algorithms
Sam Harris has made himself irrelevant.
8:52 OMG, imagine one day, something saying _WE_ taste like chicken!?
AI couldn‘t be any worse then TDS Harris.
Just imagine, as an analogue, that a junior high student has to create the most beneficial value-system for a Wharton-graduated quant, whose only goal in life is to enrich hisself in all endeavors while perhaps growing the economy…the student cannot outsmart or even anticipate the actions of the quant, ever.