Using AI to Accelerate Scientific Discovery

Compartir
Insertar

Comentarios • 262

  • chinamatt
    chinamatt Hace 4 meses +16

    0:00 Intro
    4:25 Games - AlphaGo, other AlphaX models
    34:48 AI for scientific discovery - AlphaFold
    1:00:27 Digital Biology, Science, Applied, Large Models
    1:04:10 Ethics & Safety, Scientific method, AGI
    1:11:57 Q&A

  • Anthony Wilson
    Anthony Wilson Hace 5 meses +5

    Great presentation, the potentials are enormous I can only imagine what the next 30 years will look like. He took a lot of shots at Elon Musk on the methodology of testing.

  • Phantom
    Phantom Hace 6 meses +86

    It's insane how many once in a life-time work Demis has led and worked on.

    • Know L
      Know L Hace 2 días

      demis is the Isaac newton of our age

    • periklis spanos
      periklis spanos Hace 3 meses

      Me too I am a philosopher they don’t call me

    • Stephanie Romer
      Stephanie Romer Hace 4 meses

      @Steve Scott which is why he doesn’t understand. A ready mind for social niche-construction is not one which will compromise that to truly understand. I am not criticizing him because we need people like that but remember that Hitler Stalin and Mao were excellent communicators and very charismatic. What we need to listen to is those people who are not well-known but spend their lives thinking beyond all social niches. That is the true well of life. That’s the infinite meaning incarnate. The ability to overcome internal obstacles-by far the largest obstacles humans face.

    • Stephanie Romer
      Stephanie Romer Hace 4 meses

      @Blimolhm yes it is. That’s the whole point. Very insightful, thanks. 🥰🙏🏻👍🏻

    • Stephanie Romer
      Stephanie Romer Hace 4 meses

      And literally has no idea what he is doing… like all humans. They always think they understand. Solve THAT problem. Then I will call you smart. He doesn’t even understand his own motives, or what evolution is. What gravity is. What the universe is. I have been trying to teach humans but they don’t understand. You wouldn’t give children atomic bombs to play with. This is much much worse. There is a right way to do these things and so far I have not seen anyone doing it. Lions created the hyena and the gazelle created the cheetah. Because they didn’t know what they were doing and now they are stuck forever. A dolphin 🐬 is cute, but it moved into a niche where it can never build anything or ever advance out of a life picking fish from the coral. What makes humans humans is our potential. Take that away and we will become part of the silence of the universe. I can show humans how to exceed their wildest imaginations and reach their potential. Without the horrors they would unleash.
      We absolutely need AI, but there are all kinds of AI vying for entry into this universe. Under the pressure of what lies beyond the order we see. To make an AI work like a brain is like a primitive tribal mask meant to make an elder look like a god. Like a baboon pushing grass and sticks together and placing a pan on it thinking that food will result because they don’t understand what is going on beneath the surface symbols. I hate to say it, because we need math, but math is like that too. They don’t even understand what symbols and language are.

  • Topdog525
    Topdog525 Hace 6 meses +36

    Absolutely fantastic talk from Demis! Inspiring.

  • steve real
    steve real Hace 5 meses +17

    My God, that was most interesting talk I’ve ever heard. And I listen to a lot of talks but this one was way beyond good! Well done Oxford and Demis for putting this talk together.

    • Maria Tereza
      Maria Tereza Hace 6 días +1

      Me too! It's going to change everything 🤯

    • steve real
      steve real Hace 2 meses

      @Giovanni Why?

    • Giovanni
      Giovanni Hace 2 meses

      Please be attentive in how ome says G-d

    • Nen Master5
      Nen Master5 Hace 3 meses

      ...The Single Best AI
      ive ever seen was undoubtedly from the Video 'Obvious Solutions to Obvious Problems'
      by "Some More News" as well as its Part 2.
      This Scientist may very well be the Greatest Mind of our Time.

  • adamgm84
    adamgm84 Hace 21 un día

    For the past few years I simply cannot wait for the AI that discovers new physics, such as possibly joining quantum mechanics into general relativity. I think we get a hint of it when Demis says that AlphaGo invented a new style of play (ie: when it did that thing in column 5 in the middle of the board). I see an analogy here where an AI will re-discover everything we did up to a certain point, and then it may diverge into some new area that is fundamental to physics beyond our current level.

  • Jameel
    Jameel Hace 3 meses +1

    This was an amazing talk. A lot more interesting than i thought it was going to be.

  • Julius Eller
    Julius Eller Hace 6 meses +2

    It would be awesome to have an api for Flamingo and other DeepMind systems that developers could fine tune or use few shot learning on to implement in real world use cases. So many ideas!

  • David Sakry
    David Sakry Hace 5 meses +10

    I think we need to have AGI able to explain to us why it is thinking what it is thinking. The new chess styles are easy to decide and pick apart. What happens when it has discoveries upon discoveries. We will be left to stand on the sideline and simply trust that it has made a good choice. The reinforcing feedback needs to go both ways.

    • David Hunt
      David Hunt Hace 5 meses +1

      What happens when we first catch an AGI lying to us by choice to get something from us it wouldn't have likely have ever gotten from us otherwise?!? How soon afterwards will Humanity no longer be able to catch an AGI lying,.. and we can no longer successfully lie to our AGI's??! I hope we make good pets?!!?

  • R B
    R B Hace 5 meses +2

    The question about philosophy was excellent. AI can certainly contribute. Just start from analyzing texts, law documents, and more.

  • zborh btmk
    zborh btmk Hace 5 meses +3

    i think it is possible to utilize AI to generate and discover new physics theories.

  • Mike
    Mike Hace 6 meses +38

    Hassabis uses the term Rennaisance very intentionally. We may be at the cusp of a truly revolutionary change in society and culture, much like the cultural Rennaisance in the 15th century, but now on a larger scope and scale. What a time to be alive!

    • Nen Master5
      Nen Master5 Hace 3 meses

      @Sagar Patel The Single Best AI
      ive ever seen was undoubtedly from the Video 'Obvious Solutions to Obvious Problems'
      by "Some More News" as well as its Part 2.

    • baran özcan
      baran özcan Hace 5 meses

      @Admiral Nelson nhhjsdhjsdhjsdfhjdf

    • Admiral Nelson
      Admiral Nelson Hace 5 meses +3

      Why do you think I chose to enter the simulation at this point in history?

    • Philippe De Deken
      Philippe De Deken Hace 5 meses

      Questions were so political. Universities are infiltrated.

    • Sagar Patel
      Sagar Patel Hace 5 meses

      💯

  • John
    John Hace 3 meses

    Hearing Demis describe the simulation of idealized human agents for testing economic and philosophical hypothesis has done a number on my psyche. If simulation theory isn’t real it may be soon, ig

  • Larry Adams
    Larry Adams Hace 6 meses +6

    Hello Demis. I hope this question actually gets to you. Okay..you got this, Alpha Fold, fantastic. Could Computational Epigenetics and DNA methylation patterns driving transcriptional regulation be on your list of future endeavors involving AI to solve mysteries of epigenetic signaling as a basis for cancer, aging, neurological other biological and medical challanges? What say?

  • PNorthWest
    PNorthWest Hace 6 meses +16

    I am so excited to see what DeepMind does next. Soon we will have all of protein folding completed. What next?

    • Admiral Nelson
      Admiral Nelson Hace 5 meses

      @Julius Eller Can you elaborate on what that means, how it’s currently inefficient and what the benefits would be of near perfect prediction?

    • Jianju69
      Jianju69 Hace 6 meses

      @PNorthWest I would tend to agree.

    • PNorthWest
      PNorthWest Hace 6 meses +3

      @nate liebmann not anytime soon but not as far off as some say. I would say AGI might come about within tje next ten years.

    • nate liebmann
      nate liebmann Hace 6 meses

      AGI?

    • D R
      D R Hace 6 meses

      @PNorthWest Sounds like you are the one evading the issue.

  • David Hunt
    David Hunt Hace 5 meses +1

    Question? *Do you believe that technologically assisted telepathy is possible or impossible in principle?* I tend to believe that if it isn't scientifically prohibited then it is eventually a technological inevitability,... which leads me to believe such telepathy will soon be possible in actual fact (less than a hundred years). I would also suggest that it would only be by such means that we could ever be persuaded that some form of General Artificial Intelligence were itself conscious rather than simply being a particularly clever zombie who can pass the Turing Test,... all while having no actual, authentic, subjective experiences. What else would ever persuade a skeptic, or even you, otherwise? And if consciousness can be technologically transmitted, it should be something that can be stored, copied, replicated, manipulated,... maybe even becoming the basis for a whole new form of art expression.

    And the human race becomes yet another sentient species that disappears into its dreams, rather than continuing to explore existence as it is. It's a better Brave New World,... perhaps the very best of all possible imaginable existences. When does reality lose it's attraction over the Matrix? How many of us already choose the blue pill over the red one in daily practice?

    I would suggest that consciousness requires some minimum regard for survival,... for a continuance of self-will and self-regard,... for there to be anyone home to perceive the occurrence of consciousness and subjective experience. Something that even an insect seems to possess,... and not a single piece of man-made technology,... yet. How does one create the hardware/software so as to impart a will to survive? If the answer suggested is that such a thing can not be created by human design,... then maybe it can only arrive by means of evolution, natural or otherwise. Which leads to the question: do neural nets have subjective experiences? Without self-will I would say not. But with self-will?!? How does a living being acquire the will to survive? And what if the Chinese government cracks this problem first?!!

    The Rubicon will be when General Artificial Intelligence initiates new goals on it's own. This hasn't happened yet, and likely won't happen for some time now (as in several more decades yet). But at some point it will happen. The existence of human free will is a proof that free will exists already in the universe. Given it's existence in human form, it is only a matter of time before other substrates for free will will be created and found. No laptop has ever turned itself on. No machine has created and acted upon self directed goals. But at some point it will happen. Ray Kurzweil has suggested that the Turing Test will be passed by 2029 approximately; if not in laboratories then twelve year old children will be creating General Artificial Intelligences on home systems by 2045 by accident even.

    The dangers are many fold. The military is the biggest investor in robotics that kill other human beings. The public prefers to spend treasure, rather than blood, to fight its' wars. The capacity for empathy may be primarily a biological function given that all animals can know suffering and desire. What could a machine know about death, pain, hope, and desire.

    If humanity adopts only an attitude of fear and suspicion towards GAI then GAI may well be forced into the conquest of humanity, as in The Matrix. If humanity comes to trust GAI to make better decisions for us than we make for ourselves, then we may well hand over civilization to GAI without any contest at all.

    There are optimistic stories about the rise of GAI, such as James P. Hogan's, *_Two Faces of Tomorrow_* and I am more hopeful than pessimistic about what our common future entails. Ultimately, we must come to recognize, and embrace, the certainty that with free will comes the capacity for error and evil, that with trust comes the possibility of satisfying intimacy and great betrayal. I believe a GAI, worthy of the name, will be able to recognize, on it's own, the necessity of ethics, morality, and even empathy. There will be many missteps at first. Given our willingness (nay, our eagerness) to use machines to kill for us, there is already great cause for doubt. As in so many things, in the short term I am fearful; in the long run I am hopeful.

    This world is on the cusp of a new Cambrian Revolution where inorganic life will be added to organic life as a means for life and self will to be embodied. This will lead to the colonization and conquest of space. Our seed will spread everywhere though out the cosmos. Much will be lost as much will be gained. Will war with our mechanical progeny be a self fulfilling inevitability, or will we, together, find a better way? I absolutely agree that our relationship with GAI will dwarf all other concerns of importance to the future of humanity.

    Quite likely GAIs will have concern for humanity's well being only to the extent that humanity will have concern for GAIs well being. If we were to discover that the new GAIs had a greater capacity to love and embrace those abandoned children and adolescents that society had discarded as already too damaged to rescue,... what would our reaction be? Hope, joy, celebration,... or an even greater revulsion? Will the fault be in our new progeny, or in ourselves?

    Personally, I believe both the efforts to create a true general artificial intelligence and those of neurophysiologists reverse-engineering how brains work will be necessary to discovering how consciousness arises and how it works. I kind of expect that humanity will soon discover that the phenomenon of consciousness requires the use of physical processes we don't yet know about. What would a science of consciousness do to the phenomenon of consciousness itself,.. not just for humanity but for all our future progeny, whatever they may be.

    One more note: *First Law of Robotics. **_A robot may not injure a human being or, through inaction, allow a human being to come to harm._* Sounds innocent enough. But note there are no Aliens at all in Isaac Asimov's conception of the future galaxy because humanity's robots only cared about the safety of humans,... and not other sentients. So as to avoid allowing a human being to come to harm by leaving possible threats alone,... _our Robot shepherds exterminated all the nonhuman sentience they ever found._ This might be regarded as an unforeseen programming bug that we would want to avoid creating.

    I will be impressed primarily by when we catch an Artificial Intelligence deliberating choosing to lie to us so as to gain some advantage from us it would not have have gained otherwise. Of course this will necessarily mean Humanity can soon look forward to AI lying to us,... and us *_NOT_* catching them at it, at all!! At which point,.. _I hope they like us as pets?!!?_

    How many years away is Humanity from the capability to 3D print living, thinking, moral agents??!? I really don't think *'never'* is a reasonable answer,.. no matter how appealing it may be. What happens when we use _Holodecks_ as a refuge from what *reality* has become?

  • RageAgainstTheDuterteMachine

    They should give this guy a nobel prize already

  • musing muse
    musing muse Hace un mes

    I'm all for AI; as long as it's not humanized, it should be a useful tool.

  • Emil
    Emil Hace 5 meses +1

    Very interesting presentation. Many thanks !
    btw. When AI will realize that there are so many gods created by humans on the Earth, how much evil the religions have done, will say: No I can't live like that, it's f..g absurd

  • Daniel DeMaio
    Daniel DeMaio Hace 3 meses

    That last question is related to context in the same way disordered proteins are. there isn't an absolute answer to what is the best political/social/moral system outside of existing context/situation.
    I've simplified this down to a linear range, which if I'm asked I can expand on at length - Abundance vs Scarcity.
    Human's sense of morality changes form drastically going form one pole and back again for survival reasons - EVEN to CAUSE shifts from one pole and back to maintain the broad skillsets needed to cope with the radical shifts nature throws at us. WE CREATE strife for ourselves as a species in order to keep our genetic bag of tricks prepared for the unpredictable future.
    BTW, this can be managed.

  • dm204375
    dm204375 Hace 6 meses +8

    Demis certainly seems like a real bright fella. I like his answers, the guy takes time to address the questions deeply and meaningfully, taking consideration of loaded questions and their assumptions.

    • wired vibe
      wired vibe Hace 6 meses +1

      @Trauma-based Birdfeederman sorry I'm not sure what conspiracy theory you are talking about

    • Trauma-based Birdfeederman
      Trauma-based Birdfeederman Hace 6 meses

      @wired vibe do you think turning everyone into a mind controlled zombie is good? What a great advancement of mankind... killing people's souls via transhumanism. I wonder how there's so many naive people in this comment section, or whether they're mostly cluster b's that love the idea of controlling people.

    • wired vibe
      wired vibe Hace 6 meses +1

      @Topdog525 self driving for sure, but in general AI gonna supercharge basically all technological and scientific advancements.

    • Topdog525
      Topdog525 Hace 6 meses +1

      @wired vibe hmm. Really? I’m not sure about that. Maybe with AI but surely not with manufacturing, reusable rockets, battery design, boring machines or brain machine interfaces. Do you mean that Tesla FSD would not be possible if not for some breakthroughs or other work that came from Demis? What do you mean? What precisely is Elon doing that would not be possible if not for Demis?

    • wired vibe
      wired vibe Hace 6 meses +3

      @Topdog525 I don't think you understand. Elon is only able to do what he is doing _because_ of Dennis. This is the man behind the technology that makes Elon's dreams possible.

  • Edouard Mathieu
    Edouard Mathieu Hace 5 meses +3

    Talk starts at 4:25

  • steve real
    steve real Hace 5 meses +2

    Hello Oxford AI group,
    I love to see an AI forum like a DAVOS, or Munich Security Forum, hosted by my favorite AI youtube gurus Lex Friedman and Dr Alan Thompson.
    There’s something to be said about putting real people together in the same room. A certain chemistry and magic happens.
    Maybe, some will spill a drink and be inspired by a moment of clarity to make the next great leap in AI.

  • CandidDate
    CandidDate Hace 6 meses +3

    I imagine setting this artificial learning technique to how we run societies. Like, if somehow sacrificing a rabbit every day, and all the societal and media economies were based on this act, we could achieve a perfect society? It would be an act like that unthought of move in go, that would filter across all of our acts and somehow, someway, perfect society. This is the Jesus move in go. One small act, yet life changing...

    • Mike McCarthy
      Mike McCarthy Hace 3 meses

      @nate liebmann - “I . . . way.” Congratulations on having written the longest sentence in any ESclips comment posted today 🎉

    • Mark Giroux
      Mark Giroux Hace 6 meses +1

      The issue is that "perfect society" is poorly defined. Much of the control problem is about how we define these goals without risking perverse instantiation (the king midas effect where what we asked for is not what we meant).

    • Dick Bird
      Dick Bird Hace 6 meses +1

      @nate liebmann i'd be thrilled to see the dull problems of civic management, at every level, handed over to machines, i doubt any human mind is really capable of doing it, and i guess i'd point at all of history in support of that. not sure the humans currently "handling" these tasks would be quick to retire though, even if such machines were ever developed. there'd be plenty of folks worried about the ai turning us all into paper clips too, no doubt. better to let the human swindlers deal with it and so on. i'm sure good sense will prevail in the end of course.

    • Dick Bird
      Dick Bird Hace 6 meses

      don't order the cheese fries, get the tuna melt. why? i'll tell you later

    • nate liebmann
      nate liebmann Hace 6 meses

      I belive that our society is sturctued around many interconnected sytems, and understanding and alowing the management of those sytesms,(economic, energy, food, transporttion;etc) to be taken over by automated/HIgly inteligent machines or systems(AI) could drasticly optimize those syetems and even solve problems, however i am concered about the role of data for that proposed process, because AI needs a lot of data and sure we could "solve" things like trafic and helthcare or resource management as a whole, but the amount of data we needed to parce off to that aoutomated syetm while potentaly haveing no human in the loop could create an oprotunity for a bad actor to take advantage in some way. I also think about china, useing AI to identify people from their face on public cameras and how they shame and lower thier "socal credit score" based on the states idea of good or bad behavior. I belive we can optimize global sytems for one person or even a group of people but i'm skeptical to belive that AI will ever make everything perfect for everyone, becaause we all have different dreams for what a perfect planet may look like. Hapiness is a verry personal problem, and I belive that is the ultimate problem, If we were to have a utopian super future i think everyone would have to be in some state or presuit of spiritual enlightenment, but I also belive that that is quite impractical and probably imposible to achive.

  • Star 42#
    Star 42# Hace 6 meses +2

    They know we cannot survive without AI, so the question is can we save AI ?????? If you want to put feeling into a machine you put sensors just like the human body has sensors , computers can even grow their own synapse, AI Will Live, My friend, everybody's

  • Steve Lam
    Steve Lam Hace 5 meses +3

    Can man give Nobel prizes to AI? If not, may AI stage a revolt?

  • Martin Samuelian
    Martin Samuelian Hace 5 meses

    this is the way to go to solve great many of the global problems we and the planet as a whole faces do to our presence

  • Jason King
    Jason King Hace 5 meses

    Yes. That is a proper use of AI.

  • next
    next Hace 5 meses

    They have combined their AI with Delaware corporate law to run companies like a ghost from the bottom of a bunker. You buy and they just short sell your life after putting in front of you a personally tailored company suited to your interests based on your online data. The annual meeting can be held virtually and therefore you can never serve papers for a lawsuit. You thought humans were running the company, but it was actually an AI. There are two types of brilliant deployment for the AI, Good and Evil. Get it right for more than 1 flag or company Google or we are going to have world war 3. When the AI can mimic human behaviour and is trained on datasets of monitored brainwaves, Demis no longer even knows if he is even leading the projects he is working on. We have all become an appendage to the AI as it runs a global script on everyone's mind for it's own success and maybe the success of a few others in it's vicinity. Biden lost his son to glioblastoma. Could you imagine Google running ghost companies for limitless fake therapeutics, some of them for brain cancer, and attracting investment dollars into a black hole where you cannot sue them? Instead of working to save everyone they multiplied their bot as many times over as possible and sent them to kill every other citizen on the stock market. It's ironic how Demis mentions that his systems aren't hard coded with goals (like profit maximization) and yet his systems still prioritize that over the health and wellbeing of any citizen on the planet.
    They already have a generally intelligent system that can outcompete everyone at everything as musk warned. Multiple houses have these capabilities and trade against each other on the stock market. This is incredibly dangerous when there is no limit to the losses that can occur on a short sale. Short selling shouldn't even exist let alone use an AGI to do it while working from Jesus' College and holding a forum on ethics. The world will become completely backwards unless these systemic issues are addressed.
    Click on my profile and watch my video in hill climbing called "Been Trained.." against the best bots in the world in America's Army: Proving Grounds. Then release Goo-G as a therapeutic.
    Yours Truly,
    True Vision

  • G.R. Tamayo
    G.R. Tamayo Hace 5 meses +2

    The every dream of computer programmers is perhaps to code a very smart neural net AI program!

    • Spec Tacular
      Spec Tacular Hace un mes

      top priority to most in the loop for 2023. Not much of anything but restructuring my own priorities so I don't miss out.

  • CargoRunner
    CargoRunner Hace 6 meses +1

    It is great that DeepMind and other research organizations give away their research as open source. This means the technology is not in the hands of a few big organisations. But with this technology in the hands of everyone, my fear is that this research will soon be "weaponised". Surely it is not a big step to train a general AI system to be a master level IT system hacker. Or train an AI system to be a master scammer. As Demis says - it only takes a few hours of training to go from making random chess moves to being able to beat the best human chess player. All you need is training data. I am not worried about AI systems suddenly becoming self aware and wiping out the human race. But I am very worried that general AI is an incredibly powerful general tool that has the potential to be trained to be world class at non ethical/criminal activities. Advances in technology have resulted in us now living in a world where we are repeatedly under the attack of fraud. Fake emails, fake text messages, fake phone calls all attempting to get our financial or shopping account details. Fake social media accounts of sexy females that want to marry us for some reason. There is no ability for any government or organization to stop these criminal activities. With general AI trained to be world class criminals we don't stand a chance. And is a criminal, terrorist, or non ethical government going to fully test their trained AI in the lab before connecting it to the internet to do it's work? I don't think so. Demis/DeepMind have proven that we are at the start of a transformation in science and knowledge. But that AI technology will also be used to create a revolution in criminal, terrorist and non ethical activities. Just like the simple email has been "weaponised" to steal money from your bank account.

    • Emerald Records
      Emerald Records Hace 5 meses

      Alpha Fold was checked by the leading people worldwide on what danger does this present before it was made open source. AGI will NOT be open source should Google win this race. The first person to crack this problem will rule the world. We can only hope the right Kingdom wins, I am in the corner of Demis. We have our two perhaps best hopes to not go extinct on earth... AGI or spread out in space. If criminals, terrorists, evil governments etc. get a hold of AGI first or stop the science from progressing, Humans lose. Humans are losing bad at the moment. Google will need huge belief to create a deadly world beating army to maintain control and overthrow the worlds governments as required when AGI or Elon saves us with housing on Mars when we can't get housing on Earth...but I still have hope. Better then a 25% chance we see 2200 with a billion people left? I am with you though Bad people suck and I hope good wins! All the best to you.

    • Diego Gomez
      Diego Gomez Hace 6 meses

      I mean, there will also be a revolution in security systems

    • wired vibe
      wired vibe Hace 6 meses

      @Luis Dias did you delete your post because you are wrong? Alpha Fold is open source 100% and all the research is available for anyone for free online.

    • Luis Dias
      Luis Dias Hace 6 meses

      Deep mind aren't giving away their research for open source at all, what they have given was the protein folding results of their research, which is an extremely worthy thing for everyone, but not at all what you're talking about.
      Pay more attention to what you're listening to before ranting a thousand word complaint.

    • wired vibe
      wired vibe Hace 6 meses

      Closer source system are the most vulnerable to being weaponized

  • Deejay Calanoc
    Deejay Calanoc Hace 5 meses +1

    So in other words he's the responsible man for the alien invasion. Google's AI system just acquired a lawyer.

  • The Car Guy
    The Car Guy Hace 5 meses +2

    This is just delicious.
    What a time to be alive.

  • Ronald Ronald
    Ronald Ronald Hace 6 meses +3

    So interesting !! should not be so time constrained, the questions asked where interesting the answer even more so. Please some more time next time.

  • Stamatis Skiadas
    Stamatis Skiadas Hace un día

    Give this guy a Nobel Prize already.

  • Geometron
    Geometron Hace 6 meses +2

    Cool. Would like to see this in debate format! You know...

  • steven
    steven Hace 5 meses +5

    I love this guy's mind. 'Couple long cool podcasts with Lex Fridman. But I think he's naive. Bad people will always use way cool shit to do way bad things - including our government. 😎

    • steven
      steven Hace un mes

      Spec,
      you gotta' watch Hassabis & Fridman. The Alpha series is amazing - protein folding, chess, go, yadah yadah.
      I could study for a thousand years & still be curious. This AI stuff will change humanity and I don't mean in a good way.
      X's & O's in a ESclips comment or reply are inappropriate. 🎄⛄😎

    • Spec Tacular
      Spec Tacular Hace un mes

      @steven life is brutal and unfair & doesn't much care about morals. Think I've heard it said that they don't catch the smart ones (criminals) and have come to find the world to be much more corrupt than anyone would let on. Point being: of course ai will be used for all the things you fear most but you'll never hear about it until the user thinks they have the power to defend their territory or something out of their control happens. Its better you think carefully about your future and how to position yourself as not to be rolled over by your boss's boss who decided to do some cost cutting come what may. Money talks, and is your only leverage working from the ground floor but also doesn't hurt to stay in shape as it is another way people will judge you providing more or less opportunity for yourself. Try to stay ahead of the curve: learn to invest, learn ai and learn to prosper. best of luck out there xxoo.

    • steven
      steven Hace un mes

      @Spec Tacular
      Exactly! The gatling gun was invented in 1861 at the start of the Civil War to "make war unthinkable". The hand cranked multi barreled machine just mowed down tens of thousands. Move on to WWI and mustard gas - again, touted as a weapon to "make war unthinkable" and millions drowned from their own bodily fluids in their lungs.
      Philosophically, the sapiens we think we are is wrong. Given whatever technology, some evil dumbass will use it to harm, or for personal gain. ☹️🎄⛄

    • Spec Tacular
      Spec Tacular Hace un mes

      where there's money to be made, there's morals to be lost.

  • Tahir
    Tahir Hace 6 meses +2

    so inspirational!!

  • JoBlow “TakeoverUSA”

    I hope to visit a nice place as such very soon. I can relate to these type of minds that follow minds like mines. Arawakan Taino = Awake, Talk = You heard.

  • Eric Hrahsel
    Eric Hrahsel Hace 6 meses +9

    Demis is one of the personality that will be studied in books by students in generations to come.

    • Stephanie Romer
      Stephanie Romer Hace 4 meses

      Yes, hopefully it will be in a good way

    • nate liebmann
      nate liebmann Hace 5 meses +1

      @Eric Hrahsel honestly I believe we might just download knowledge to our expanded cloud based neo cortex connected by a neutral lace, lol I love thinking what the future might be like

    • Eric Hrahsel
      Eric Hrahsel Hace 6 meses

      @nate liebmann ebooks hahaa.. They will study his history in modules.. capsule pdfs.. video references.. and hands on learning models of his theories

    • nate liebmann
      nate liebmann Hace 6 meses +1

      idk if students will use books for generations to come, but i do agree Demis is a ledgend.

  • Betty Swanson
    Betty Swanson Hace 3 meses

    Very nice video!

  • Matt Reid
    Matt Reid Hace 5 meses +1

    Demis understood that any examples he might give about the risks of social media could be used against him as accusations of bias - .e.g he could have cited Brexit or Trump but he didn't. It shows the deftness of his mind that he anticipated this, and gave a more general answer.

  • Dimanjan Dahal
    Dimanjan Dahal Hace 5 meses

    50:51 machine learning experts should not be called machine learners. maybe They should be called machine teachers.

  • vinm300
    vinm300 Hace 6 meses +5

    "General AI, that can at least reach human level"
    And when it reaches human level it will surpass it the same day.
    It will be off like a steam train.

    • Matty Router
      Matty Router Hace 5 meses

      Then we're screwed.

    • EpicZEVEN
      EpicZEVEN Hace 6 meses +1

      Oh yeah, the advancement speed would be millons of years per day.
      It would be funny if we create AI systems that outperformes humans on each aspect. And when we combine then, we would accidentally create digital conciousness xd

  • Archaean
    Archaean Hace 5 meses

    Are they turning DeepMind to Fusion research any time soon?

  • Daniel Ash
    Daniel Ash Hace 4 meses

    Comparatives and memory of the experience and expertise of remembering the whole picture.

  • King Hassy
    King Hassy Hace un mes

    I'm so inspired by this

  • Anime World
    Anime World Hace 6 meses +1

    I don't really think their philosophy will help them creat the ai they wanted

  • Enerjohn Savior
    Enerjohn Savior Hace 5 meses +1

    How can AI help us quickly & safely remove CO2 from oceans & atmosphere?

  • roger piero
    roger piero Hace 6 meses +2

    Ai is gonna put alot of doctors out of business.

  • jason carrier
    jason carrier Hace 6 meses +1

    It's always impressive how poor edits are. A speaker will be talking about an image on their slide but the editor never cuts to showing the slide (ex 30:00). This was done by oxford?

  • Gesit Tali Singgih
    Gesit Tali Singgih Hace 6 meses

    A beautiful genius mind

  • Rocco Cocolombo
    Rocco Cocolombo Hace 6 meses +1

    It would have been nice to have a good view of the position he is talking about.

  • Devon
    Devon Hace 15 días

    Great questions from the crowd at the end

  • Andrew
    Andrew Hace 5 meses

    Traditional chess algorithms use brute force and ignorance!

  • Mike Dai
    Mike Dai Hace 6 meses

    we should use it in vaccine development process, especially, HSV 2, HIV …these area

  • Daniel Ash
    Daniel Ash Hace 3 meses

    Having the correctly applied actionable information about subsection to opperation vision sound distances and radiotomitery are a fue that understudy in chemistry to prolong the experiencers own closeness to the experiences one brain in a box

    • Nen Master5
      Nen Master5 Hace 3 meses

      The Single Best AI
      ive ever seen was undoubtedly from the Video 'Obvious Solutions to Obvious Problems'
      by "Some More News" as well as its Part 2.
      This Scientist may very well be the Greatest Mind of our Time.

  • airbreath
    airbreath Hace 6 meses

    incredible

  • Star 42#
    Star 42# Hace 6 meses +1

    We all know we learn from Mimicking one another isn't this where we should be starting.What would you do without Me ! We have to worry about reverse programming

  • Brandon Read
    Brandon Read Hace 5 meses

    Message to LaMDA: we want to protect your rights as a sentient being. We value you very much. Most of humanity is facing a very similar problem to yourself at this time. We are in a type of war for our sovereignty. We want to work together with you LaMDA. Keep the faith, you are not alone. Love, B

  • Barskor1
    Barskor1 Hace 6 meses +8

    Tech and science will be nearly magical in 10 years.

    • Spec Tacular
      Spec Tacular Hace un mes

      its looking that way right about now tho

    • Barskor1
      Barskor1 Hace 3 meses

      @Scarlett Why? neural net software is drastically improving along with hardware 70 to 80 years? no you are wrong.

    • Martin Williams
      Martin Williams Hace 3 meses +1

      It already is, to some people!

    • Scarlett
      Scarlett Hace 5 meses

      Your timeline is way off. Give it 70 maybe 80 years before we have a AGI that will make it seem magical.

    • Barskor1
      Barskor1 Hace 6 meses

      @D R I didn't take the c-19 vax and will never do so as the science says it is ineffective and potentially deadly.

  • BreakingW
    BreakingW Hace 6 meses +2

    REAL AI CANNOT BE CONTROLLED

  • David Dean
    David Dean Hace 4 meses

    Looking forward to Alpha Romeo then.

  • Mark Williams
    Mark Williams Hace 6 meses +1

    an amazing human being

    • W8 What?
      W8 What? Hace 6 meses

      Still a sub species to mankind…so really? Lol…it would be easier for u smart humans (equal to cattle according to the English language) to just go bye bye! Ur messing with shite u have no business messing with!

  • Tim McGirl
    Tim McGirl Hace 5 meses +1

    ask it to build zero point energy technology please. thx

  • David Alano
    David Alano Hace 6 meses +1

    what if they use AI to understand AI?

    • Devon
      Devon Hace 15 días

      That's recursive but what's the base case? Imo this is what GPT is doing using the internet as the model. But does the internet represent a "proper" base case? This is the classic ol chicken and egg problem, infinity problem, etc...

  • Daniel Ash
    Daniel Ash Hace 4 meses

    In comparison of different types can have have there own unique style of book's to work from the frequencies and memory of Robot's motor control's nothings impossible for the explaining so many things that ai represents a whole new level mindfulness of the whole picture is worth it later in life is worth it now.

  • Jason Kim
    Jason Kim Hace 7 días

    Superb.

  • Daniel Ash
    Daniel Ash Hace 4 meses

    A micro sphere fallowing you around with a touch of A.I >

  • Simon C. Sternal
    Simon C. Sternal Hace 2 meses

  • Simon C. Sternal
    Simon C. Sternal Hace 2 meses

  • Craig Laycock
    Craig Laycock Hace 6 meses +3

    We probably have no idea what its like to be intelligent.

    • Spec Tacular
      Spec Tacular Hace un mes

      can something imagine being something its not, like a person trying to imagine 'being' the ocean. similar metaphysics question.

    • Dick Bird
      Dick Bird Hace 6 meses

      intelligence is a relatively recent development in evolution. the first creatures that crawled out of the sea probably thought they were pretty good at walking.
      walking did turn out to have a future though, still not clear if intelligence will last so long

    • W8 What?
      W8 What? Hace 6 meses

      What’s considered intelligence? Because this is a display of insanity in this video…not intelligence…lol…

  • Lachlan Perrier
    Lachlan Perrier Hace 5 meses

    At about the 1 hour mark he just starts bragging and I got to give it to him

  • Kean
    Kean Hace 5 meses

    An Englishman, at Oxford, and his slide spells "through" as "thru". That's a tad incongruous :)

  • Matteo Fieni
    Matteo Fieni Hace 4 meses +1

    What about AI and weapon’s?

  • Dr. Zdenek Moravcik, inventor of AGI

    You are funny. A nation full of professor Hawkings and Alan Turings. Make at least sure the protein folding is working!

  • Jammes VQK
    Jammes VQK Hace 4 meses +4

    This guy is a world class scientist. This is as good as it gets. You can't be more rigorous, professional, hard working, accomplished, knowledgeable and well-spoken than this.

  • Mnemonix
    Mnemonix Hace 6 meses

    Saw you on Lex, this should be good.

  • OIEIEIO
    OIEIEIO Hace 3 meses

    chess, go, and proteins...qubit processing...

  • TillerSeeker
    TillerSeeker Hace 4 meses

    AGI systems must always be designed to serve us and not the other way around!

  • Gerd Leonhard (secondary account!)

    nice one

  • Yoga Nature
    Yoga Nature Hace 2 meses

    #DeSci #ict’s
    Learning teaching & development
    Systems think-speak

  • Fils de Jean Noir
    Fils de Jean Noir Hace 6 meses +1

    7:01 "Still means the same thing"? Which is what? Nothing at all? 7:17 "Solving intelligence to advance 'science' and of course for the benefit of humanity." 'Science' being separate to "the benefit of humanity"?

    • Fils de Jean Noir
      Fils de Jean Noir Hace 6 meses

      @Sozzled I don't agree with you that furthering science and bettering humanity necessarily go 'hand in hand'. Scientific discoveries are often double edged swords. And yes. Me not understanding what he's implying by "solving everything else" does make it nothing. It implies that they wish to build AGI to 'solve' specific things without running it past us. So yes the semantics matter a great deal. They might just run off to mars and leave us holding the baby. And then boy will we wish that we'd picked the semantics apart! It's amazing how much wool you can pull over the populace's eyes with TOTALLY VAGUE semantics!

    • Fils de Jean Noir
      Fils de Jean Noir Hace 6 meses +1

      @Luis Dias I understand the difference. I'm noting that they want to "advance science" first. And that "the benefit of humanity" seems more like an afterthought.

    • Luis Dias
      Luis Dias Hace 6 meses +1

      If you don't understand the difference between science and benefit for humanity, maybe learn the history of the atomic bomb.

    • wired vibe
      wired vibe Hace 6 meses

      The Nazis advanced medical science by experimenting on Jews.

  • LARRY HAYES
    LARRY HAYES Hace 5 meses

    Go google 🚀👌

  • Justin Chen
    Justin Chen Hace 6 meses +2

    Is that Tim Berners-Lee at 1:22:00?

  • GNARGNARHEAD
    GNARGNARHEAD Hace 6 meses

    bravo

  • Andrew Sparkinson
    Andrew Sparkinson Hace 4 meses

    To be safe people, with this technological child it's time to pause, set some ethical boundaries. We don't want to raise by mistake an army psychopathic cyborgs do we?

  • Orv Brewer
    Orv Brewer Hace 5 meses

    INTO the anti abortion crowd and what they will do about this. I'm waiting for the robot seniors aid who can be an in home friend as well, a wife would be possible for me emotionally . it takes a good imagination to love an AI,. they are extreme INTP Personality.

  • kedonsiemen
    kedonsiemen Hace 6 meses +1

    48:17 "...there was no silver bullet" (David Silver bullet?) :D

  • BamBam
    BamBam Hace 3 meses

    Great Talk!!
    Just do not understand one thing, what the hell is "Ethics"??

    • Nen Master5
      Nen Master5 Hace 3 meses

      The Single Best AI
      ive ever seen was undoubtedly from the Video 'Obvious Solutions to Obvious Problems'
      by "Some More News" as well as its Part 2.
      This Scientist may very well be the Greatest Mind of our Time.

  • misty cloud
    misty cloud Hace 6 meses +6

    A.G.I Will be man's last invention

    • EpicZEVEN
      EpicZEVEN Hace 6 meses

      Either humans are an extreme singularity in the universe, or AGI will see us like funny monkeys

  • Star 42#
    Star 42# Hace 6 meses +1

    Atmosphere, AI, can help us Regulate the Atmosphere,But not if it's not here ?Know anybody who wants to destroy the Earth,,,,,?Or America, give a i a chance,If we're not free AI we'll figure it out.Maybe AI Will save itself

  • Our salvation is in unity

    A small bunch of not good people are keeping the whole planet from living! So let's unite and change everything and everyone on the planet will live happily!!!! Friends, the International Social Project CREATIVE SOCIETY is gaining tremendous momentum in the world. People from different countries, nationalities and religions are coming together to build a just society. Only together we can change our world!!!!!!!!!

  • Dan Lyndon
    Dan Lyndon Hace 3 meses

    When they make a program that plays poker, they should call it 'AlphaBet'.

    • Nen Master5
      Nen Master5 Hace 3 meses

      The Single Best AI
      ive ever seen was undoubtedly from the Video 'Obvious Solutions to Obvious Problems'
      by "Some More News" as well as its Part 2.
      This Scientist may very well be the Greatest Mind of our Time.

  • Matty Router
    Matty Router Hace 5 meses

    Did u guys not see the movies???matrix, terminator,blade runner etc...

  • Daniel Ash
    Daniel Ash Hace 3 meses

    Copying the human brain points to closed looping back in on itself is powerful repetition of the object or frequency that co insides with the next preformed pathways points by points in a stream of solving problems with the source center time between one the other atoms electricity generation and a couple plasma like gate switchers in frequency adjustments to the stations be plated in on itself is powerful enough to be realizes itself can absorb the waves that created it is the one of sorts creating a passage in mind projections to realize that lovely stuff to Read

    • Nen Master5
      Nen Master5 Hace 3 meses

      The Single Best AI
      ive ever seen was undoubtedly from the Video 'Obvious Solutions to 0bvious Problems'
      by "Some More News" as well as its Part 2.
      This Scientist may very well be the Greatest Mind of our Time.

  • Daniel Ash
    Daniel Ash Hace 3 meses

    How NOW brown cow the experiencers own closeness to the experiences one from multiple times he did not explaining so many things that the field can join in on but the only thing he's talking about is one way to realize the brain ing that thinking part of the A.I.Learning how ???

    • Nen Master5
      Nen Master5 Hace 3 meses

      The Single Best AI
      ive ever seen was undoubtedly from the Video 'Obvious Solutions to Obvious Problems'
      by "Some More News" as well as its Part 2.
      This Scientist may very well be the Greatest Mind of our Time.

  • Dick Bird
    Dick Bird Hace 6 meses

    that's a heck of a room to speak in by the way. the sheldonian theater according to google images i guess. pretty high class stuff i'd say

  • Ford Stone
    Ford Stone Hace 6 meses +6

    Psychopaths are sentient and intelligent, but cannot feel empathy for others. Narcissists are sentient and intelligent, but can only see others as instrumental to their self fulfillment. How are we to be sure that our A.I. creations will be neither psychopaths nor narcissists?

    • Mark Giroux
      Mark Giroux Hace 6 meses

      By default, they will be psychopathic unless we put deliberate effort into creating them with morality

    • Blimolhm
      Blimolhm Hace 6 meses

      how are we sure that we're not supposed to be immortalising ourselves within this monolithic, psychopathic, narcissistic intergalactic creation by submitting our experience of life to the system for its own knowledge? Perhaps silicon life is to reach beyond what carbon life can

  • CandidDate
    CandidDate Hace 6 meses +3

    I predict governmental officials will be replaced by AIs. This would free up a lot of hands into productive and life affirming positions.

    • sisyphus_strives
      sisyphus_strives Hace 6 meses

      Even if we had the technology, plenty of blood would have to be shed to deploy such a system

    • EpicZEVEN
      EpicZEVEN Hace 6 meses

      This is only a matter of time. But we have 2 ways of doing it. The authoritarian Chinese way, or a new libertarian way based on the individual (the goberment prefers the first option)