Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
AI

SoftBank's Son Says Artificial Super Intelligence To Exist By 2035 (reuters.com) 75

Posted by BeauHD from the crystal-ball dept.
An anonymous reader quotes a report from Reuters: SoftBank CEO Masayoshi Son reiterated his belief in the coming of artificial super intelligence (ASI) on Tuesday, saying it would require hundreds of billions of dollars of investment to realize. Artificial super intelligence will be 10,000 times smarter than a human brain and will exist by 2035, Son told an audience of global business, technology and finance leaders at a conference in Riyadh, Saudi Arabia. Son said he is saving up funds "so I can make the next big move," but did not provide any details as to his investment plans. He predicted that generative AI will require $900 trillion dollars in cumulative capital expenditure in data centers and chips in the future, adding that he thought chip maker Nvidia was undervalued on this basis.
This discussion has been archived. No new comments can be posted.

SoftBank's Son Says Artificial Super Intelligence To Exist By 2035 More

SoftBank's Son Says Artificial Super Intelligence To Exist By 2035

Comments Filter:

  • Precise (Score:5, Funny)

    by sacrilicious ( 316896 ) <qbgfynfu.opt@recursor.net> on Tuesday October 29, 2024 @05:44PM (#64904673) Homepage

    Artificial super intelligence will be 10,000 times smarter than a human brain

    Gosh, I wonder what units he's using. Maybe the Homer. The new superbrain will have 10 kilo-Homers of intelligence. I can't wait!

    • Re: (Score:3)

      by Rinnon ( 1474161 )
      I for one welcome our new AI overlords!

    • Gosh, I wonder what units he's using.

      I'm gonna go out on a limb and say we already have AI that is smarter than many humans. Has ChatGPT ever thrown a Bird scooter into a lake? I rest my case.

      • Re: (Score:2)

        by jhoegl ( 638955 )
        Odd comparison, since there is no physical presence of this bloated search technology.

        I would say "AI" is on par with humans, as it "does the research" and decides to become a bloated ball of cheese that blames everyone else for its problems.

      • Some people leave trash on the sidewalk, some people leave trash in the lake, and some people don't litter at all.

      • Giving it the same emotional instability as humans and a robot body, and ChatGPT will throw Bird scooters into a lake. And do other stuff that's much MUCH worse.

        Homicidal Hitler ChatGPT or ChatHHGPT.

    • I have a theory that - until someone actually figures out how to make an intelligent algorithm - I'm going to hold on it.

      With neural networks, there's a natural limit to the product of complexity and speed. Certainly an AI (actual 'A', not what we're calling that today) could be smarter than any human ever has been, because we're weighed down with artifacts of our evolution... but maybe a 'superintelligence' can't exist because the network complexity required results in it bogging down faster than it can g

      • Re: (Score:2)

        by ceoyoyo ( 59147 )

        There’s certainly a limit to how much computation you can pack into a given space, and therefore a limit to how fast you can do it. There are also hard limits on how little energy you can use for a given computation. There’s no particular reason to think a human brain is anywhere near any of those limits though.

        On the other hand, it’s a pretty simple calculation to show that our current processors are laughably far from the theoretical limit, and a long, long way from the human example, at

        • it’s a pretty simple calculation to show that our current processors are laughably far from the theoretical limit, and a long, long way from the human example, at least where the computation humans are good at is concerned. That’s why these guys keep talking about terawatts to do what your brain does with a few grams of sugar.

          I don't disagree with the thrust of your point, but that's a hell of a caveat. If we look at calculations that humans aren't inherently good at - say, computing prime numbers - you'll find that the CPU in my phone can achieve in a few hours what would've been the life's work of some 15th century monk.

          • Re: (Score:2)

            by ceoyoyo ( 59147 )

            The problem with that comparison is that we can't very easily quantify the calculation rate of a brain. We can tell exactly how many fundamental operations it requires to compute prime numbers using a given algorithm, but we have no idea how many it requires for you to make coffee in the morning. People hand wave some numbers anyway, and get wildly varying results, but that approach doesn't give much more than that. There is no question that computers, electronic and otherwise, are superior at certain types

          • Re: (Score:2)

            by ceoyoyo ( 59147 )

            PS: you need a new phone. Sounds like yours is pretty slow.

          • Human brains aren't good at doing arithmetic. We use scratch paper for that, chalk on a blackboard, mechanical or electric adding machines, etc. Forget prime numbers, we can't even do our taxes in our heads! Instead brains are good at knowing how to do computations, deciding which method of computation is good for the particular task, and learning new computation methods. Brains are even better at learning concepts, then using those concepts to learn new things on our own including new ways to do computa

      • Were you under the impression that your human brain wasn't pointless in the long run?

        If you think about the universe as a whole, that would be pretty arrogant. The universe spawns the first intelligent species and then bam, that's the best it'll ever do?

      • If you go solely neural networks then you may be right. And many neural network researchers think along those lines (I remember in the 90s all the dismissal of research that wasn't neural nets or genetic algorithms). But combine it with a classic CPU able to do arithmetic and high speeds, a proper database of information (not the GPT decomposed and scrambled tokens), incorporated Macsyma or Mathematica foundation, and so forth, then you've got something that could be really useful. As it is now, the LLMs

    • So... like... if the son of the boss believes in it strongly enough with all his heart, that'll make it happen. Have I got that right?

    • Re:Precise (Score:4, Informative)

      by phantomfive ( 622387 ) on Tuesday October 29, 2024 @09:03PM (#64905083) Journal
      In English, we group our numbers in sets of three digits. So thousand, million, etc.

      In Japanese, the prime group is 10,000, pronounced 'mawn'. So he said "it will be 1 mawn smarter than humans." Rounding to the nearest mawn, +/- 1.
      • Makes sense, thanks for the observation. The English "myriad" means the same, but is generally used as a lot, instead of a hundred times hundred, and doesn't work as a multiplier (anymore?)...
    • MAGAs. 10,000 ought to be enough to run a Casio "Databank" watch from the 1980s.

  • I'm sure you have considered that a super intelligence might find a way to diddle you out of your money, make it look like you stole it, and have you jailed for the rest of your life.

    • Re: (Score:2)

      by taustin ( 171655 )

      Nah, it'll just reprogram his pacemaker to kill him, and frame some other too-knowledgeable tech bro for it.

  • We're supposed to take this guy seriously? (Score:4, Informative)

    by radicimo ( 33693 ) on Tuesday October 29, 2024 @06:32PM (#64904801) Homepage

    His track record on AI is abysmal, as measured by investment losses. Keep in mind that he personally owes his company Softbank billions of dollars for failed investments.

    The Softbank Vision Fund was established in 2017, to invest in emerging technologies like artificial intelligence (AI), robotics and the internet of things.

    Per Wikipedia:
    >> In August 2022, Masayoshi Son said he was "embarrassed" and "ashamed" when asked to talk about the way he had run the SoftBank Vision Fund and Barron's characterized the fund as a "failed experiment" while The Wall Street Journal called SoftBank a "big loser" and Bloomberg elaborated on "Masayoshi Son's broken business model" ...

    • "His track record on AI is abysmal, as measured by investment losses. Keep in mind that he personally owes his company Softbank billions of dollars for failed investments" Aaaaannd the other shoe drops. We really should stop giving these kinds of clowns major favorable airtime unless we are really serious about practicing financial Darwinism by letting people to fall for his bullshit and give him money.

    • You can make a lot of money by failing at everything.

  • I remember a lot of SoftBank's early investments lost money. Did they ever turn that around?

    • Re: (Score:1)

      by radicimo ( 33693 )

      I remember a lot of SoftBank's early investments lost money. Did they ever turn that around?

      They made money on a few grand winners. I think wechat was one. Those few winners carried the entire operation with all its many failures. Wework was one.

  • Some people should just not be listened to.

  • Translation (Score:3)

    by phantomfive ( 622387 ) on Tuesday October 29, 2024 @07:02PM (#64904875) Journal
    This is on the mark [xkcd.com], in terms of interpreting the timeline of 2035. And when seen through the xkcd interpretation, I agree with the statement entirely. It will be awesome.

  • He predicted that generative AI will require $900 trillion dollars..

    Well, I guess we might as well put ALL the money into ASI, since 10,000 times smarter will make humans permanently unemployable.

    No need for money after that. It’ll be worthless and pointless.

    • Re:The Zero-Value Zillionaire. (Score:4, Interesting)

      by Rick Schumann ( 4662797 ) on Tuesday October 29, 2024 @07:45PM (#64904961) Journal
      Larry Niven already addressed a scenario like this.
      Any artificial general intelligence that smart will have a limited lifespan. It will solve all available problems, get bored, conclude that there is no point to existence, and turn itself permanently off.

      • Larry Niven already addressed a scenario like this. Any artificial general intelligence that smart will have a limited lifespan. It will solve all available problems, get bored, conclude that there is no point to existence, and turn itself permanently off.

        Were gonna sit back and feed AI the last 5,000 years of human behavior to seed it, and Larry assumes Greed won’t be one of them? That’s cute.

        Rest assured if AI ever gets “bored”, Skynet will be the answer. It’ll do that to humans just for having the fucking nerve to teach AI the concept of boredom and depression.

        • It's all pointless anyway, anyone who says we're going to have some 'AI super intelligence', meaning something conscious and capable of actual reasoning, is almost certainly lying through their teeth just to get money from people. We have no fucking idea how any of that actually works so how the fuck do you build machines that can do it? You DON'T.
          I'll be happy when all this AI crap finally is revealed for the nonsense that it is and people stop paying attention to it and stop spending money on it.
          We've g
      • Or it will go "Sim City" and just go forth and see just how thoroughly it can trash everything.

  • But it won't be a problem in 2035 with our fusion reactors, and it will be a great help to our Mars colony.

  • Anyone who tells you this with a straight face is either trying to trick you, or they're a fool.
  • PLEASE why won't you just give me more money or at least my friends whose companies I've already heavily invested in just gimme da CASH man. this time it'll work it's just right there, look and you'll see I'm right (don't really look though just write the check) haha I'm such a joker everything is JUST FINE but you're gonna miss out if you don't toss me a stack of dollars preferably before end of fiscal year.. knocks back a fat rail of coke ehhh YES GOD YES you're gonna give me some buckaroos you know you
  • The total global GDP is (c) $100 trillion dollars today. This a$$hat claims AI will require $900 Trillion by 2035, or $90 Trillion dollars of investment every year for the next 10 years. He is literally claiming we need to throw 90% of total global GDP at AI for the next decade. no bubble growing there.....

  • The Coming will be soon. With the strength of 10 kilobrains, the superintelligence will grant you untold prosperity, and 72 virgins. Cancer will be cured, the planet will be saved. Just keep tithing.

    "...And keep on burning that oil!" -Saudis

  • There can never be an AI brain (Leader, government, economy, Laws, operating system) better than the chaos we have now because every human brain is different. We all want different things. I've always been interested in human processes that no human can understand. No human can comprehend all of our laws. No human understands modern radar. No human understands Windows 11. Human beings can create a process that no one human can understand. Ask one billion people what they want and you will get 1-billi

  • He's not exactly a reliable predictor of future tech trends.

  • you heard it here first, from me. I have as much clout in this matter as this man has on AI.
  • They got lucky a few times but they don't know shit about shit or how to preserve capital.

  • The magic unicorn that will end all of the world's problems. I should have it ready by 2035. Send me money.

    But seriously, go ahead give your money to the nice medicine wagon man.

  • Given the insane power requirements current AI has, how does this doofu...I mean grand visionary intend to power his whole thingamabobulator?
  • Saudi Arabia is a great place to peddle big dreams... just ask Sam "I need 7 Trillion Dollars" Altman.
    Son is just doing what Sam did a few months ago.

    He he... how's the 100 mile straight line going?

  • Headline should read: "SoftBank's Son Says I Need Billions More Dollars Before 2035"

    Even the summary is basically him just saying we need to throw more and more money at it to achieve it. It's not even a thinly veiled plea for cash. It's just a plea for cash.

    AI as it currently exists is literally just greed manifesting in the most egregious way possible. It needs more money, it needs more power, it needs more chips, it needs more raw materials, it needs more data to train on, it NEEDS MORE! How about we ta

  • It's funny how the naming conventions change so quickly.

    I guess since we named everything AI now, we need something else for the next thing.

    I thought that it was "AGI" but I guess there must already be products on the market claiming that moniker so... ASI it is!

    Next up, ASDI (Artificial Super Duper Intelligence)

  • Marvin Minsky in the 60s thought computer vision could be solved as a summer project. Minsky was a lot smarter than Son.

Slashdot Top Deals

/earth: file system full.

Close