Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
Facebook

Mark Zuckerberg Plans To Shake Up Meta's AI Efforts, Again 34

Posted by msmash from the behemoth-sized-pivot dept.
Meta announced today that it is splitting its Meta Superintelligence Labs into four divisions focused on AI research, superintelligence development, products, and infrastructure. The reorganization accompanies potential downsizing of the AI division's thousands of employees and executive departures, according to New York Times.

Vice President of Generative AI Loredana Crisan is expected to announce her departure Tuesday. The company is exploring third-party AI models for its products rather than relying solely on internal technology. Chief AI Officer Alexandr Wang's team has abandoned Meta's previous frontier model Behemoth and is developing a new model from scratch, the report added.

Mark Zuckerberg Plans To Shake Up Meta's AI Efforts, Again More | Reply

Mark Zuckerberg Plans To Shake Up Meta's AI Efforts, Again

Comments Filter:

  • Apologize first (Score:4, Funny)

    by OrangeTide ( 124937 ) on Tuesday August 19, 2025 @02:03PM (#65600348) Homepage Journal

    Apologize for the Metaverse first.

  • reorganize reorganize again.

    • In fairness... (Score:4, Insightful)

      by Brain-Fu ( 1274756 ) on Tuesday August 19, 2025 @03:27PM (#65600618) Homepage Journal

      A lot of smart people bought into the AI hype. If Meta didn't try to position themselves as a leader in the field, they would have been pummeled in the stock market. Zuckerberg may be rich, but he is still beholden to the will of the shareholders. If they are caught up in a bubble, then so is he.

      The reality is kicking in. We just had an article right here on Slashdot today about how 95% of attempts at using AI in the corporate workplace were failing. I think the bubble is starting to pop. So, Meta is re-positioning for that event, now, since there really isn't much else they can do.

      Realizing the dream of AGI is going to require a fundamentally different breakthrough. This supercharged pattern-matching-and-prediction tech we have now is pretty cool, but it just can't get where we want it to be, and no scaling-up of training efforts will fix that. Producing machines that can visualize and reason will require a different approach, and the path to get there is far from clear. Maybe some of the ridiculously-paid AI specialists will come up with something, and maybe they won't. There is still enough belief in this possibility, at the moment, to try, of course. So the big players are still trying. But the goal posts no longer appear nearly as close as they once did.

      • Re: (Score:2)

        by HiThere ( 15173 )

        I don't think that anything fundamentally different is required, but it may well require a robot body with lots of seinsors, (This should be doable in simulation.)
        And it's going to require continual learning, not session based learning.

        Lots of "short cuts" that were taken to make things cheaper will need to be undone. The basic training has to be on real facts, not on whatever text you find on the internet. And those "real facts" need to map onto the data from the sensors. Don't feed your AI on lies and

        • Who decides what are the facts?

          • Re: (Score:2)

            by spitzak ( 4019 )

            The GP is suggesting that the machine has physical sensors and compares any information it gets with what it can see/detect in the real world, and reject anything that does not agree as being a lie. For instance you would be unable to convince it that objects fall away from the Earth. And it could possibly figure out detailed tests for really obscure things since it is some kind of super-powerful AI. Furthermore it could decide that information from a source that has only provided proven-true information be

            • That might work for physical facts, but when we get into matters of the mind, government, economics etc. where the answers to questions are a matter of opinion or open to doubt, the AI would be out of luck. It would also presumably be just as vulnerable to scams and lies. Most of the "bad" behaviors are a result of bad data. I do not see how the system you are suggesting would protect the AI from this

    • Speed things up: break things and then move fast.

  • Err, the metaverse, err, AI, no, wait, let's do AI again.

  • Chief AI Officer Alexandr Wang

    CAIO? more like ciao, amirite?

    Seriously though, how long before the trap door button gets pushed for this guy?

  • Zuckerberg is an incompetent boob (Score:3)

    by rsilvergun ( 571051 ) on Tuesday August 19, 2025 @02:58PM (#65600542)
    Everything he has tried to do since Facebook has failed miserably with that virtual reality nonsense being the cherry on top. He can fail as much as he wants though without any punishment because once you are admitted to the club you don't get kicked out as long as you don't screw with fellow club members.

    Anyone else who was as bad at their job as Zuckerberg would have been fired ages ago.

    It won't matter though because Facebook is one of the major platform holders and because of that they are going to eventually have a commanding lead with llms because they're going to be able to keep feeding training data into their llms . Meanwhile they're competitors are going to have to keep scraping websites which is getting increasingly difficult between antibot technologies and AI slop
    • That Hot or Not app he made was pretty cool.
      • He didn't come up with the concept though it came from a website. Honestly I wasn't even aware of zuckerberg's companies have put out a Hot or not app. The original website from 2000 looks to be owned by the same company that owns Ashley Madison

    • Instagram is the world's most used image platform.
      Threads have 100s of millions of MAUs.
      VR headsets have on a single product sold more units than some popular consoles. - Yeah it's a money sink, but it has bought about an entire industry that is still making content and hardware to this day.

      Anyone else who was as bad at their job as Zuckerberg would have been fired ages ago.

      I just looked, Facebook made record profit and revenue in Q4 2024 (I don't have 2025 figures), the previous record was Q3 2024, the record before that was Q4 2023 with it being steady in Q1/Q2 2024, the previous record w

  • Some people rail on Zuck.

    But he did Fuck the Winklevii losers.

    For that I say he's not all bad.

  • It's amazing how these super-rich dicks keep fucking up yet stay rich. I guess they built up a fat margin of error, and/or mesmerize investors by being big and sounding bold.

    • It's the lack of shame regarding doing fraud, willing deception, and just being a dick that is a prerequisite to being an exorbitantly rich asshole.
  • “As the personalization loop kicks in and the AI starts to get to know you better and better, that will just be really compelling. It’ll help you achieve your goals, create what you want to see in the world, be a better friend to those you care about, and grow to become the person you aspire to be.”

    The myopathy of the vision. Reminds me of billg and his road to nowhere.

  • " Chief AI Officer Alexandr Wang's team has abandoned Meta's previous frontier model Behemoth and is developing a new model from scratch"

    All the AI that Meta had done previously is third rate and therefore worthless. Zuck realized that and poached the best talent from the competition (at enormous cost but he's swimming in money). The new folks have reviewed that legacy code and wisely decided to start from scratch. I think it will take a while.

  • That's makes sense.......see 4 times the failures coming !!!!
  • Pop already, goddamn you

Slashdot Top Deals

This is a good time to punt work.

Close