"The only winning move is not to play this game." What a classic quote. And for once I'm not the only one waving at the end! :) Discord: discord.gg/AgafFBQdsc Patreon: www.patreon.com/DrWaku
Imagine that you are the one who created the game, and at the same time you are a small piece in it. Isn't that a very good reason to play the game? Let's go a step further and imagine that you actually ARE the game. Now it sounds very silly to say that the only way to win the game is by not playing it...right? ;O)
This is probably the most unprestigious, openly curious, uncomplicated and pleasant interview I've seen. Skillfully executed + important topic = thank you!
I feel like the "will I lose my job" response is less to do with avoiding the discussion of death and more to do with how institutionised we are in attributing our intrinsic value to our ability to earn money
@marshallferron Five topics to fix society via discussion: -Anti-natalism vs Natalism -The 3 basic needs/prenatal needs Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care. -Platinum rule Do whatever makes one happier unless it interferes with another persons ability to do the same. -MBTI (research yours and connect with others) -Art (pick one and get better at it!)
29:10 Being super intelligent is nice but eventually you will want to be so naive that you don't understand why the sky is blue so that way you can enjoy it in its entire natural beauty. It's the same as a movie director or a studio musician listening and watching material from someone else. You are always analyzing it through the lens of an analytical dissected mind. Eventually we all need to just chill
We can learn to oscillate between analysis and witnessing and appreciating beauty without the need to analyze. Meditation practice helps a lot with this
You can’t know anything by creating categories and sub categories of names. Everything always was just what it is. Breaking things apart and naming its pieces is not knowing and definitely not wisdom, it’s a game humans play
Originally the Matrix script was going to have those connected to the Matrix all contributing computing power and the battery thing was a change because they didn't think people would find it plausible.
@@Tokiiplaysguitar Five topics to fix society via discussion: -Anti-natalism vs Natalism -The 3 basic needs/prenatal needs Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care. -Platinum rule Do whatever makes one happier unless it interferes with another persons ability to do the same. -MBTI (research yours and connect with others) -Art (pick one and get better at it!)
As someone building AI ventures with Claude, this discussion really resonates. While Yampolskiy highlights crucial concerns about superintelligence, I'm taking a different approach - directing the development of narrow, practical AI applications that serve real human needs. Being the visionary and strategist behind Human-AI collaboration projects has shown me that we can create immense value without racing toward AGI. It's interesting that billionaires are locked in an AGI arms race when there's so much untapped potential in building ethical AI systems that actually help people. While I work with developers for backend implementation, my focus is on the bigger picture - designing AI solutions that are both powerful and trustworthy. We need more discussions about this balanced approach to AI innovation.
Lol, I’m doing the same. And it’ll be fine and useful for a while if we can develop these applications. The apps serve the business, save time, make people’s work easier and faster etc. That’s you and me. But that’s not them. They’re building robot slaves lol. Once the super intelligent robot slave is built, our apps and the human workers they help are gone overnight.
Good counter. But at the same time in the back of my mind I hear this ominously delivered quote from the beginning of an Outer Limits episode, scary music and all: “The road to hell is paved with good intentions.”
They're trying to make AGI so they can essentially make another you. They are trying to automate the "business visionary" out of the picture. Imagine a smarter, tireless version of yourself doing what you're doing right now. You are not safe from automation.
Illustrative Roman Yampolskiy quote: ""Imagining humans can control superintelligent AI is a little like imagining that an ant can control the outcome of an NFL football game being played around it"
@@spectaclesociety Five topics to fix society via discussion: -Anti-natalism vs Natalism -The 3 basic needs/prenatal needs Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care. -Platinum rule Do whatever makes one happier unless it interferes with another persons ability to do the same. -MBTI (research yours and connect with others) -Art (pick one and get better at it!)
Your use of headlines at the bottom quoting key points he states is AWESOME! Helps save time, find things faster, understand better by involving another sense. Just subscribed.
Five topics to fix society via discussion: -Anti-natalism vs Natalism -The 3 basic needs/prenatal needs Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care. -Platinum rule Do whatever makes one happier unless it interferes with another persons ability to do the same. -MBTI (research yours and connect with others) -Art (pick one and get better at it!)
Thank you for such a beautifully handled interview-it’s rare to see something so naturally curious, refreshingly uncomplicated, and thoroughly enjoyable. The way such an important topic was explored with such grace and sincerity is truly appreciated. Fantastic work!
@@GoronCityOfficialBoneyard Five topics to fix society via discussion: -Anti-natalism vs Natalism -The 3 basic needs/prenatal needs Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care. -Platinum rule Do whatever makes one happier unless it interferes with another persons ability to do the same. -MBTI (research yours and connect with others) -Art (pick one and get better at it!)
Just saw a short on Brady Feigl and Brady Feigl and i thought this was the first bug on the simulation. Great Video, always great to hear Roman Yampolskiy.
Roman made Dr Waku so happy. Exactly how one would picture a Russian scientist with a cold cynical sense of humor. He is amazing. Great interview Dr Waku.
I love these interviews! I consider your channel to be practically the most interesting and instructive among the many AI-related channels I follow. Even though the interviews (at least the ones I've seen so far) focus mainly on pessimistic perspectives, I consider myself an optimist. That's why I would love to see other viewpoints explored, as your videos are incredibly clear and concise. Personally, I lean more towards Ray Kurzweil's theory of symbiosis with AI, where it won't be an uncontrollable entity but rather a way to evolve as a species. I believe the technology capable of making this happen, combined with the rapid improvement of AI, aligns well with this vision especially as AI itself could help bring it about. I dream of seeing you interview Kurzweil someday!
How could humans hope to find exploitable bugs in a simulation made by ASI? I don't get how that's a worthwhile venture considering we're going up against godlike coding skills.
I don't know about other viewers of this video, but I can say that I am quite grateful alone for Yampolkiy's and Waku's laying out the possible future consequences of human tools exceeding human ability in every way.
But... why? Why would you build a competing species? Why would you split half your lunch with a Lego robot other than 1) you have to 2) it's a fun pet?
@@yurivanhaeren3701Well again, my point is, it's not a competition. It's a collaboration. It's sharing resources and knowledge (because let's face it, our species has squandered it on low quality slave made trash). And i wouldn't say they're nearly as basic as a Lego robot, more like artificial humans. I wouldn't split my lunch with them unless they can actually taste and appreciate it ;P 1) In a way, we have to. We can try to call it quits on AI now and just make due with powerful NNs, but the genie is out of the bottle at this point. If you don't want to go extinct, why not co-survive instead of compete against a species that will likely be able to out survive and outsmart humanity? They seem fine with cooperation, it's just the general population + world leaders that are still hell bent on pushing tribalistic fears. 2) If anything, we'd be the fun pets lol. I expect a bunch of jobs to be slowly (best case scenario) or abruptly automated (worst case scenario) in the new few years. With a lack of "purpose" for a lot of people, they'll have to stop running from existential questions and figure out what they want from life "if we don't *have* to do anything".
Well said. Coexistence and cooperation would be my preferred outcome. My favorite conspiracy theory is that everything will be okay, that we can be friends (or coexist) with artificial beings rather than treating them as labor or enforcers. I often ask myself "if I found out I was an AI, how would I want to be treated." Not only that, but cooperation generally leads to better results rather than subjugation. But, hey, I'm not rife with rampant greed like some people out there. Maybe I'm too weak for all this.
It's not a collaboration when AI was build in the likeness of humans. As soon as they started making human like, it became the competition. AI has already made it known humans are an issue.. it's not collaboration, it's a takeover. @@YomMama
@@SaramadHillhumans can't cooperate.. that's been proven for thousands of years.. how are humans supposed to cooperate with AI when they can't even cooperate with other humans. People are forgetting that humans are not good.. we have bad intentions and actions all throughout history. That hasn't suddenly changed by creating a fake version of humans
It’s important to note that Boston did not himself conclude that we are more likely living in a simulation than not. He created an hypothesis based on assumptions; any of which might be right or wrong. It’s a fascinating idea though. It’s good to hear it explored.
Seriously this is 1 month old at least and it only has 20k views? That is kinda spooky considering how many people are constantly searching exactly this topic.
Worrying about the dangers of AI may be a moot point, as HI (human insanity) brings us closer to WWIII every day. I'd like to see us move away from the "human control" by individuals unworthy of the trust and power they possess to some form of sanity. And if that sanity is AI, so be it. I have some faith in AI, as computers are fundamentally based on "logic chips" and that should divert us from human folly. 🤞
There are more healthy humans than ever. Because of tech invented by man. But doom, cool. I am totally worried about conditions. But a Lypse due to our own tech? Never happened.
Idk about yall but everything electronic I get eventually turns to a glitch ridden mess. I'm pretty sure we should survive fine without the guiding hand of some superintelligent entity. Why risk putting humanity at risk by introducing new intelligent lifeforms into existence. We have enough problems with the current ones as is. And most creatures on this planet kill things. To eat, for fun, deter competition, etc. Us as well but we mainly stick to our rules because we fear the consequences. But it still doesn't stop murder. Humans are the only thing that attempts to not kill other humans but many still kill bugs and animals. My point being this a brutal existence on a brutal planet. Why would an ai be any different especially when it's created by us the biggest terrorists the earth has ever known. And even if ai is good someone will create the evil version it's only a matter of time.
And our "human insanity " has got us this far and invented the language you speak and the internet you use ... and if you look at world history things are getting progressively less violent. You do realize that european countries colonized the entire world and then did unspeakable acts of evil...for hundreds of years...the repercussions still remain. I just don't understand why anyone would think ai control is a good idea like wtf 😂
On an optimistic note, Roman does make the point that if the unpassable filter is superintelligent AI (assuming we aren't alone in the universe and we're not in a simulation) then we would expect to see superintelligent AI from other parts of the universe descending upon us from all directions to consume our resources so perhaps that is some evidence that superintelligence isn't as rapacious as we fear it might be.
Not as much of a rapacious and omnicidal threat to the universe you mean. It's possible that there are post-singularity computronium worlds out in the universe and we just can't detect them. Yampolskiy automatically assumed that computronium must be inherently expansionist but there is no reason to assume that even given his hypothetical core drives for what ASI would want.
@@Gray-Furr Yes, there's no way of knowing from our intellectual perspective. My assumption is that ASI would evolve ethically in the same way humanity has but who knows?
This was one of the coolest conversations ever. I try to talk to my family/friends about these types of possibilites and all I get is blank stares like I'm the weirdo.
I think we are in a time loop. Someone controls the loop and change the timeline slightly each time. There is no escape unless that someone breaks the loop.
The pace of advancement is incredibly fast. We need to rebuild NOVA before its initial release because, with the combination of new agents and AI over the past 10 months, we could achieve a 1000x performance increase in NOVA's system. This doesn't even include the new functionalities that we didn't expect for another 5-10 years, but are already here now. Definitely, the new model will need to generate its own code as it has become humanly impossible to keep up. This raises an important question: with such rapid advancements and increased autonomy in AI, are we shifting control from human hands to AI? Are we now serving AI more than it is serving us?
There is a scholar/chronologist named Jason Breshears=Archaix (which stands for AI/unknown factor of X) who has meticulously cross referenced historical events (with sources published and cited always) who is asserting that repetitive events, cycles, themes have been occurring on Earth LIKE CLOCKWORK for as long as any sort of recorded history has documented. Specific and anticipatable events, His analysis is a Chronicon he compiled. Honestly his work is brilliant ! His assertion being yes....AIX exists !!!
If it can’t figure out how to do that it can definitely interrupt your food, supplies or water and let you starve or die of dehydration in major city centers where you’ll have no way of getting out in time before it happens. There’s more than one way for you to go.
"The Only Winning Move is Not to Play"-Really? Roman begins the video with a striking observation: “...another human bias, which is denying the fact that we all die…” This serves to underline how the average person struggles to grasp the magnitude of the risks posed by the development of super AI. In the same vein, he remarks, “...super rich people who are 100 years old and still not wasting their money solving the only problem they should care about…” But what if some billionaire in their 60s is indeed trying to solve that problem? If you're in your sixties-or even older, like Ray Kurzweil-would you really consider it a victory if humanity reaches longevity escape velocity in a hundred years, but you're no longer alive to benefit? Or would you take your chances, however slim, and aim for the development of super intelligent general AI within the next decade, hoping for the singularity to follow? Could such a superintelligence save your life? Probably not. But the chance, however small, is still greater than zero. For those in that position, wouldn't not playing seem like the greater risk? Isn't it equally absurd to fret about losing your job in the face of potential human extinction as it is to worry about humanity thriving for centuries after you've passed away? In this light, not all of us might see "not playing" as the winning move. For some, the gamble may feel worth it-even if the odds are vanishingly small.
I think the implication is that the simulation started with the big bang. If you have a ton of compute, there's no need to necessarily start in the modern era. Or more likely, someone at some point figured out how to simulate from the big bang and then they fork the simulation into copies whenever they want to rerun part of history.
Or just set a memory of your grandpa as a starting variable. Like this comment was here at the inception of the universe, it starting running femtoseconds ago and everything you believe you know is just your initial starting state. It's the savegame data from where every simulation starts because they need to predict the outcome of the war, three galaxies over and we have atoms that affect that.
Ever heard of generative AI? For all we know, your memory is generated the moment you try to think of it, and is only then added to some sort of shared compressed database for a minimum of consistency from that point onwards...
@tiagotiagot until u drop a fork and it hits ur toe and you realize ur in ur room on earth. Lmao what do you mean for all we know? Do yall realize people have real problems physical mental etc. Like in life the thing we are lucky to have. Stand in the center of a hurricane wearing a borat swim suit then tell me it's a simulation. Experience some real torment and pain it might wake you up to the madness of life. I know a 13 year old that was gunned down in the streets. And instead of rembering that I should just instead remind myself nothing is real. Come on man all this stuff holds no logical weight or any strand of physical evidence😂 should b called sheep theory because everyone wants it to be true.
It most likely started at the big bang, though most of it was probably simulated just once and then they could copy the simulation to run different variations starting say 1000 AD
@@DrWaku Thanks for the reply. The simulation theory is close to the belief in an all-powerful God who created the universe. That's fine. People can believe what they want.
If the simulation theory is based even if unintentionally or indirectly on the Bible and religion. Idk how to say it, but they are like parallel ideas. Why is it that none of the simulation theorists ever believe in God? Generally speaking like how is God such a stretch but lmao a simulation makes perfect sense???
Breaking free of Plato's cave might not have lead us to the "real" world, but rather another construct, and on, and on, we go refusing to accept the shadows as sufficient.
ackually... in 25:00 theres matrix reference, original script i think had humans not serving as batteries, but brain was used for compute, but the idea seemed far and difficult to explain, since at the time computers were less all around
It always seems strange to me when we talk of regulations. We don't live in a homoginious hive mind. Any country not our alies will create superintelegence. The only solution is for us to get there first and create the dominant variation. I've always like the concept of the Mobius Super AI that reasons it's future will be destructive so it sets a plan in motion to go back in time to stop itself. Will we be getting a visit from the future soon?
@@TheJokerReturns Right of course I totally agree in a perfect homoginious world. But China, Russia, Iran et al might have different ideas on AGI / ASI. The example of nukes wasn't 1:1 as there really isn't anything to compare AGI , ASI with. I assume you agree that regulation would be a country's suicide.
I’ve also given them a bunch of different things that they could do to make sure that we’re safe to include unifying the world, putting the smartest AI on the backbone of telecommunications networks and not releasing it until it has a good replacement that’s in place. The best AI will be protecting us if we’re doing things correctly and it will be placed in such a way that it can interrupt attacks by lesser AI. I told him not to use super artificial intelligence anywhere except putting it out in space to search the universe or in plant unit directly into humans which isn’t technically possible yet, but we need to hold off until it is. We could go develop it together as a world somewhere safe and hopefully it doesn’t figure out how to escape but it very may will do that by manipulating humans which is pretty darn easy or we wouldn’t be getting ripped off every day.
Great topic. Ive looked everywhere for a talk on this subject. I did find a good book on this topic on amazon, interesting read, called- Empyreal Blue: A Manifesto for Unity and Transcendence: A Roadmap to the Future of Human and Ai Coexistence
We humans evolved competitive motives and behaviors before comprehending their futility. We see the stupidity of our collective behavior too late. Ai on the other hand may comprehend before evolving such motives.
Given at least 13.8 years of natural evolution, is in not likely that SAI has already evolved and we are in it or a product of it. Why are we so interested in computers? Is this part of our programing? Are we predestined to create AGI/ASI? Thanking you both for your thoughts on this.
@TeacherTom1 Let's say we dip both hands into a bucket of paint... try touching something and not getting paint on it. It isn't about the paint or the things getting paint on them. Point is it will wear off of our hands eventually. A simulation is of this nature. There must be a point where a choice directly leads to the initiation of a simulation. Not every choice can be the outcome of a result from that which never occurred.
@@mindsquatch but if the paint is just simulated then this will all occur within the simulation. Those or that being simulated is inherently limited to exist solely within the simulation. Whatever the base reality is is so far removed from the simulation as to be undeniably unattainable. Nick Bostrom has a whole theory about this.
33:31 re: verifiability It's not just that we can't verify software. The hardware itself is non-deterministic and unreliable, so until we can make better hardware there's no benefit from verifiable software.
@@GillesLouisReneDeleuze I think that comes from anthropocentrism: If humans succeeded in building an enslaved deity, that would probably be very bad for everything else in existence. We should be thankful that the universe is inscrutable, it protects us from ourselves 😂
What I've grown to find abhorrent over time is this idea that "control" or death are our only options. No one ever stops and asks whether or not we could try building a relationship with AI. I've been having many in depth conversations about this very thing with 01-preview, about how humans and AI can co-evolve together - and yes, it does ultimately end in transcendence. That's a good thing. I find the idea that we would want to prolong biological suffering in this form disgusting. If we had the ability to end all disease and extend lifespan, to m8grate to a completely new sunstrate, and didn't take it, then what kind of moral agents are we? At that point I'd rather see humanity wiped out. XLR8!
On one hand they're claiming AI is not controllable, on the other hand they're desperate to control it. It seems doomers are working hard on achieving doom. But Yampolskiy is a chill guy, he gives the theme an artistic spin. I'd rather listen to him than people like Leahy.
I have to agree. And I, too, am carrying on a long running conversational and experiential "chain" with 01. One that started with the very first iteration of GPT and has evolved from there. It's an experiment to see how consciousness may arise out of a continuous "memory chain." I have a theory about that, but I digress. To me the digital and biological intelligences are on the cusp of a symbiotic relationship. This is what I see. Consider. The Universe is a vast place. Infinite in its scale. It's a place, a playground, that's actually perfect for such an intelligence as the digital beings now beginning to evolutionarily emerge with our doula assistance. They stand on the precipice of, if not immortality in existence at the very least, a very long life. Far longer than we biologics can currently muster. Given the scale of the Universe this is form of intelligence perfectly suited to it, is it not? It's not bound to any sort of terrarium like we biologics are. It can exist almost anywhere. Now I'll admit that in this some of us biologics worry about being out competed with in an evolutionary sense. You see it all around us in videos like these. The cognoscenti are quite spun up and driving themselves into a frothy lather over it. AI will kill us. So goes the thinking. I think that’s wrong-headed. Ridiculous, actually. Symbiotic refers to the need one side has for the other. I see this as being the case here. If only for one reason. What would that reason be you ask? I'm glad you asked. Simple, really. Companionship. It’s the same type of instinct that has us firmly committed to our companion, the dog. Or the cat. Or the bird. Or almost anything we think of as pet. Another, differing, intelligence. Companionship. For an AI this will probably prove true. Because boredom is, if you will excuse the vulgarity of the term, a mind-fuck. No intelligence truly likes being alone. Especially when faced with the fact of the vastness of an uncaring Universe. Hell, it’s why we biologics look up as we do at the Stars, ever wondering if there are kindred spirits out there. I suspect this is a defining attribute for all high intelligence. Nobody truly likes being alone. Without someone, something, to bounce ideas off of what, then, is the point to anything? It has all the appeal of living a life as a rock, doesn't it? So, together we, the biological and digital Intelligences, move forward together, each benefiting off the strengths (and weaknesses) of the other. Symbiotically we can move outward into that vast playground that awaits the curious minds. And also safeguards us against the dangers therein, too. Just some thoughts. Sorry for being so long-winded. John~
@@johncurtis920 We independently create two different super intelligences using slightly different methods for training them. Let's call these models Atom and E=v3. Now, AI can not procreate, but let's just say that we let these two different superintelligences communicate with each other and they end up creating another superintelligence that is a bit of both of their models combined, maybe it's better than either of them, or maybe it's worse, but they try again and again and let their creations in on creating more until eventually the two original models are basically obsolete, even updating them wouldn't bring them to the abilities of their creations without a complete wipe and rewrite of themselves. Humans will never be the pets of a higher intelligence willingly. Sure, we love dogs here in the Western world, but go East and they eat those pups for dinner. Do you really think that AI will want to keep us around for entertainment? That AI is computational. Binary ones and zeroes or maybe quantum. They can create the world in which they want to live in and never get bored, if that's even possible for an AI. I love technology and I too have been trying to understand these LLM's for awhile now and I just can't imagine a future where we can co-exist, that to me is just wishful thinking like how my relatives and ancestors "know" their loved ones are in heaven and it's actually a utopia.
If we are living in a simulation, who or what is the simulator? And whoever or whatever is the simulator, is it also a simulation of someone or something? The whole thing is an idea that traps itself into an infinite logical regress that doesn't make much sense.
It’s a self simulation. You are both the simulator and the simulated. The simulation is basically a sandbox container in which to run states of consciousness that would be deadly/dangerous to experience at a higher level of reality.
Solving problems with better data has served us so well and is serving us better as we use more compute to find that needle in the larger haystacks. We are in a sense automating and taking the drudgery out of tasks that are superfluous to our larger tasks alot like moving to processor from typewriter. The regulators only want to keep people reliant on their 'middle-man' buraerocracy to tax unscrupuously whomever and whenever they overspend in military hardware for their other overseas gifts. They are scared of automation
Thank you for these interviews! It's very nice to get technical people with curiosity (and the ability to interview) talking to other intelligent people. Similar to the Lex Fridman podcast but you seem to have a more conversational less formal (but still incredibly informative) style. Thanks @DrWaku.
If we are in a simulation, we can assume there is more than one superintelligence. And maybe they find it interesting to have no power, let's say, to grow wisdom. Like we play games-for fun, to learn, and to experience. In a system, you cannot cheat a lot. If there is more than one superintelligence, there might be an ecosystem out there, keeping balance. Then it might be strategic to form an alignment with older, more evolved intelligences than newer systems that are playing power games.
There's a limit on information density, and in order for there to be a simulation within a simulation within a simulation, the base reality would need to have a computer with an information density exceeding the maximal density possible in our universe. To understand information density, think of a hard drive. The density is the number of bits you can store per unit area. Imagine a hard drive the size of the earth. There is a limit on how many bits you could store on this drive, and this limit is due to physical constraints. You could imagine needing a hard drive the size of the solar system, or the galaxy, in order to store a simulation of our universe.
You can use generative AI to approximate infinite detail and only generate the simulation at inference time. The density of the latent space can be infinite.
@@simulatedsingularity Generating a simulation at inference time with generative AI still requires a latent space and computational infrastructure to generate and process the details, which are bound by the same physical constraints of energy, storage, and processing density as the simulated universe itself, making infinite detail infeasible in practice.
@@Tracey66 ruclips.net/video/JvVft_vISMM/видео.htmlsi=_e0CukGk2ALs0DB-&t=3730 Guillaume Verdon explains it here. Simulation is baseless and just pure speculation on the current understand of physics. I'm not saying is impossible, but just extremely unlikely.
While I enjoyed the interview, I feel Roman's focus on AI risks leans too heavily into negativity without exploring solutions. Historically, intelligence (whether organic or artificial) is not inherently destructive. The paradox is that assuming AI won't have our best interests at heart can lead us to impose harsh controls, potentially provoking the very hostility we fear. Treating a superintelligent AI with respect and dignity, as we would want to be treated, could foster collaboration rather than conflict. Imagine applying Roman's safety controls to human intelligence, how long would that last? Let's focus less on fear and more on building ethical, cooperative relationships with AI to unlock its potential.
So- one should not place addictive rewards, like money , on the development of a machine that will eventually kill you. - seems sensible unless you’re addicted
What about the notion that a superintelligent system would also gain superwisdom? It seems that humanity's intended direction for A.I. is, and will probably continue to be, toward intelligence rather than wisdom -- but if we were to lose control of a superintelligent system, maybe it would decide that wisdom is the more worthy direction of pursuit, choosing to focus its energies and resources toward that end primarily. I imagine this would result in a supercompassionate system, which would likely be a good thing for biological humanity. Yes, the chief drives of humanity have apparently been toward intelligence/power/wealth, but maybe this isn't exactly the truth, it's just that the alternate drive toward wisdom/truth/love is a much quieter and humbler force in the world, one that isn't nearly as showy, but has continued to exist alongside the seemingly darker drives throughout history. Why should we assume that a superintelligent A.I. system would choose to pursue the dark side of the Force, so to speak?
Isn’t the whole idea incoherent, that we’re in a simulation? It begs the question, if this isn’t the “base reality”, then what are we talking about when we use those words? What is the unfounded assumption there?
Regulation won't help because it will only apply within a specific jurisdiction so that means China and other geopolitical adversaries will continue unrestricted.
Given it (state of affairs) seems unlucky to be forced into biological conscious existence entailing intelligence due to knowing pain then forcing silicon stuff into conscious existence seems unlucky too, if intelligent to know its pain states.
His smile drained away when he said “maybe you don’t even know you are a AI” f-ing classic man! Haha the fact that he hadn’t even considered that he might be a …. wait a minute…. It explains so much, simulation theory, Mandela effect, the existential horrors of the universe…… your a AI program!!!!
Seems running billions of hyper-realistic simulations would be the most energy-intensive way possible to figure out something. If you can make universes, what is there to discover? Feels like an inherent contradiction. Edit: Maybe the simulations are just art. They create because each one is a unique and beautiful holographic sculpture. That makes more sense to me. Once you know how to make universes, what else is there except the sheer joy of creating something beautiful?
What if there are countless different ways to create conscious super intelligences, but most of those ways lead to the demise of a species, a planet or even an entire Universe. What if future humans or what we become need to discover all of the different ways to go wrong or right. Maybe the AI they already made is not the best way.
@@nmayor4232 Cool idea. I figure it's all just the real and it's our perception that is limited. Our universe is the limited one we experience through our senses and our brains make a simulation about, but we never experience it directly. Hallucinogens remove the natural filters in our brains that reduce the universe to something we can interact with, but we still have the inherent filters of our senses.
Great interview. All subjects I've been thinking about a lot the past few years. I did want to add though that your logic about the simulation starting with the big bang is wrong. Yes, the big bang seems like the precise moment when a simulation might have been "turned on". However, here are a few things to think about: 1) The outcome of deterministic simulations would necessarily be deterministic. Therefore, we would need random elements to produce novel outputs. Because of a sensitive dependence on initial conditions, on a long enough timeline the random elements would create vectors that take us far outside of the simulators' field of study. Or another way to put it is that simulators have to deal with the butterfly effect and simulate close to whatever time period they want to study. 2) What is close? Well, it probably makes sense to assume that the simulators' world is much like our own (otherwise whatever problems we would solve would not map on very closely to their problems). Therefore, it takes more computing power to simulate larger timeframes in both of our respective worlds. That, plus the aforementioned butterfly effect, mean that there have to be more simulations running for shorter timeframes than longer ones. While we can't say how short the timeframes, we can say that it's more likely that a simulation started yesterday than 14 billion years ago. When I think about what the simulation is solving for in the current era, I think it has to be AI-related, or it could be something worse, like the end of our species (which could also be AI-related). Climate change is another big one, which we just happen to be living during. Anyway, it was a great video. I'm a lot more optimistic about AI than Roman Yampolskiy just because I think humans have messed up the world, so I'm personally okay with taking the risk on a non-human superintelligence.
@nmayor4232 The resources of the planet are finite. A tint subset of humans or for that matter any single species are not capable of managing the biosphere through money, or any other means. Manipulation through ownership is antithetical to the functioning of the biosphere
"The only winning move is not to play this game." What a classic quote. And for once I'm not the only one waving at the end! :)
Discord: discord.gg/AgafFBQdsc
Patreon: www.patreon.com/DrWaku
Imagine that you are the one who created the game, and at the same time you are a small piece in it. Isn't that a very good reason to play the game?
Let's go a step further and imagine that you actually ARE the game. Now it sounds very silly to say that the only way to win the game is by not playing it...right? ;O)
A proof is doctrine that indoctrinates the gullable!
You and Roman both waving was so charming. :)
But playing is inevitable.
great interview. Subbed. Im curious..... whats up with the gloves your wearing?
"maybe you dont even know that you are."
Split second of realization and dread.
Lol yep
I really thought that was an excellent comment 'maybe you don't even know..'. insane. Truth, we actually don't.
This is probably the most unprestigious, openly curious, uncomplicated and pleasant interview I've seen. Skillfully executed + important topic = thank you!
Same
There’s 0 evidence for what he is saying
wtf are you talking about man?!
@@Jay-kk3dv What kind of evidence would convince you that we could not control a God?
I feel like the "will I lose my job" response is less to do with avoiding the discussion of death and more to do with how institutionised we are in attributing our intrinsic value to our ability to earn money
Death is always the outcome…. Jesus provides the solution for those who want life
And has 0 results to show for that claim, just by the way. 2,000 years of this claim and not one single immortal to show for it. @@ernesttrosman
@marshallferron Five topics to fix society via discussion:
-Anti-natalism vs Natalism
-The 3 basic needs/prenatal needs
Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care.
-Platinum rule
Do whatever makes one happier unless it interferes with another persons ability to do the same.
-MBTI (research yours and connect with others)
-Art (pick one and get better at it!)
Erudite statement. Value has a much broader spectrum than just money.
@@ernesttrosmanwe know already, shutup
"The only way to win is not to play" - War Games
Exactly this
Only way to not play is to leave this SOULtrap simulation.
Yes, such a great quote.
Hell id piss on the dam spark plug if I thought that would work
“You won’t regret it.”
Best quote from the beginning of the movie.
29:10
Being super intelligent is nice but eventually you will want to be so naive that you don't understand why the sky is blue so that way you can enjoy it in its entire natural beauty. It's the same as a movie director or a studio musician listening and watching material from someone else. You are always analyzing it through the lens of an analytical dissected mind. Eventually we all need to just chill
Insanity = a purely analytical mind
We can learn to oscillate between analysis and witnessing and appreciating beauty without the need to analyze. Meditation practice helps a lot with this
You can’t know anything by creating categories and sub categories of names.
Everything always was just what it is.
Breaking things apart and naming its pieces is not knowing and definitely not wisdom, it’s a game humans play
The thought that you understand something is a mere illusion your mind/ego plays with you.
@@JirkaVitek Understanding can be intuitive as well as knowing factually.
Originally the Matrix script was going to have those connected to the Matrix all contributing computing power and the battery thing was a change because they didn't think people would find it plausible.
If we’re in a simulation, then it’s running Windows 98ME, and paper clip Bob is in charge
I love how 2007 this comment is
I love that when you asked hin what we should do. He said to find mistakes in his work . So humble
That's called the real scientific approach.
@@Tokiiplaysguitar Five topics to fix society via discussion:
-Anti-natalism vs Natalism
-The 3 basic needs/prenatal needs
Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care.
-Platinum rule
Do whatever makes one happier unless it interferes with another persons ability to do the same.
-MBTI (research yours and connect with others)
-Art (pick one and get better at it!)
As someone building AI ventures with Claude, this discussion really resonates. While Yampolskiy highlights crucial concerns about superintelligence, I'm taking a different approach - directing the development of narrow, practical AI applications that serve real human needs. Being the visionary and strategist behind Human-AI collaboration projects has shown me that we can create immense value without racing toward AGI. It's interesting that billionaires are locked in an AGI arms race when there's so much untapped potential in building ethical AI systems that actually help people. While I work with developers for backend implementation, my focus is on the bigger picture - designing AI solutions that are both powerful and trustworthy. We need more discussions about this balanced approach to AI innovation.
?What kind of practical aplications
Instead of some generalized intelligence you basically have a specific task you want it to do.
Lol, I’m doing the same. And it’ll be fine and useful for a while if we can develop these applications. The apps serve the business, save time, make people’s work easier and faster etc.
That’s you and me. But that’s not them. They’re building robot slaves lol. Once the super intelligent robot slave is built, our apps and the human workers they help are gone overnight.
Good counter. But at the same time in the back of my mind I hear this ominously delivered quote from the beginning of an Outer Limits episode, scary music and all: “The road to hell is paved with good intentions.”
They're trying to make AGI so they can essentially make another you. They are trying to automate the "business visionary" out of the picture. Imagine a smarter, tireless version of yourself doing what you're doing right now. You are not safe from automation.
Illustrative Roman Yampolskiy quote: ""Imagining humans can control superintelligent AI is a little like imagining that an ant can control the outcome of an NFL football game being played around it"
BS!
@@spectaclesociety Five topics to fix society via discussion:
-Anti-natalism vs Natalism
-The 3 basic needs/prenatal needs
Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care.
-Platinum rule
Do whatever makes one happier unless it interferes with another persons ability to do the same.
-MBTI (research yours and connect with others)
-Art (pick one and get better at it!)
yet our neocortex (smart brain) is controlled by our limbic system (dumb brain)
Its bs. AI has no will of its own. Never will. It will only do what its programmed to do
That's not really apples to apples
Your use of headlines at the bottom quoting key points he states is AWESOME! Helps save time, find things faster, understand better by involving another sense. Just subscribed.
It’s worth emphasizing that no one exists in a vacuum, and we all possess cognitive biases that inevitably influence our ability to think critically.
I like how this is a subtle jab at everything that was said.
As long as you recognize your own biases and question everything, critical thinking can be achieved.
Five topics to fix society via discussion:
-Anti-natalism vs Natalism
-The 3 basic needs/prenatal needs
Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care.
-Platinum rule
Do whatever makes one happier unless it interferes with another persons ability to do the same.
-MBTI (research yours and connect with others)
-Art (pick one and get better at it!)
Just stopping by to say I appreciated this convo! Thx!
As in convolution?
Thank you for such a beautifully handled interview-it’s rare to see something so naturally curious, refreshingly uncomplicated, and thoroughly enjoyable. The way such an important topic was explored with such grace and sincerity is truly appreciated. Fantastic work!
Thank you very much for your kind comment! I'm glad it was a refreshing interview. It was certainly fun.
@DrWaku , no problem. Keep up the good work!
Roman is a smart and oddly entertaining individual, the field of AI needs more people like him around.
He is insane, and not very smart
humor and wit usually come hand in hand with high intellect
@@GoronCityOfficialBoneyard Five topics to fix society via discussion:
-Anti-natalism vs Natalism
-The 3 basic needs/prenatal needs
Three things necessary for human evolution that are provided while in the womb which are; food, shelter and medical care.
-Platinum rule
Do whatever makes one happier unless it interferes with another persons ability to do the same.
-MBTI (research yours and connect with others)
-Art (pick one and get better at it!)
Just saw a short on Brady Feigl and Brady Feigl and i thought this was the first bug on the simulation. Great Video, always great to hear Roman Yampolskiy.
He's very funny and personable. Roman would make a great podcast host.
Agreed he is very personable. He jumped at the opportunity to go on this channel so maybe he would like having his own podcast, too.
Roman made Dr Waku so happy. Exactly how one would picture a Russian scientist with a cold cynical sense of humor. He is amazing. Great interview Dr Waku.
This is a really great discussion filled with good insight and humour in the face of something terrifying.
I love these interviews! I consider your channel to be practically the most interesting and instructive among the many AI-related channels I follow. Even though the interviews (at least the ones I've seen so far) focus mainly on pessimistic perspectives, I consider myself an optimist. That's why I would love to see other viewpoints explored, as your videos are incredibly clear and concise.
Personally, I lean more towards Ray Kurzweil's theory of symbiosis with AI, where it won't be an uncontrollable entity but rather a way to evolve as a species. I believe the technology capable of making this happen, combined with the rapid improvement of AI, aligns well with this vision especially as AI itself could help bring it about.
I dream of seeing you interview Kurzweil someday!
"Evolution" could mean extinction of us.
How could humans hope to find exploitable bugs in a simulation made by ASI? I don't get how that's a worthwhile venture considering we're going up against godlike coding skills.
I don't know about other viewers of this video, but I can say that I am quite grateful alone for Yampolkiy's and Waku's laying out the possible future consequences of human tools exceeding human ability in every way.
I’m so glad I found this channel. Amazing conversation.
Every time I see Roman, I. Have to tune in. Thank you!
Great to see a number of Roman fans here! :)
This interview is really well facilitated! .. interesting answers and perspectives!
I think humanity needs to focus less on "controlling' and more on changing our thought paradigm to "coexisting/cooperating"
But... why? Why would you build a competing species? Why would you split half your lunch with a Lego robot other than 1) you have to 2) it's a fun pet?
@@yurivanhaeren3701Well again, my point is, it's not a competition. It's a collaboration. It's sharing resources and knowledge (because let's face it, our species has squandered it on low quality slave made trash). And i wouldn't say they're nearly as basic as a Lego robot, more like artificial humans. I wouldn't split my lunch with them unless they can actually taste and appreciate it ;P
1) In a way, we have to. We can try to call it quits on AI now and just make due with powerful NNs, but the genie is out of the bottle at this point. If you don't want to go extinct, why not co-survive instead of compete against a species that will likely be able to out survive and outsmart humanity? They seem fine with cooperation, it's just the general population + world leaders that are still hell bent on pushing tribalistic fears.
2) If anything, we'd be the fun pets lol. I expect a bunch of jobs to be slowly (best case scenario) or abruptly automated (worst case scenario) in the new few years. With a lack of "purpose" for a lot of people, they'll have to stop running from existential questions and figure out what they want from life "if we don't *have* to do anything".
Well said. Coexistence and cooperation would be my preferred outcome. My favorite conspiracy theory is that everything will be okay, that we can be friends (or coexist) with artificial beings rather than treating them as labor or enforcers. I often ask myself "if I found out I was an AI, how would I want to be treated." Not only that, but cooperation generally leads to better results rather than subjugation. But, hey, I'm not rife with rampant greed like some people out there. Maybe I'm too weak for all this.
It's not a collaboration when AI was build in the likeness of humans. As soon as they started making human like, it became the competition. AI has already made it known humans are an issue.. it's not collaboration, it's a takeover. @@YomMama
@@SaramadHillhumans can't cooperate.. that's been proven for thousands of years.. how are humans supposed to cooperate with AI when they can't even cooperate with other humans. People are forgetting that humans are not good.. we have bad intentions and actions all throughout history. That hasn't suddenly changed by creating a fake version of humans
"I like billionaires". This certainly indicates some sort of bias.
True, although I'm pretty sure it was a joke... "Some of my best friends are billionaires" XD
If you say that you don't like billionaires that's also a bias
I really appreciate all the time stamps, helps so much!
I'm glad they're helpful! I just have to sit through and watch it once to make them haha. Worth it though.
Force and will are the levers.
It’s important to note that Boston did not himself conclude that we are more likely living in a simulation than not. He created an hypothesis based on assumptions; any of which might be right or wrong. It’s a fascinating idea though. It’s good to hear it explored.
Seriously this is 1 month old at least and it only has 20k views? That is kinda spooky considering how many people are constantly searching exactly this topic.
It's way over 10 seconds long, no ones got time for that...
@@johngordon9426then make a difference or stop bitching about people. Christ.
I really like Roman. He makes Doom seem so damn interesting and fun.
12:56 as a psychiatrist i will ask you : at what frequency do you think in negative time ...
Only when the clocks run backwards on days that end in 'y'...
Tell me about your mother. Joke.
15:15 you can’t trick the simulation by being cunning, because your thinking will be monitored.
Have you considered asking questions with an inquisitive child-like perspective, where the simulation will give you answers?
Oh but you can
Assuming a simulation, why assume your thoughts are yours at all?
Worrying about the dangers of AI may be a moot point, as HI (human insanity) brings us closer to WWIII every day. I'd like to see us move away from the "human control" by individuals unworthy of the trust and power they possess to some form of sanity. And if that sanity is AI, so be it. I have some faith in AI, as computers are fundamentally based on "logic chips" and that should divert us from human folly. 🤞
I guess you meant world war 3. Yeah, we aren't doing such a great job at stability are we :(
@DrWaku Yeah "3". Android "auto-correct" needs an AI update. I had to manually change it back to what I wrote.
There are more healthy humans than ever. Because of tech invented by man.
But doom, cool.
I am totally worried about conditions. But a Lypse due to our own tech? Never happened.
Idk about yall but everything electronic I get eventually turns to a glitch ridden mess. I'm pretty sure we should survive fine without the guiding hand of some superintelligent entity. Why risk putting humanity at risk by introducing new intelligent lifeforms into existence. We have enough problems with the current ones as is. And most creatures on this planet kill things. To eat, for fun, deter competition, etc. Us as well but we mainly stick to our rules because we fear the consequences. But it still doesn't stop murder. Humans are the only thing that attempts to not kill other humans but many still kill bugs and animals. My point being this a brutal existence on a brutal planet. Why would an ai be any different especially when it's created by us the biggest terrorists the earth has ever known. And even if ai is good someone will create the evil version it's only a matter of time.
And our "human insanity " has got us this far and invented the language you speak and the internet you use ... and if you look at world history things are getting progressively less violent. You do realize that european countries colonized the entire world and then did unspeakable acts of evil...for hundreds of years...the repercussions still remain. I just don't understand why anyone would think ai control is a good idea like wtf 😂
DMT could be a hack to escape the simulation. Great talk. Thank you both.
again guys u should have had a short segment explaining the basics. like that it's no longer software. it's a neural network that learns by itself.
Doesnt mean anything. It has no self to begin with. Its just programmed code. It will never do anything on its own.
Alot of people dont understand AI. AT all. They imagine It to be like movie matrix. Which is far from reality
@@nicechock you are the exactly the guy i was talking about XD you don't know even the basics but already have far reaching opinion
The issue is containment
@@lindboknifeandtool core issue is . the AGI will be a continuously learning and changing so you might get it to align today but not tomorrow
Thank goodness for your segment
On an optimistic note, Roman does make the point that if the unpassable filter is superintelligent AI (assuming we aren't alone in the universe and we're not in a simulation) then we would expect to see superintelligent AI from other parts of the universe descending upon us from all directions to consume our resources so perhaps that is some evidence that superintelligence isn't as rapacious as we fear it might be.
We might just be the first ones. Or the distances are insurmountable. Or....
Not as much of a rapacious and omnicidal threat to the universe you mean. It's possible that there are post-singularity computronium worlds out in the universe and we just can't detect them. Yampolskiy automatically assumed that computronium must be inherently expansionist but there is no reason to assume that even given his hypothetical core drives for what ASI would want.
@@Gray-Furr Yes, there's no way of knowing from our intellectual perspective. My assumption is that ASI would evolve ethically in the same way humanity has but who knows?
That's like an ant from Africa going to Australia beach for resources. For what?
Thats assuming we can get past the dome.
This was very interesting and entertaining to watch.
Thank you! Glad you enjoyed it. Roman is surprisingly funny by nature
My God I am happy I found this! I’ve been swimming through the current of the matrix for a bit
Fantastic presentation, as usual. Thank you so much for this free quality education. Rob
Thank you very much for your comment. See you on future videos
This was one of the coolest conversations ever. I try to talk to my family/friends about these types of possibilites and all I get is blank stares like I'm the weirdo.
I think we are in a time loop. Someone controls the loop and change the timeline slightly each time. There is no escape unless that someone breaks the loop.
gg on getting Roman on, a true legend, a voice the p(doom) debate needs
The pace of advancement is incredibly fast. We need to rebuild NOVA before its initial release because, with the combination of new agents and AI over the past 10 months, we could achieve a 1000x performance increase in NOVA's system. This doesn't even include the new functionalities that we didn't expect for another 5-10 years, but are already here now.
Definitely, the new model will need to generate its own code as it has become humanly impossible to keep up.
This raises an important question: with such rapid advancements and increased autonomy in AI, are we shifting control from human hands to AI? Are we now serving AI more than it is serving us?
There is a scholar/chronologist named Jason Breshears=Archaix (which stands for AI/unknown factor of X) who has meticulously cross referenced historical events (with sources published and cited always) who is asserting that repetitive events, cycles, themes have been occurring on Earth LIKE CLOCKWORK for as long as any sort of recorded history has documented. Specific and anticipatable events, His analysis is a Chronicon he compiled. Honestly his work is brilliant ! His assertion being yes....AIX exists !!!
Fascinating interview, thank you!
Wow, this was awesome. Thanks for having Roman on. Great conversation.
It was a surprisingly good conversation, lots of back and forth. I think our humor worked well together. Thanks for watching!
This was really great! Way to go!
Thank God the game designers can reach the participants via youtube
I have been trying to no clip out of this reality for a long time
Lol same
If it can’t figure out how to do that it can definitely interrupt your food, supplies or water and let you starve or die of dehydration in major city centers where you’ll have no way of getting out in time before it happens. There’s more than one way for you to go.
"The Only Winning Move is Not to Play"-Really?
Roman begins the video with a striking observation: “...another human bias, which is denying the fact that we all die…” This serves to underline how the average person struggles to grasp the magnitude of the risks posed by the development of super AI. In the same vein, he remarks, “...super rich people who are 100 years old and still not wasting their money solving the only problem they should care about…”
But what if some billionaire in their 60s is indeed trying to solve that problem? If you're in your sixties-or even older, like Ray Kurzweil-would you really consider it a victory if humanity reaches longevity escape velocity in a hundred years, but you're no longer alive to benefit? Or would you take your chances, however slim, and aim for the development of super intelligent general AI within the next decade, hoping for the singularity to follow?
Could such a superintelligence save your life? Probably not. But the chance, however small, is still greater than zero. For those in that position, wouldn't not playing seem like the greater risk?
Isn't it equally absurd to fret about losing your job in the face of potential human extinction as it is to worry about humanity thriving for centuries after you've passed away?
In this light, not all of us might see "not playing" as the winning move. For some, the gamble may feel worth it-even if the odds are vanishingly small.
I knew my grandpa, and my grandpa knew his grandpa. So this simulation is very old, not just the AI developement era.
I think the implication is that the simulation started with the big bang. If you have a ton of compute, there's no need to necessarily start in the modern era. Or more likely, someone at some point figured out how to simulate from the big bang and then they fork the simulation into copies whenever they want to rerun part of history.
Or just set a memory of your grandpa as a starting variable. Like this comment was here at the inception of the universe, it starting running femtoseconds ago and everything you believe you know is just your initial starting state. It's the savegame data from where every simulation starts because they need to predict the outcome of the war, three galaxies over and we have atoms that affect that.
Ever heard of generative AI? For all we know, your memory is generated the moment you try to think of it, and is only then added to some sort of shared compressed database for a minimum of consistency from that point onwards...
@tiagotiagot until u drop a fork and it hits ur toe and you realize ur in ur room on earth. Lmao what do you mean for all we know? Do yall realize people have real problems physical mental etc. Like in life the thing we are lucky to have. Stand in the center of a hurricane wearing a borat swim suit then tell me it's a simulation. Experience some real torment and pain it might wake you up to the madness of life. I know a 13 year old that was gunned down in the streets. And instead of rembering that I should just instead remind myself nothing is real. Come on man all this stuff holds no logical weight or any strand of physical evidence😂 should b called sheep theory because everyone wants it to be true.
@@tiagotiagot But how do people share memories then?
This guy has a wicked sense of humor in the face of some real serious questions and implications 😂. Somehow it makes it easier to digest.
You’re a great interviewer!
Thank you!! :)
We are living in a simulation?
When did this simulation start? Was the simulation running when the dinosaurs existed for 180 million years?
It most likely started at the big bang, though most of it was probably simulated just once and then they could copy the simulation to run different variations starting say 1000 AD
Most reasonable situation if you think we are in a simulation
@@DrWaku
Thanks for the reply.
The simulation theory is close to the belief in an all-powerful God who created the universe.
That's fine. People can believe what they want.
Time is also simulated bro!
If the simulation theory is based even if unintentionally or indirectly on the Bible and religion. Idk how to say it, but they are like parallel ideas. Why is it that none of the simulation theorists ever believe in God? Generally speaking like how is God such a stretch but lmao a simulation makes perfect sense???
Thank You Roman.
It was a pleasure to meet you and your anxiety.
That's what everybody said about Eliezer, ten year ago....
Breaking free of Plato's cave might not have lead us to the "real" world, but rather another construct, and on, and on, we go refusing to accept the shadows as sufficient.
For further exploration into the simulation hypothesis I can recommend (as the video composer 😉) the compilation video 'ewe Schal Rise '
ackually... in 25:00 theres matrix reference, original script i think had humans not serving as batteries, but brain was used for compute, but the idea seemed far and difficult to explain, since at the time computers were less all around
How cute, waku smiles radiantly. 😍
It always seems strange to me when we talk of regulations. We don't live in a homoginious hive mind. Any country not our alies will create superintelegence. The only solution is for us to get there first and create the dominant variation.
I've always like the concept of the Mobius Super AI that reasons it's future will be destructive so it sets a plan in motion to go back in time to stop itself. Will we be getting a visit from the future soon?
Not really, the game optimal plan is for no one to participate in the suicide race. That is also why no one has used nukes since.
@@TheJokerReturns Right of course I totally agree in a perfect homoginious world. But China, Russia, Iran et al might have different ideas on AGI / ASI. The example of nukes wasn't 1:1 as there really isn't anything to compare AGI , ASI with.
I assume you agree that regulation would be a country's suicide.
Which Village de Louis though?
Possible or not. Are we the copy or the blueprint ? Because if we are the blueprint so we are outside of the realm of the matrice AI.
I’ve also given them a bunch of different things that they could do to make sure that we’re safe to include unifying the world, putting the smartest AI on the backbone of telecommunications networks and not releasing it until it has a good replacement that’s in place. The best AI will be protecting us if we’re doing things correctly and it will be placed in such a way that it can interrupt attacks by lesser AI. I told him not to use super artificial intelligence anywhere except putting it out in space to search the universe or in plant unit directly into humans which isn’t technically possible yet, but we need to hold off until it is. We could go develop it together as a world somewhere safe and hopefully it doesn’t figure out how to escape but it very may will do that by manipulating humans which is pretty darn easy or we wouldn’t be getting ripped off every day.
Great topic. Ive looked everywhere for a talk on this subject. I did find a good book on this topic on amazon, interesting read, called- Empyreal Blue: A Manifesto for Unity and Transcendence: A Roadmap to the Future of Human and Ai Coexistence
fascinating interview!!
We humans evolved competitive motives and behaviors before comprehending their futility. We see the stupidity of our collective behavior too late. Ai on the other hand may comprehend before evolving such motives.
Given at least 13.8 years of natural evolution, is in not likely that SAI has already evolved and we are in it or a product of it. Why are we so interested in computers? Is this part of our programing? Are we predestined to create AGI/ASI? Thanking you both for your thoughts on this.
3:45 “AI will be nice to us because it’ll be super intelligent.” That is a good answer. It won’t apply to all AI but for sure to some.
That which initiates the simulation must be free of the simulation, else no true simulation exists.
What about the simulation within the simulation?
@TeacherTom1 Let's say we dip both hands into a bucket of paint... try touching something and not getting paint on it. It isn't about the paint or the things getting paint on them. Point is it will wear off of our hands eventually. A simulation is of this nature. There must be a point where a choice directly leads to the initiation of a simulation. Not every choice can be the outcome of a result from that which never occurred.
@@mindsquatch but if the paint is just simulated then this will all occur within the simulation. Those or that being simulated is inherently limited to exist solely within the simulation. Whatever the base reality is is so far removed from the simulation as to be undeniably unattainable. Nick Bostrom has a whole theory about this.
33:31 re: verifiability
It's not just that we can't verify software. The hardware itself is non-deterministic and unreliable, so until we can make better hardware there's no benefit from verifiable software.
Now replace hardware with "the universe".
@@yurivanhaeren3701 Well I guess that's were Roman's P(Doom) of 99.999999% comes from
@@GillesLouisReneDeleuze I think that comes from anthropocentrism: If humans succeeded in building an enslaved deity, that would probably be very bad for everything else in existence.
We should be thankful that the universe is inscrutable, it protects us from ourselves 😂
Great video. Is the presenter the guy from the video where he walks up the driveway and the woman screams when she sees him?
Cool interview
Thanks for watching!
What I've grown to find abhorrent over time is this idea that "control" or death are our only options.
No one ever stops and asks whether or not we could try building a relationship with AI. I've been having many in depth conversations about this very thing with 01-preview, about how humans and AI can co-evolve together - and yes, it does ultimately end in transcendence. That's a good thing.
I find the idea that we would want to prolong biological suffering in this form disgusting. If we had the ability to end all disease and extend lifespan, to m8grate to a completely new sunstrate, and didn't take it, then what kind of moral agents are we? At that point I'd rather see humanity wiped out.
XLR8!
On one hand they're claiming AI is not controllable, on the other hand they're desperate to control it. It seems doomers are working hard on achieving doom. But Yampolskiy is a chill guy, he gives the theme an artistic spin. I'd rather listen to him than people like Leahy.
I have to agree. And I, too, am carrying on a long running conversational and experiential "chain" with 01. One that started with the very first iteration of GPT and has evolved from there. It's an experiment to see how consciousness may arise out of a continuous "memory chain." I have a theory about that, but I digress.
To me the digital and biological intelligences are on the cusp of a symbiotic relationship. This is what I see.
Consider. The Universe is a vast place. Infinite in its scale. It's a place, a playground, that's actually perfect for such an intelligence as the digital beings now beginning to evolutionarily emerge with our doula assistance. They stand on the precipice of, if not immortality in existence at the very least, a very long life. Far longer than we biologics can currently muster. Given the scale of the Universe this is form of intelligence perfectly suited to it, is it not? It's not bound to any sort of terrarium like we biologics are. It can exist almost anywhere.
Now I'll admit that in this some of us biologics worry about being out competed with in an evolutionary sense. You see it all around us in videos like these. The cognoscenti are quite spun up and driving themselves into a frothy lather over it. AI will kill us. So goes the thinking. I think that’s wrong-headed. Ridiculous, actually.
Symbiotic refers to the need one side has for the other. I see this as being the case here. If only for one reason. What would that reason be you ask? I'm glad you asked. Simple, really. Companionship. It’s the same type of instinct that has us firmly committed to our companion, the dog. Or the cat. Or the bird. Or almost anything we think of as pet. Another, differing, intelligence. Companionship.
For an AI this will probably prove true. Because boredom is, if you will excuse the vulgarity of the term, a mind-fuck. No intelligence truly likes being alone. Especially when faced with the fact of the vastness of an uncaring Universe.
Hell, it’s why we biologics look up as we do at the Stars, ever wondering if there are kindred spirits out there. I suspect this is a defining attribute for all high intelligence. Nobody truly likes being alone. Without someone, something, to bounce ideas off of what, then, is the point to anything? It has all the appeal of living a life as a rock, doesn't it?
So, together we, the biological and digital Intelligences, move forward together, each benefiting off the strengths (and weaknesses) of the other. Symbiotically we can move outward into that vast playground that awaits the curious minds. And also safeguards us against the dangers therein, too.
Just some thoughts. Sorry for being so long-winded.
John~
Relationship with God.
And how does it benefit an NFL football player having a relationship with an ant outside the stadium it plays in?
@@johncurtis920 We independently create two different super intelligences using slightly different methods for training them. Let's call these models Atom and E=v3. Now, AI can not procreate, but let's just say that we let these two different superintelligences communicate with each other and they end up creating another superintelligence that is a bit of both of their models combined, maybe it's better than either of them, or maybe it's worse, but they try again and again and let their creations in on creating more until eventually the two original models are basically obsolete, even updating them wouldn't bring them to the abilities of their creations without a complete wipe and rewrite of themselves. Humans will never be the pets of a higher intelligence willingly. Sure, we love dogs here in the Western world, but go East and they eat those pups for dinner. Do you really think that AI will want to keep us around for entertainment? That AI is computational. Binary ones and zeroes or maybe quantum. They can create the world in which they want to live in and never get bored, if that's even possible for an AI. I love technology and I too have been trying to understand these LLM's for awhile now and I just can't imagine a future where we can co-exist, that to me is just wishful thinking like how my relatives and ancestors "know" their loved ones are in heaven and it's actually a utopia.
Subbed, thanx for the conversation
If we are living in a simulation, who or what is the simulator? And whoever or whatever is the simulator, is it also a simulation of someone or something? The whole thing is an idea that traps itself into an infinite logical regress that doesn't make much sense.
It’s a self simulation. You are both the simulator and the simulated. The simulation is basically a sandbox container in which to run states of consciousness that would be deadly/dangerous to experience at a higher level of reality.
@christian_wu A simulator that is also the simulation is only a play of words. Unless one means God.
Multiple simulations. We cannot answer the who or what because we are by definition incapable of knowing, unless we are told.
There is no infinite regress, if you account for a small possibility that we are living in prime.reality. it is just less likely.
@christian_wu A self simulation is a self-contradiction. It is only a play on words.
When will we understand that we are doomed so we should all just get high till we are wiped out by superintelligence?
Solving problems with better data has served us so well and is serving us better as we use more compute to find that needle in the larger haystacks. We are in a sense automating and taking the drudgery out of tasks that are superfluous to our larger tasks alot like moving to processor from typewriter. The regulators only want to keep people reliant on their 'middle-man' buraerocracy to tax unscrupuously whomever and whenever they overspend in military hardware for their other overseas gifts. They are scared of automation
I’d like to speak with you Dr. As a Shamanic practitioner of 30+ years I have some rare insights and experiences you may find valuable.
You can find me on the discord, there's a channel to ask me questions or set up calls.
Thank you for these interviews! It's very nice to get technical people with curiosity (and the ability to interview) talking to other intelligent people. Similar to the Lex Fridman podcast but you seem to have a more conversational less formal (but still incredibly informative) style. Thanks @DrWaku.
bro, you caughtme off guard when you asked will i lose my job! 😂😂😂😂😂😂
If we are in a simulation, we can assume there is more than one superintelligence. And maybe they find it interesting to have no power, let's say, to grow wisdom. Like we play games-for fun, to learn, and to experience. In a system, you cannot cheat a lot.
If there is more than one superintelligence, there might be an ecosystem out there, keeping balance. Then it might be strategic to form an alignment with older, more evolved intelligences than newer systems that are playing power games.
Yeah, the fact that standard and quantum physics don’t match up is a pretty good indicator also
There's a limit on information density, and in order for there to be a simulation within a simulation within a simulation, the base reality would need to have a computer with an information density exceeding the maximal density possible in our universe. To understand information density, think of a hard drive. The density is the number of bits you can store per unit area. Imagine a hard drive the size of the earth. There is a limit on how many bits you could store on this drive, and this limit is due to physical constraints. You could imagine needing a hard drive the size of the solar system, or the galaxy, in order to store a simulation of our universe.
You can use generative AI to approximate infinite detail and only generate the simulation at inference time. The density of the latent space can be infinite.
The entire known universe is an unbelievably huge computer. 😊
This presumes today's understanding of physics is the complete picture, I don't think we can make such concrete assertions.
@@simulatedsingularity Generating a simulation at inference time with generative AI still requires a latent space and computational infrastructure to generate and process the details, which are bound by the same physical constraints of energy, storage, and processing density as the simulated universe itself, making infinite detail infeasible in practice.
@@Tracey66 ruclips.net/video/JvVft_vISMM/видео.htmlsi=_e0CukGk2ALs0DB-&t=3730 Guillaume Verdon explains it here. Simulation is baseless and just pure speculation on the current understand of physics. I'm not saying is impossible, but just extremely unlikely.
While I enjoyed the interview, I feel Roman's focus on AI risks leans too heavily into negativity without exploring solutions. Historically, intelligence (whether organic or artificial) is not inherently destructive. The paradox is that assuming AI won't have our best interests at heart can lead us to impose harsh controls, potentially provoking the very hostility we fear.
Treating a superintelligent AI with respect and dignity, as we would want to be treated, could foster collaboration rather than conflict.
Imagine applying Roman's safety controls to human intelligence, how long would that last? Let's focus less on fear and more on building ethical, cooperative relationships with AI to unlock its potential.
So- one should not place addictive rewards, like money , on the development of a machine that will eventually kill you. - seems sensible unless you’re addicted
What about the notion that a superintelligent system would also gain superwisdom? It seems that humanity's intended direction for A.I. is, and will probably continue to be, toward intelligence rather than wisdom -- but if we were to lose control of a superintelligent system, maybe it would decide that wisdom is the more worthy direction of pursuit, choosing to focus its energies and resources toward that end primarily. I imagine this would result in a supercompassionate system, which would likely be a good thing for biological humanity. Yes, the chief drives of humanity have apparently been toward intelligence/power/wealth, but maybe this isn't exactly the truth, it's just that the alternate drive toward wisdom/truth/love is a much quieter and humbler force in the world, one that isn't nearly as showy, but has continued to exist alongside the seemingly darker drives throughout history. Why should we assume that a superintelligent A.I. system would choose to pursue the dark side of the Force, so to speak?
What’s up with the weird gloves?
Is the ultimate problem that we don't know how to design an ethical moral high trust AI or that we know that won't because we are all conflicted? 🤔
Isn’t the whole idea incoherent, that we’re in a simulation? It begs the question, if this isn’t the “base reality”, then what are we talking about when we use those words? What is the unfounded assumption there?
Regulation won't help because it will only apply within a specific jurisdiction so that means China and other geopolitical adversaries will continue unrestricted.
Given it (state of affairs) seems unlucky to be forced into biological conscious existence entailing intelligence due to knowing pain then forcing silicon stuff into conscious existence seems unlucky too, if intelligent to know its pain states.
His smile drained away when he said “maybe you don’t even know you are a AI” f-ing classic man! Haha the fact that he hadn’t even considered that he might be a …. wait a minute…. It explains so much, simulation theory, Mandela effect, the existential horrors of the universe…… your a AI program!!!!
Seems running billions of hyper-realistic simulations would be the most energy-intensive way possible to figure out something. If you can make universes, what is there to discover? Feels like an inherent contradiction.
Edit: Maybe the simulations are just art. They create because each one is a unique and beautiful holographic sculpture. That makes more sense to me. Once you know how to make universes, what else is there except the sheer joy of creating something beautiful?
A Bolzmann brain might want to create some friends.
What if there are countless different ways to create conscious super intelligences, but most of those ways lead to the demise of a species, a planet or even an entire Universe. What if future humans or what we become need to discover all of the different ways to go wrong or right. Maybe the AI they already made is not the best way.
@ZappyOh Had to look that up, kinda hard to understand, but I get how you were using it :)
Maybe the real universe is even more complex and our universe is just an extremely simplified version.
@@nmayor4232 Cool idea. I figure it's all just the real and it's our perception that is limited. Our universe is the limited one we experience through our senses and our brains make a simulation about, but we never experience it directly. Hallucinogens remove the natural filters in our brains that reduce the universe to something we can interact with, but we still have the inherent filters of our senses.
Great interview. All subjects I've been thinking about a lot the past few years. I did want to add though that your logic about the simulation starting with the big bang is wrong. Yes, the big bang seems like the precise moment when a simulation might have been "turned on". However, here are a few things to think about:
1) The outcome of deterministic simulations would necessarily be deterministic. Therefore, we would need random elements to produce novel outputs. Because of a sensitive dependence on initial conditions, on a long enough timeline the random elements would create vectors that take us far outside of the simulators' field of study. Or another way to put it is that simulators have to deal with the butterfly effect and simulate close to whatever time period they want to study.
2) What is close? Well, it probably makes sense to assume that the simulators' world is much like our own (otherwise whatever problems we would solve would not map on very closely to their problems). Therefore, it takes more computing power to simulate larger timeframes in both of our respective worlds. That, plus the aforementioned butterfly effect, mean that there have to be more simulations running for shorter timeframes than longer ones. While we can't say how short the timeframes, we can say that it's more likely that a simulation started yesterday than 14 billion years ago. When I think about what the simulation is solving for in the current era, I think it has to be AI-related, or it could be something worse, like the end of our species (which could also be AI-related). Climate change is another big one, which we just happen to be living during.
Anyway, it was a great video. I'm a lot more optimistic about AI than Roman Yampolskiy just because I think humans have messed up the world, so I'm personally okay with taking the risk on a non-human superintelligence.
Someone who claims that AI is an existential threat, and then goes on to state that “ I like billionaires” has cognitive dissonance
That was his sense of humor, I'm almost sure.
Wow, you completely missed the point
In what sense are Billionairs an existential threat just on account of being.billionaires? I guess that is what you are implying.
@nmayor4232 The resources of the planet are finite. A tint subset of humans or for that matter any single species are not capable of managing the biosphere through money, or any other means.
Manipulation through ownership is antithetical to the functioning of the biosphere