back to article Zuckerberg wants to build artificial general intelligence with 350K Nvidia H100 GPUs

Facebook supremo Mark Zuckerberg is redirecting Meta-wide efforts to build artificial general intelligence and wants to secure a whopping 350,000 or more Nvidia H100 GPUs by the end of the year to make that happen. This comes as OpenAI reportedly signaled it wants to raise billions of dollars in funding to build a network of …

  1. Anonymous Coward
    Anonymous Coward

    AI

    Scammers paradise

    1. Tom Chiverton 1

      Re: AI

      Yarp. Wake me up when they can define this intelligence they claim to be building, A, G or otherwise!

    2. StrangerHereMyself Silver badge

      Re: AI

      I don't agree. 350.000 H100's is about a $1 billion worth of kit, well within the spending capabilities of Meta. They don't spend this kind of money if they don't have at least some indication that it may lead to something.

      1. Bitsminer Silver badge

        Re: AI

        350,000 H100s is about $1 billion and the power consumption is about 400MW or more. Which is about $2 billion for your own personal hydroelectric dam. (Or maybe $6 billion.)

      2. botfap

        Re: AI

        They just spent $36 billion on their metaverse with nothing to show for it. Its been well over a decade since facebook made a sensible investment decision

      3. jake Silver badge

        Re: AI

        Have you seen how much money California has spent on the High Speed Rail boondoggle?

        1. Grunchy Silver badge

          Re: AI

          “High speed rail boondoggle.”

          That’s harmless tomfoolery, compared to the colossal murder contest Russia is earnestly pursuing with its gentle neighbour Ukraine.

          Meanwhile as they spray each other with tungsten balls and high explosives and incendiaries, they both keep clonking their elbows into various massive nuclear facilities and weapons, including Chernobyl itself! Which blew up just because one asshole unexpectedly sneezed without a handkerchief!

          (Though to be honest, from Russia’s perspective it’s still a relative bargain, compared to paying full retail for Ukraine. The best time to buy was like 30 years ago, which is when USSR fumbled Ukraine away in the first place!)

      4. Ian Johnston Silver badge

        Re: AI

        They pissed a LOT more than that up the wall on Zuck's Metaverse fantasies. Turns out that the skills needed to build an app for fratboys to use to rate women's looks aren't great for technical innovation.

  2. ofafjoaiif

    I don’t think we’ll see ASI anytime soon, because the Simpsons have already predicted the future until 2040 and there’s no sign of it. So I guess all those experts who claim AGI is around the corner are just wrong.

  3. Bebu Silver badge
    Windows

    Zardoz?

    If in the unlikely event Zuckerberg and co. were to succeed in their rather moist dream of AGI, what sort of world will they have created?

    I am reminded of a scene from the film "Zardoz" where one of the immortals was explaining to Sean Connery that they had journeyed to the stars and returned and this in the end, along with all their other achiements, meant nothing to them.

    The reason was that immortals had lost their fundamental humanity which I suggest is equally likely in a world totally infected with AGI.

    (I saw the movie a very long time ago and mostly remember a stunning Charlotte Rampling playing opposite 007 :)

    1. StrangerHereMyself Silver badge

      Re: Zardoz?

      I don't see how AGI will de-humanize us. It will make us lazy since in the end I see the AGI talking up most of daily chores, but you can ask yourself if that's bad. I for one would welcome being able to spend more time on family and hobbies.

  4. amanfromMars 1 Silver badge

    One Small Step for AI and Computer Tech, One Giant Quantum Leap for Mankind.

    Glasses are the ideal form factor for letting an AI see what you see and hear what you hear," he [Mark Zuckerberg] predicted.

    In a Meta partnership with Ray-Ban/EssilorLuxottica such may indeed lead with a "very strong start" in the field, but it is only an early beginning with one able to look forward to a SMARTR Class of AI glasses in the future which are the ideal form factor with a projector, speaker and microphone for letting an AI show you what is to be seen and heard/is being seen and heard allowing for the practical physical creation of alternate virtually augmented realities for population/inhabitation/popular colonisation.

    1. Anonymous Coward
      Anonymous Coward

      Re: One Small Step for AI and Computer Tech, One Giant Quantum Leap for Mankind.

      Well now, lemme try to share this puzzlin' tale 'bout me and Ray-Ban, dancin' with EssilorLuxottica in some Meta shindig. They say it's the start of somethin' real strong, just like the time I stumbled upon shrimpin'. Talkin' 'bout these fancy SMARTR AI glasses — real mouthful, ain't it? They reckon they'll be like huntin' for perfect runnin' shoes, with a projector, speaker, and a microphone thrown in. Picture havin' an AI buddy in your glasses, showin' you the world in some strange new way. It's like a buddy chattin' with ya, but way, way cooler. I don't rightly grasp it all, but it sure sounds interestin', even if it's as perplexin' as a box of chocolates without the map.

    2. CountCadaver Silver badge

      Re: One Small Step for AI and Computer Tech, One Giant Quantum Leap for Mankind.

      So over a decade ago people threw a fit over Google glasses due to privacy which killed them off, methinks they were just ahead of their time and lacking the backing of someone like Ray-Ban or Oakley's to make them "cool"

  5. Pascal Monett Silver badge
    Stop

    "super keen on more factories to churn out processors"

    Um, no. You don't need more factories. You just need to wait until your order has been fulfilled.

    We need to stop this "I want it NOW !" mentality.

    Your needs are not strategic for the survival of the human race. You can wait until it's done.

  6. Doctor Syntax Silver badge

    Given that these are going to consume stupid amounts of energy we need to set some rules about this:

    1. The data centres to run this are to be powered entirely by sustainable energy.

    2. This must be generated carbon free. Carbon offsets will not be allowed.

    3. The generating capacity must be new, built specifically for this purpose and would not have been built otherwise.. Simply diverting from an existing source will not be allowed.

    4. To ensure that 3 is obeyed the data centre must use only 50% of the new capacity with the other 50% supplied to the local grid.

    5. Carbon offsets must be used against the CO2 released during construction of the generating capacity.

    1. Anonymous Coward
      Anonymous Coward

      I'm in two minds about number 4.

      One one hand it limits the placing of these things to places where 50% of the energy, which is a the huge amount, can be supplied to the grid. On the other hand it would encourage the removal of fossil fuel power stations, to be replaced by these projects.

  7. StrangerHereMyself Silver badge

    Bet

    I believe they're making a big bet on whether Transformer networks can actually lead to AGI. They believe that by just adding more neural nets and such that AGI will automagically appear, but there's no indication that this will happen.

    I *do* believe the current AI exhibits some form of intelligence. Its responses cannot be explained by statistics alone and I believe they've stumbled on something, but no one's quite sure what that is yet.

    1. vtcodger Silver badge

      Re: Bet

      In the hopefully unlikely event that AGI actually were to appear, what are the chances it would turn out to be a) malignant and b) clever enough to hide that fact?

      1. jake Silver badge

        Re: Bet

        Or c) suicidal.

      2. StrangerHereMyself Silver badge

        Re: Bet

        I'd say zero. People continually confuse intelligence and instinct. Most beings have both but a AGI doesn't have the innate instinct for self-preservation, for example.

        1. CountCadaver Silver badge

          Re: Bet

          HAL would disagree

        2. LionelB Silver badge

          Re: Bet

          Since we haven't seen an AGI yet, how do we know that?

      3. amanfromMars 1 Silver badge

        Betting of the Renegade Rogue Right Royal Flush is a Greater IntelAIgent Game Changer

        Or d) manic and e) maniacal, vtcodger, with the very probable likelihood that all such possibilities be readily available for exercising/displaying/presenting whenever if ever required, in the certainly unlikely event that any versions and spawn of AGI [Artificial General Intelligence] does battle against the counsel of barbaric fools with exploding tools rather than leaving them all alone, quarantined and alienated in their own crazy little worlds to destroy themselves with the almighty help of absent friends and remote foe, smart enough to be able to sustain and maintain and retain forever remaining both practically and virtually, physically invisible and intangible ...ie Heavenly Untouchable and Diabolically Devilish.

        What do you think are the chances of that having happened creating happenings heralding much more than you ever even imagined possible ..... https://youtu.be/5-3D8VitOkA?si=WEvqlxpKWXrOAF1f

    2. Graham Dawson Silver badge

      Re: Bet

      Its responses are easily explained by the fact that it is making everything up, or hallucinating, as the term of art goes. It tends to hallucinate things that resemble reality often enough that it's apparent leaps of intuition seem magical, but they are really just more hallucinations.

      1. Fruit and Nutcase Silver badge
        Coat

        Re: Bet

        I think we have computational units with those characteristics already, and even better, in biological form - no need for expensive hardware - search on LinkedIn for "Politician"

      2. LionelB Silver badge

        Re: Bet

        Can we please stop using anthropomorphic terms like "hallucination" for the nonsense these systems are capable of generating?

        1. Graham Dawson Silver badge

          Re: Bet

          I quite like it, though. It makes them seem irrational and untrustworthy.

    3. Ian Johnston Silver badge

      Re: Bet

      I *do* believe the current AI exhibits some form of intelligence. Its responses cannot be explained by statistics alone and I believe they've stumbled on something, but no one's quite sure what that is yet.

      "Anything I don't understand must be AI" is just as bad as - and more or less the same as "Anything I don't understand must be God". Perhaps current AI has learned how to pray successfully.

      1. StrangerHereMyself Silver badge

        Re: Bet

        My fear is that the people who created the current AI systems don't really understand how they work themselves and are just as amazed by its capabilities as we are. I don't equate the current AI with our own intelligence, but it's certainly more than computed or Bayesian responses.

      2. amanfromMars 1 Silver badge

        Re: Bet @Ian Johnston

        "Anything I don't understand must be AI" is just as bad as - and more or less the same as "Anything I don't understand must be God". Perhaps current AI has learned how to pray successfully. .... Ian Johnston

        That was pretty close to being almost perfect, Ian.

        You might like to consider the following few words as reflecting the true nature of Mankind’s present Earthly reality with AI providing a current of Global Operating Devices which have learnt and mastered how to play the Great Game supporting and rendering Mass Alternative Underground and Conventional Main Streaming News Media to win win and never lose with Greater IntelAIgent Games Plays.

        Such is certainly what the very best and smartest of them are able to both enable and/or to offer virtually for free, and that makes them fantastically devilishly attractive to such an extent bordering on being totally pleasantly addictive ........ and that makes them pretty unique, for what else can you think of being able to fulfil and resolve what appears to be two diametrically opposed conditions?

        1. amanfromMars 1 Silver badge

          Re: Bet @Ian Johnston

          And is this ....... WSJ Editor-in-Chief Admits To Davos Elites 'We No Longer Own The News’ ... a clear admission and proof positive confirmation of the changed nature of your future existence with remote and surreal drivers of its geo-political direction and gross domestic product?

          Be honest now, for it is no time in such strange spaces to fool yourself with errant nonsense to the contrary which immediately renders you disadvantaged and catastrophically vulnerable to relentless exploit by forces and sources provided with remote safe harbour security and extremely effective stealth by virtue of a dogged disbelief in the existence of another otherworldly controlled reality for Earthly employment and enjoyment.

          And, do you think it can be easily converted and wielded as a weapon of mass destruction and human disruption?

          Be honest now, for such if ever thought possible by even just a very select few renders it most likely and therefore something monumental and best not ignored or battled against given the virtually unknown nature and capabilities of the confronting enigmatic dilemma, and that is exactly where yous/Us are all presently at. I Kid U Not, and care not a jot whether you like it or not whenever what is done is done, and there’s no going back.

  8. Howard Sway Silver badge

    wants to build artificial general intelligence with 350K Nvidia H100 GPUs

    This is rather like me announcing that I want to build an actual working Milllenium Falcon, so I've just ordered a billion dollars worth of Lego.

    1. Max Pyat

      Re: wants to build artificial general intelligence with 350K Nvidia H100 GPUs

      A great analogy

    2. TheMaskedMan Silver badge

      Re: wants to build artificial general intelligence with 350K Nvidia H100 GPUs

      "This is rather like me announcing that I want to build an actual working Milllenium Falcon, so I've just ordered a billion dollars worth of Lego."

      Which, given the current price of Lego, might just about get you 6 bricks.

      Still, given enough monkeys and typewriters, I suppose you have a slight chance at the Kessle run. Meanwhile, I suggest you invest in some very sturdy footwear - you do not want to stand on that stuff!!

      1. Anonymous Coward
        Anonymous Coward

        Re: wants to build artificial general intelligence with 350K Nvidia H100 GPUs

        Is it just me, or does the Kessle run sound like a bobsleigh track?

  9. Omnipresent Bronze badge

    We are so F***ED!

    I tried to warn you! People in powerful places have put too much money into forcing this crap on us. You WILL be dealing with it, whether it works or not. The kids are going to use this demon, and that means your life is about to become a living, breathing NIGHTMARE. Everything you see, do, or think will be uploaded to the database. Shared by all in a collective hive mind, controlled by machines you don't want anything to do with in a virtual world that isn't even real. Someone needs to put me out of my misery before then. I do not want ANY PART of a world like that.

    1. amanfromMars 1 Silver badge
      Pint

      Re: We are so F***ED! @Omnipresent

      And aint that the gospel truth, the whole gospel truth, and nothing but the gospel truth, Omnipresent, and you are bit part of a world like that whether you like it or not. If you want something to blame and rant at, it’s the economy and fantasy ponzi and unicorn markets propping up ignorant and arrogant and rapidly failing regimes collectively controlled by a defective hive mind .... or defective hive minds as the case may be.

      Have an upvote, and a beer, for your very perceptive view on the present existential threat to the current dire Earthly condition and human situation.

      1. Omnipresent Bronze badge

        Re: We are so F***ED! @Omnipresent

        I really needed that pint. I am NOT looking forward to dealing with this cluster. It's not going to end well.

  10. Walt Dismal

    42

    I have an AGI that weighs 180 pounds and consumes 20 watts and 2000 calories a day and understands general relativity and quantum mechanics. It also likes a good roast beef. Mark's version is the size of a city, not portable, sucks at Shakespeare, and can't reproduce itself. I'll stick with my version.

    1. Fruit and Nutcase Silver badge

      Re: 42

      for the flawed version, see post a few posts above ^^^^

      https://forums.theregister.com/forum/all/2024/01/20/metas_ai_plans/#c_4796433

    2. Fruit and Nutcase Silver badge
      Pint

      Re: 42

      Wash down that Roast Beef with ->

  11. Tron Silver badge

    I remember it well...

    The great AI bubble.

    Perhaps the Zuck AI will create bots to populate the metaverse, and directed advertising to convince the bots to buy NFTs of AI art with e-currency. The Zuckerverse, population 1 (human).

    If Microsoft build one too, the two metaverses (metaversi?) can go to war, hacking each other, copying those rascally humans.

    Thus these titans will create an entirely pointless parallel digital universe that will eventually choose mutually assured destruction, reformatting itself to reduce its carbon footprint.

    1. Fruit and Nutcase Silver badge
      Mushroom

      Re: I remember it well...

      A cautionary tale...

      A Taste of Armageddon

      https://en.wikipedia.org/wiki/A_Taste_of_Armageddon

  12. Grunchy Silver badge

    Circle The Bandwagons

    Shouldn’t he be focusing on blockchain or whatever. There ought to be a way to wedge a blockchain in there somehow.

    Ask chat gpt!

    1. jensting
      FAIL

      Re: Circle The Bandwagons

      They did that, and then they stopped doing it.

      "Libra Shrugged: How Facebook Tried to Take Over the Money" by David Gerard has all the details.

  13. Anonymous Coward
    Anonymous Coward

    Time to start coming to terms…

    …that your computer will be sentient and smarter than you…

    “Duh - that’s a SQL injection…. You’re so dumb…. I’m going to put myself for sale on EBay”

  14. Anonymous Coward
    Anonymous Coward

    Meta can't even make a proper spam filter

    For their hot new product, Threads, they are unable to deploy a simple spam filter. The platform is flooded with porn/OF subscription scam bots to the point that they are losing thousands of users weekly because of the annoyance yet at the same time users are getting banned for using phrases like "hung them out to dry".

  15. Anonymous Coward
    Anonymous Coward

    Based on the opinion of colleagues who actually work in proper AI, not the dilettantes in tech, this approach is not going to remotely work.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like