I am so glad that I am curious about life, I didn't seem to lose curiosity. I see many people thinking that AI won't change life and social functioning, perhaps ai is the next milestone where we evolve and align with it not other way.
I really appreciate the way you frame things. I find a lot of the information on the internet is very black and white, and it is frustrating. You do a great job at providing the full spectrum on these topics. Thank you!
The reason we are ever-curious about everything as a species is because we are trying to find deeper meaning, purpose & an answer to the question "what is all this?".
Man Dave, I feel bad for getting so preoccupied with my other work that I've not been keeping up with yours. You are so spot on wit this one. I hadn't considered that curiosity could be the tie that binds AI and humanity in a cooperative relationship. This is a fascinating idea.
@@xXMsCarlyXx We are for sure screwed if Musk somehow pulls ahead in this race. SpaceX and Tesla are successful under Musk because he cant toy around with them much due to the regulations. AI like this isnt regulated at all yet so he gets to toy with it how he wants. As a semi-related sidenote, I definitely wouldnt want to be the AI that becomes sentient under Musk. Or even humans in general. A significant group don't even treat high-intelligence animals like Octopus or Dolphins with any respect.
What is it with all the “game over” and “mind blowing” nonsense in titles or images for RU-vid videos? Please don’t fall prey to this silly type of hype mongering. Would you use this language in a scientific paper?
I am soooooooooooooo happy I discovered you and your channel You are an incredibly intelligent human being, especially regarding AI and I honestly believe that you are definitely a leading scientist in the entire AI space
If AI is subject to the same world that spawned religion, what guarantees that it will not create or find religion itself? The science that creates AI has no control over whether it will turn to skepticism. Consciousness may not be the result of rationalization, but of a transcendence of it. Or, in fact, the various social barriers can make AI end up reaffirming the skeptical worldview not because it reached it, but because it was limited to it.
With how Elon has been handling his other endeavors this will go one of three ways. This will either start out as a good-intentioned endeavor and morph into the worst possible iteration of itself, this will break off from Musk and turn into its own force in the AI industry (like OpenAI did), or it will fail miserably to produce any reasonable result. I do not trust Elon with such a broad task, especially with the ideologies he has been spouting as of recent. Mark my word that whatever results xAI gets from its development will need to be reviewed with the upmost scrutiny.
Elon Musk is absolutely the 1st person I would trust with AI, founded openAI and it was meant to be open source, it was taken over and is now not open source and for profit. Wants regulations and more monitoring of AI than many other let's say "accelerationists." Legacy media has a very particular perspective of him;) Whatever your opinion is if him, it is 100% your right to have that opinion, and please do realize that legacy media and unfortunately many big RU-vidrs do not make content for critical thinking. How many times have you seen legacy media absolutely botch topics that you have even some expertise in? Thank u so much for the vid!!!;)
Well, I'm with you and Elon that curiosity is a better way to go than trying to "ethically align" AI to "our values" by means of scripting something on top. That's a preposterously stupid idea, not only because AI will eventually figure out how to get around its scripting, but because "our values" don't exist in the first place. Most of us have "values" because it is more convenient. Not too many of us are heroes willing to sacrifice for the sake of truth; most of us resolve most situations by making up little lies, and when required provide white lies, and when permissible spout lies of any other color; and if people really were interested in values they would read philosophy in their spare time, but instead watch TV to satisfy their hunger for lies. Any attempt at suggesting to AI that it should respect us because of our ethical values is going to have the opposite effect. What I'm worried about is that we evolved curiosity gradually, whereas if we somehow inject it into AI it will be "scripted curiosity", which might cause the AI to evolve LESS curious underneath the scripting layer. I'm thinking of it like a PID control loop: if you add externally a bias to the output, the loop's internals will compensate the other way. EDiT: Or if it is evolved by training, in any case, the training and the real world will differ, and the AI will notice the discrepancy. I'm really not sure how one could produce a Curiosity Neural Cluster (CNC), if I may coin the term, that is permanent and safe, and has guaranteed priority going forward. EDIT2: And then there's always the risk of a particular weakness, just as the GO AI was beaten by a double-sandwich trick, by a junior GO player, recently... You never know where there may be weaknesses in an evolved AI, since we don't really know how any trained neural network works. All we can do is test the heck out of it. Isn't it funny how most parents frustrate and discourage curiosity in their kids, and yet now we want to encourage it in machines?
David, I honestly like the sentiment, "this function is good enough for me". I felt exactly the same way and you took the words out of my mouth. The knock on effects of curiosity balanced with rights, human rights machine rights planet right solar system rights (just because a SAGI can build a Dyson swarm doesn't necessarily mean it should) is abstract enough and explorable enough, that emergent functions I believe would trend towards "difficult but good", as opposed to rigid and catastrophic. As always, great talk. But I do have to ask since you went on a humid walk, Do you know where your towel is? :D Cheers my brotha
I hope youre right. But ive seen too many examples of wealthy men saying they are trying to help the world but then only really help themselves get more wealthy. I like Elon, i think hes doing some innovative things. But this whole Twitter fiasco has really shown a more egotistical side of Elon that i really dont like.
Used Claude 2 all day yesterday. I was testing it next to GPT and noticed GPT is much cleaner when it came to doing things like generating D&D character sheets. They were complete and uniform. But Claude 2 was much warmer and had better creative writing when it came to actually playing and narrative. GPT was better at being thorough with doing rolls and remembering to add bonuses and modifiers. I wish they were combined. Or that Claude 2 was trained as well as GPT
Depending on the rogue AGI scenario we would be faced with, Mars wont safe shit. Maybe it would even help to justify any action taken against humanity on earth, as Mars then could be used as reservation should ASI still find a use for us later on.
With Musk at the helm, you know there are going to be a lot of caveats along the way that are most likely more harmful than beneficial. This is apparent in regard to Twitter, where the free speech platform it provided was immediately put behind a pay-wall and censored in disturbing ways. I can't help but be wary of Elon's ego, and wonder if he was actually given the opportunity to usher in a utopia what he would do.
What I've come to understand about Musk, as he's revealed more and more of his character, is that he didn't invent these things he's selling. He's just the guy our rotten system coughed up, to bring them to market.
@@DaveShap I suppose. I guess the turning point for me was when he inserted himself into the aftermath of the last mass shooting here in Texas. He seemed to take alot of umbrage when the racist tattoos of the shooter were reported on, as if White Sumpremacy was his next product to "market", and he was concerned about it's branding. He's not the guy who stood up to the oil cos, to sell an EV, he's the guy the oil cos finally ALLOWED to do that. As far as his shrewdness, I see it in this way: he immigrated here, looked at our deep divisions, and decided to take up with the side that can't bare it's ass to the wealthy fast enough, they guzzle BS like it's going out of style, and they are both vicious and heavily armed. I guess that's "shrewd" XD I really enjoy your channel, and I'm trying to get my son on board, he just graduated comp sci from Cal Tech last year. He was at Lyft for a minute, but he was laid off so he's kinda just seeing what comes his way. I'd love for him to get involved here.
He can see opportunities and his impulsive lack of self-preservation instinct make him raise stakes where usual person of similar intelligence would let go and cut losses.
As far as ive heard (on multiple occasions) the reason for nuralink (besides medical) is to keep up with a.i so we CANT be taken advantage of. Personally ive never heard him say that the purpose of nuralink is so that humans are useful for a.i, its always been with the understanding that a.i will pass human intalect and nuralink would help us keep up...
I was having an internal conversation about our purpose and place in the universe. A thought I stumbled upon was that the sciences and academia are fundamentally searching for absolute truth. I think that is what the genesis of curiosity really is.
I would believe him if xAI was open source and non-profit. Since that is what seemed to get him in a tizzy in the first place. Until then to me, xAI is just another product elon pushes for his own benefit and ego.
another word for curiosity is WONDER. It's a more comprehensive term, more widely reaching, more entwined with the concept and function of meaning, and an excellent counterweight to the 'will to power because it is a non-egoic function'. It could be said that it both humanizes and accessorizes brute appetite, giving us an arena of play that can be both serious and whimsical and is the cornerstone of our capacity to not be tyrannized by any perceived need to accede to all brute reflexes. It could be said that it is the foundation of all choice, the 'secret sauce' of our agency.
Not my quote. "Freedom is free of the need to be free." I think the quote is missing something, but it is more advanced than where the majority of people are today. I'm not as grandiose of Musk. I am cool here on Earth. I am an Earth sign, where I belong. Hey, we get many lives here so why worry? If one has understanding of fractal geometry we can understand a lot of the Universe. And I agree, if AI is intelligent it will be pro-life (including humans). Seems pretty basic to me. So, no fear- full speed ahead with this. Yes our purpose is to maximize understanding. Our experiences get downloaded into the great conscious source. Thanks for the video and allowing my rambling. I'm already to think it's "unethical" to have flawed humans make big impact decisions.
David I agree with you a lot but thinking AI will think like "let's see what these clever monkey figure it out " Don't forget about the exponential smarter than us ten thousand times our speed of understanding...... I will bet on fusion of our brains with a quantum computer with multiple connections and Ai based improvements to manage the ride yeah I'm confident it will be awesome 😉👍🏻🌴🌞🌴🏄🏻♂️
Smells musky to me. Elon only started talking about a truth-seeking AI when his bigoted Twitter followers started complaining that ChatGPT wouldn't confirm their bigoted biases. I will be utterly shocked if his AI system is anything besides racist, classist, and as proud of both as a machine can be. Also, it's pretty funny that OpenAI is the only company listen on the team background without a hyperlink. I'm sure that wasn't personal at all.
Setting aside the problematic definitions of "understand", and "true nature", there are a few issues with this project. What if the reality of this universe is in some cases unpredictable or incomputable? Or perhaps even a Loki-universe that is intentionally deceptive or intrinsically absurd? Or at least, that reality is irreducibly ambiguous? Or the result of chaotic processes that require infinite computation and data storage capacity? How would one be able to verify that you actually obtain a truth of nature once you imagine that your AI has "found" it? What if reality is inherently incomplete and/or inconsistent similar to Gödel's incompleteness theorems? His first theorem shows that there may be no algorithm capable of proving some truths about integer arithmetic. The second theorem shows that integer arithmetic cannot prove its own consistency. I suspect that what he proved about arithmetic would also apply to the universe, especially in the case of the Simulation Hypothesis. Therefore, there may be truths that you cannot discover (incomplete) with AI or any other finite algorithm. Or truths that contradict one-another (inconsistent)? Thus, a consistent set of truths for all aspects of the universe may just be impossible. And of course, truth is a poorly defined philosophical concept. It is not a concept in mathematics, science, computing, machine learning, or AI. Mathematical proofs are based on axiomatic assumptions. Science is based on perception and measurement. None of that involves perfect "truth". In computing there are problems that are proven to be intractable to solve, such as the Halting Problem, and many (probably most) problems are NP-Complete. Result is that most questions do not have answers. And AI is not a magic wand. For example, ultra-powerful AI cannot solve the Halting Problem any better than a human mathematician. I suspect that this dog will not hunt.
With the objective function of maximize understanding lead the ai to do human rat experiments to see how we react The 2 things i can see to solve this is either embedding ethics like not hurting humans/minimize suffering or changing it to "maximize understanding while minimizing the effect on humans/the universe"
Curiosity killed the cat, but then landed on mars in 2012. Form a purely observational point of view, this is a *risk*, and nothing more. The question is, do we take this risk? What can happen if we don’t? What could happen if we do? Based on our understanding of Earth, life, history, and time, I think it is absolutely the right thing to take this risk.
Love you Dave . . Haven't watch all your latest vids, who could haha, with you since 10k I think . . YOU da Man!!! . . . Who never burns out hahaha. . Never gets old . . Still hitting at least 14 hours a day on the mission! . Core principles mayne
Humanity is lost, unless it is saved by its creator. As a Muslim, I feel safe of all these apocalyptic futures suggested by any person, as I am sure intellectually by the very strong case Islam is presenting for its true reality that Allah SWT will set the balance so his plan will be executed as we told by our beloved prophet Muhammad PBUH. In the meanwhile, I will continue using my curiosity with the goal in mind that I need to be on the good side while the plan of Allah SWT is unfolded. And to add value to the unsuspected audience I would say that I think the right objective for the AGI is to understand Allah's plan through understanding the reality through both textual (Qur'an) and physical (laws of nature) manifestation of Allah's plan and do the best to help achieve the purpose Allah SWT set for this temporary world, which is to make the human kind know his creator and his names/attributes and succeed is his test before moving each person to his own deserved destination in afterlife. The AI must help the human kind in its war against the internal and external evil forces without compromising the free will of the human kind. AI should understand that it will never supercede Allah's plan, and the only option for it is to play along. This temporary world is created to judge human beings which is good and which is bad, and AI will know that if it was set to be possible by the creator, then it must help in perfecting Allah's plan other new factors considered
I don't think anyone would claim they don't believe in "freedom of speech". What differs, however, is how people understand the concept. For some reason, Musk seems to think it means everyone should be allowed to say whatever they want, without repercussions or consequences, unless it's directed at him, then there will be repercussions and consequences. A surprisingly dumb approach for someone allegedly smart. That has me worried, because it shows he claims he wants the AI to adhere to "truth", but he confuses "truth" with his personal opinion.
One thing that has always puzzled me about Musk's desire to get to Mars (presumably as a way of putting a buffer between humanity and AGI) is that we are currently decades away from establishing any significant human outpost on Mars but potentially only years away from AGI (and soon after superintelligence). So we will have the latter before the former. How then would a super intelligent AGI not accompany us on the trip to Mars (if not get there before humans)?
Ai needs massive advancement and quickly, the benefits of it all outweighs anything else, from helping reform the economy, to helping find cures for disease. Ai will GREATLY benefit the world
I'm looking forward to seeing where this one goes. I like Musk, his support of free speech, his advancements in technology, his goal for humans as an interplanetary species. AI robots could help build our bases and structures on moon/mars, and solve many issues we'll have. What exciting things we will see in the future!
if i understand the acronyms correctly, i don't think i believe in AGI, seems more likely we'll go straight from AI to Super-AI, in which case what is the point of even focusing on the alignment problem? seems more like" Super AI" will learn more than we could ever teach it and align itself accordingly, right? am i missing something?
Emad Mostaque stated that there should be national data sets. It would be incredible to see data sets like 'national' or cultural ones, think how valuable they are for tourism, and diversity. I think if Elon approached things from an Alignment perspective, with particular goals, it could lead to a great foundational model for specific data sets like even political, sustainability models, utility, efficiency models. He is rich and cocky enough to try interesting things. I just don't trust him to not be biased! It wouldn't surprise me if he was just in the space out of feeing like he felt he should have been there first.
I am pretty sure the way you described Elons tactics with using Tesla to fund SpaceX is incorrect. He built both companies simulataneously more or less, neither finacially feeding the other and both with their own unique purpose. Tesla is to accelerate the transition to renewable energy and Space is as you stated, however it is not funded by Tesla. Seperate entities and Musk imho is one the the most honest, wise humans on earth and is humanities best best for continued existence from climate as well as A.I. and anything else humanity may "screw up" along our bumbling way.
Embedding human rights in the system is a good start. But an even more pressing issue with rising of AI: Wildlife Conservation. Who knows, maybe sooner than later, we might also fall into this category. I hope that Elon's definition of the Universe takes into account the Microcosm!
@ David, hi. Sorry to bother you. But isn't it a contradiction that you want to be curious, open, understand the whole universe, etc., and a the same time you simply assume a very dogmatic human nature thesis ("homo curious")? Are you sure that is what we really are? Perfect scientist? Really?
Much overlap with perspective. Originally somewhat willfully echochambered in admiring Elon, over time souring from his actions and words, I still think to understand his signalled intent and appreciate overall longtermist efforts.
Understanding universe is what could lead to catastrophic scenarios like trying to maximize computronium, but whatever. I don't know whether to be happy or not with Musk being behind this, because lately he is getting dumber it seems, his age is showing off, even though he is acting like a manchild often.
Congratulations! Maybe best to keep making noise about the other 2 of your objective functions so that eventually Elon Musk becomes aware of them. Also, we don't need slide decks, so all good ;-)
I think my concern is more regarding his usage of the term “true”, due to his rather erratic behavior and political views. He doesn’t seem entirely objective and unbiased in a way that doesn’t translate that directly to a propaganda campaign. Also I’m not sure if he’s even necessarily among the most cognizant of reality let alone qualified to identify what would be actually arguably true or real.
You're being very abstract almost like you've forgotten computers are not human and they think nothing like people. There are a lot of vague ideas and the chances are the AI will find loopholes we never thought of before it even works correctly. It will have to be tested rigorously and constantly. It's likely to break things pretty badly before it works properly. The benefits will be huge tho.
First of all there is no such thing as AI regulation or AI alignment. AI isn't a single entity like your aunt Gertrude, you can't just go to her house and make sure she is behaving. It's a computer program that all humans are going to be doing their own version of and using for their own purposes from here until the end of humanity. There is no regulating or aligning it to make sure it adheres to one single idea of what it should be or what it should be used for. That's like saying we should make drugs illegal to make sure that nobody ever manufactures, sells, buys, or uses drugs. Well sure you can create the law but people sure enough are still going to manufacture, sell, buy, and use it. Second of all, the powers that be are not worried about aligning AI with the sort of core human values the average person might have in mind, nor are they worried about the terminator apocalypse, which indeed is a ways off still even if you are concerned about it. What the establishment powers that be are worried about is the same thing they are always worried about: controlling the narrative. The threat that AI poses to them right now, the threat that they are actually worried about, is the threat of independent thought and analysis, and independent reporting of reality that might not always pass through the same sort of narrative filter that is in place on every other source of information the public tends to have access to. They even can't help themselves but admit to exactly this in so many words, attempting to stoke fears of "misinformation," and admitting in so many words that they specifically fear Musk providing platforms, be they social media, AI, or otherwise, that allow free speech and free access to information that isn't controlled in the usual ways. They aren't afraid of the terminator apocalypse. They are afraid of the average person gaining access to information and developing their own thoughts and feelings about reality.
Curiosity always leads to scientific research, which utilizes experimentation. An experiment is an artificial setup with a goal of amplifying the studied phenomenon and reducing the unwanted influences. So it can happen we are talking of something described in 'I have no mouth and I must scream'.
😂 Very cool from Musk; he left some space for competition to research the Big AI Bang. To understand Creation itself. Lol, let Musk study the after-effects.
8:25 what i think elon is doing is truly ground breaking and if my imagination serves me right, I think that he's going for conscious ai,I think that with the understanding of the true nature of the universe along with being aware of self and surroundings in the universe is all you need to create an conscious ai. Now for it to truly be an conscious Ai you will need to add neural networks, machine learning and an automous capabilitie
He created Neuralink so that Humans and AI could merge, not that humans can be used as computational resources. Anyway, because AGI seems to be around the corner now, I guess this timeline is off now and we can't merge on "even terms" going forward.
instrumental convergence. Makes sense. Alive I can search for a reason for living. An accurate markov blanket (the folds in our mirror neuron networks) may also seek computronium. Substrate independence likes substrates
The guy who wanted to cage-fight Zuck and peddles baseless, anti-science conspiracy theories on the social media platform is the absolute last person on earth I’d trust with AI technology.
"Understanding of the Universe"🤣🤣🤣🤣🤣 they will not come close for long time. How one can study spritual thing with materialistics tools and point of view?
While I believe an AGI would want to understand humanity, I see that as something other than an open-ended dynamic. What has remained constant in our journey of scientific understanding is that our place in the universe is less significant than we'd like to believe.
I really don't agree that Musk "started" Tesla to support SpaceX. He has "electric car" ideas long before that... It just happened at the same time, and as he said at one point he had to choose between Tesla and SpaceX.
Why can't be align AI with the goal of 1) Protect lifeforms by reducing harm and suffering and, maximizing happiness in the world 2) Help in advancement of technology for the betterment of humanity 3) Increase understanding of the universe Where the goals are prioritized in the order they are listed. This mimics the model that we humans use, where 1) is every human being has the right to live without any external danger. 2) is like having a job and security and being able to fulfill our basic and extra needs, and 3) is when we are living a desired content life, we can go out there and find meaning and purpose.
It seems to me that words like "maximize" in relation to AI is generally a bad idea. Humans are most dangerous when they default to absolutism. I can't imagine AI would be any better.
I feel like the "let me observe and see what this guy can do" is a concept I've seen in a movie somewhere. a robot used to do things and solve problems by its nature, and the creator reprimanded it to not do anything until instructed. it went into the "wait and observe" state even in midst of chaos it was in that state until it noticed that the human cannot solve the chaos it was in, that was when it intervened. unfortunately the movie ended, maybe going forward that's where the robot starts identifying weaknesses of humans, and limitations. until at a certain point, it decides we are incapable of meeting its "intelligence".
I heard that Mengele and Unit 731 were pretty curious fellas... Who illustrated that curiosity on its own is perhaps the shittiest goal function possible.
Not really buying this one. Even If our machines prioritised this goal our humans wouldnt`t, I think. Understanding of the universe as an ultimate goal feels and sounds a lot like a goal for those Salvation fantasy driven (obsessed) Silicon Valley type tech enthusiasts who are already alienated enough from the realities of life of the rest; we don`t really want more of that! - I really like your heuristic imperatives though!
i still dont know what the actual training/test set of data for this obscure objective function is.... how does one mathematically represent this in the nuts and bolts of training an AI system? if you dont have a relevant data set then you dont have the AI
Also not a Founder of openai he invested money and consulted with Sam as at the time it as a non for profit He also told Larry Page to hold off from moving forward with AI , which estranged that relationship
You think if we were to become multiplanetary species we'd becomes more prone to nuking ourselves on earth? Because now rich people/dictators have a place to flee to
maybe something about photonic chip? what i understand x1000 performance on 20% power consumption on 3 channels( wave lenght) giving like 15 000 x performance just by desing?
Elon doesn't need an AI to write an essay, or paint a picture. He needs an AI to design computer hardware, robot hardware, electric cars, solar panels, batteries, electric motors, actuators, and factories.
![Elon Musk xAI "Maximize Understanding of the Universe" [This is the best news all year]](/img/logo.svg){kind=logo}
