Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
Businesses Facebook AI

Meta Lays Off 600 From 'Bloated' AI Unit (cnbc.com) 71

Posted by BeauHD from the scaling-down dept.
Meta is laying off about 600 employees from its AI division as part of a restructuring to streamline operations and solidify Alexandr Wang's leadership over the company's AI strategy. "Workers across Meta's AI infrastructure units, Fundamental Artificial Intelligence Research unit (FAIR) and other product-related positions will be impacted," notes CNBC. "However, the cuts did not impact employees within TBD Labs, which includes many of the top-tier AI hires brought into the social media company this summer." From the report: Those employees, overseen by Wang, were spared by the layoffs, underscoring Meta CEO Mark Zuckerberg's bet on his expensive hires versus the legacy employees, the people said. Within Meta, the AI unit was considered to be bloated, with teams like FAIR and more product-oriented groups often vying for computing resources, the people said. When the company's new hires joined the company to create Superintelligence Labs, it inherited the oversized Meta AI unit, they said. The layoffs are an attempt by Meta to continue trim the department and further cement Wang's role in steering the company's AI strategy. Following the cuts, Meta's Superintelligence Labs' workforce now sits at just under 3,000, the people said.

Meta Lays Off 600 From 'Bloated' AI Unit More | Reply

Meta Lays Off 600 From 'Bloated' AI Unit

Comments Filter:

  • It shows monopolies have already formed (Score:5, Insightful)

    by rsilvergun ( 571051 ) on Wednesday October 22, 2025 @09:37PM (#65744542)
    In the past they wouldn't risk firing that many engineers because they would end up at potential competitors. Facebook has already figured out they can either run those competitors out of business or make back room deals with them if they are one of the big guys.

    What annoys me as we all grew up with TV and movies telling us how bad corporations are and if you learn how bad they really are in the real world it's actually worse than TVs and movies make it out to be.

    And despite all that it takes almost nothing for people to become pro corporate ass wipes. A few moral panics. Violent video games or trans girls in sports or dei woke or whatever the fuck they're calling political correctness today and blammo we all forget corporations are overpowered and evil as long as they protect us from the big scary whatever.

    It's almost as if deregulating our news media so that it can be owned by two people was a bad idea

    • Re:It shows monopolies have already formed (Score:5, Insightful)

      by Tony Isaac ( 1301187 ) on Wednesday October 22, 2025 @09:48PM (#65744558) Homepage

      Or maybe, just maybe, the "AI" teams *were* actually bloated.
      Maybe they were staffed with people who claimed they knew how to build AI products but couldn't actually deliver.
      Or maybe they figured out that the stuff they were promising to accomplish, was mostly vapor.
      Or maybe it was just politics in a big, bureaucratic organization.

      • Re: (Score:2, Insightful)

        by Anonymous Coward

        This is fun...

        Or maybe the “AI revolution” costs more than it earns.
        Or maybe Alexandr Wang wanted to look decisive.
        Or maybe research doesn’t impress Wall Street.
        Or maybe streamlining sounds better than panic.
        Or maybe the hype cycle peaked, and now it’s cleanup time.
        Or maybe they needed scapegoats for slow progress.
        Or maybe someone’s bonus depends on cutting headcount.
        Or maybe they’re betting on fewer people and more buzzwords.

      • I doubt the ML engineers were not good enough since they were hired by Meta and worked there for years. More plausible is that they overspent on VIP hires and then decided to cut from other ML roles. I am curious to see in 6 months if Meta can catch up. Anyway, the majority ML work nowadays is not training models, it's prompt and agent stuff, this type of work is complementary to model training.

        • The model improvement seems to be dying down. What seems to be being worked on is trying to find more data to stuff in the model, especially now that people are closing off access to training data, and trying ways to use the model in new applications. This is why there is such a push to harvest stuff from as many sources as possible.

          • Re: (Score:2)

            by HiThere ( 15173 )

            Perhaps that should be "The model improvement for chatbots seems to be dying down.", as I think that's the correct statement of what you mean. And that's probably correct. Once you get beyond chatbots, though, the model improvement is continuing.

            FWIW, I think chatbots have an intrinsically limited capability. However if you use chatbots as an interface to some other capability (i.e. robots, in various meanings of that term) then the limitation changes drastically.

      • Re: (Score:3)

        by AmiMoJo ( 196126 )

        Beware these big tech companies, especially when they are on a hiring frenzy. It's almost always a bubble, and you can only expect to be there for a few years maximum. Plan your exit strategy before the big layoffs come.

      • For decades high tier engineers from in demand Fields would be snapped up by the big players to prevent them from working for competitors regardless of whether or not those big companies had anything for them to do. Nvidia was particularly bad about that starving AMD of engineers.

        It's basically a very effective way to prevent competition when you have unlimited money from another line of business.

        These crazy hiring sprees were specifically designed to starve startups and other companies of potential

      • That's a lot of "or maybe"..
        I'm sure it's a multi prong decision. Hire everybody away from the competition due to deep pockets. See which ones out perform. Remove the ones who can be replaced with AI.

        • Agree, it's a lot of maybe. My point is, it's not easy to attribute such a decision to one root cause (monopoly).

      • Or maybe, just maybe, the "AI" teams *were* actually bloated.
        Maybe they were staffed with people who claimed they knew how to build AI products but couldn't actually deliver.
        Or maybe they figured out that the stuff they were promising to accomplish, was mostly vapor.
        Or maybe it was just politics in a big, bureaucratic organization.

        Or maybe AI research is really hard, and Facebook is learning that simply hiring a bunch of people and throwing money at them isn't enough. AI is still in the research stage and not the product stage, so it's hard to have deadlines and product roadmaps. Nonetheless, "visionaries" like Zuckerberg demand progress and excellence by doing what they can, which is to yell a lot and fire the expendable guys in the red shirts.

    • I would agree with this view if it was a leader in the field like Google doing the mass layoffs in Deep mind. But Meta's AI program is troubled and it really does seem to be just restructuring as a part of integrating the ScaleAI leadership, however misguided that move may be.

    • No it's nothing to do with monopolies having formed in the space. Especially considering there's a wide range of players producing a wide range of models.

      The reality is Facebook has kneejerk management. Always has. They over spend, over hire, and when they under deliver they let people go.

      600 engineers in a 3000 strong group isn't an indication that they are firing the brightest with potential for competitors to come in. It's an indication that either:
      a) their knee jerk hiring spree didn't do sufficient qua

    • Re: (Score:3)

      by dfghjk ( 711126 )

      "...how bad corporations are..."

      And by corporations you really mean people, because corporations are merely a reflection of the people who lead them.

      "...so that it can be owned by two people..."

      There it is.

      Why so many words to say that our society is threatened by the greed of billionaires?

      • Re: (Score:3)

        by HiThere ( 15173 )

        Corporations are not merely people, they're people acting in an environment that rewards particular motivations and discourages others. They're essentially a minimally regulated bureaucracy. They positively select for those who are power hungry and immoral to be rewarded.

    • Re: (Score:1)

      by DarkOx ( 621550 )

      In the past they wouldn't risk firing that many engineers because they would end up at potential competitors

      In the very recent past maybe. That was a function of cheap credit and in the case of AI irrational exuberance that is now *starting to wane.

      There is nothing normal about companies keeping employees they don't need just to deny them to competitors. That isnt in any way economic efficent either, and I would argue not even really good for the employee, who is either on the bench or doing some kind of make-work.-project.

      Your posts are just ridiculous, it literally does not matter what the news is - your warped

  • bad management (Score:5, Interesting)

    by gurps_npc ( 621217 ) on Wednesday October 22, 2025 @10:23PM (#65744586) Homepage

    They got impressed with the new guys, so they bought them out.

    Their old employees were almost certainly good employees. Otherwise they would have fired them BEFORE the take over.

    But after they bought out the new guys they thought that it would be stupid to buy new cow if old cow was good cow. So old cow must be bad cow. Sell old cow.

    The truth is most likely that all the people involved EXCEPT the management are probably extremely competent. The management thought to save money in the place they were trying to spend money. That has to be the stupidest idea ever. If it was worth it to buy the company, then they needed all the smart people, including the old employees.

    Why? Because there is the myth of the singular genius that invents the product. Science is not engineering. Engineering is not science. The AI 'inventors' are hiring engineers, not scientists. They are figuring out how to do something the scientists already have theorized is possible.

    Any AI improvements will be done by a whole team of very qualified engineers doing the hard work of bringing the scientific ideas to reality. No one man - or group of men - is going to be that much better than other people. More smart people will however speed up the process.

    Management was stupid. If hiring them was smart, then keeping their old employees was also smart.

    • Good comment. Progress is based on team work, nobody is genius enough to do it alone. And I agree management is stupid. Just look how Meta is fumbling with LLMs, how the Chinese Qwen and DeepSeek do laps around them. It was a clear management issue and the poor guys became the scapegoat.

    • I love your cow analogy! Here on slashdot we used to have car analogies, but these days the cars drive themselves, so what's the point?

      A cow doesn't milk itself! It's perfect!

      Cow: An analogy for the next generation.

    • Their old employees were almost certainly good employees. Otherwise they would have fired them BEFORE the take over.

      This is far from certain. Often when a company wants to be bought, they hire people to lead purchasers to believe they are growing faster than reality. Caveat emptor.

    • Re: (Score:2)

      by HiThere ( 15173 )

      The only thing wrong with your comment is that science isn't a "lone genius" driven activity either. That it sometimes seems to be is an artifact of the way histories get written. Without a massive support group, Einstein wouldn't have accomplished anything.

  • In fact (Score:3, Funny)

    by GrahamJ ( 241784 ) on Wednesday October 22, 2025 @10:29PM (#65744590)

    Every leader at meta is a wang

    • Especially their large systems. Running cobol on that big brown Wang. It could always keep up with every request. It never failed. It just kept going all night long until all the paper in the printer was spent.

  • Is that wise? (Score:1)

    by Anonymous Coward

    An AI that is already broken having its staff cut?

  • What was that popping noise and why is everything now covered in... is that soap?

  • the AI will program itself now

  • Oops (Score:2)

    by DrXym ( 126579 )
    So Meta employees were literally training their replacements.

    • Re: (Score:2)

      by allo ( 1728082 )

      AI engineers are not simple programmers. If we're at AI writing better AI, we will see much different systems than we have now.

  • Sigh. (Score:2)

    by ledow ( 319597 )

    Just wait for it.

    Headline: "Meta uses AI to replace hundreds of jobs."

    Reality: "Meta stupendously over-hired for the AI fad and is only now realising that it has no real profitable value to it and it's all just hype."

  • The bubble is bursting.

  • What happened to the MetaVerse? (Score:5, Interesting)

    by monkeyxpress ( 4016725 ) on Thursday October 23, 2025 @04:41AM (#65744904)

    Seriously why are investors so quick to forget about Zuckerberg's all-in bet on the Metaverse? He spent billions on it, and I believe it was part of the reason for the whole renaming/restructuring of FaceBook as Meta.

    I guess it's the same with Musk and his endless promises, and now even Apple with their iPhone 16 debacle. It's like markets just reward making huge claims, not delivering, and then declaring that 'oh but don't worry about that previous mistake - i've found an even bigger next big thing'. Capitalism needs discipline in the form of a market beat down when you get things wrong. Without this it just rewards stupidity. At a minimum there may be much more talented entrepreneurs getting starved of capital because whatever idea a FAANG gets sucks up all the money.

    • It seems to me that there is an enormous amount of people who think a person is right just because they have money and a lot of those people happen to have money. Like how Musk said that cameras were enough for self driving. No one stopped to ask if that made sense, they just threw more money at him even though he is where he is because he was born wealthy and is not a particularly smart person.

      • Re: (Score:2)

        by HiThere ( 15173 )

        Well cameras & lights should be enough for most safe driving. Or at least as safely as people drive. (I'd add mics, but a lot of those people are driving in noisy enough environments that they can't register a loud noise off to their left.) Maybe add vibration sensors.

        OTOH, if you want to drive more safely than people, it probably helps a lot to have accurate distance ranging.

        • Cameras and lights are not enough if they are in a fixed position on the vehicle, obviously. This was called out years ago, now admitted by Musk.

          • Re: (Score:2)

            by HiThere ( 15173 )

            Drivers sit in a fixed position. So "should be able to be as safe as humans" seems reasonable. This doesn't mean or imply that the actual implementations are as safe.

            Besides, I'd prefer if automated driving were safer than human drivers.

      • Re: (Score:2)

        by ebunga ( 95613 )

        They pivoted to video, err, I mean they pivoted to AI.

    • Re: (Score:3)

      by mjwx ( 966435 )

      Seriously why are investors so quick to forget about Zuckerberg's all-in bet on the Metaverse? He spent billions on it, and I believe it was part of the reason for the whole renaming/restructuring of FaceBook as Meta.

      I guess it's the same with Musk and his endless promises, and now even Apple with their iPhone 16 debacle. It's like markets just reward making huge claims, not delivering, and then declaring that 'oh but don't worry about that previous mistake - i've found an even bigger next big thing'. Capitalism needs discipline in the form of a market beat down when you get things wrong. Without this it just rewards stupidity. At a minimum there may be much more talented entrepreneurs getting starved of capital because whatever idea a FAANG gets sucks up all the money.

      Investors are stupid and easy to manipulate. Particularly when it comes to leaders, add to that the fact that it's near impossible for a small investor lead movement to unseat a board or executives as they themselves will hold significant quantities of the shares, in particular voting shares. The whole point of stock buy-backs was to consolidate power back in the hands of the few who support the board and C-levels. Ergo, the current management becomes more or less unaccountable. See Boeing, QANTAS, Intel an

      • Investors are stupid and easy to manipulate.

        In what way were they manipulated? Zuckerberg was very up front that he was going to pull all stops to invest in VR, and he did, all the while returning continuous healthy profits to shareholders and investing a not out of line proportion of revenue into R&D. It just is shock and awe because it's a big number, but then this is Meta... everything is a big number.

        add to that the fact that it's near impossible for a small investor lead movement to unseat a board or executives

        Why is this relevant? You invest in a company whose direction you agree with and who you think will return you profits. The idea that you need t

    • They haven't forgotten anything, quite the opposite, they support it. It sounds scary to you and me because billions is a big unfathomable number. But the reality is Zuck spent less on the metaverse (and actually spent less on R&D in general) than many peers in the tech industry. It's just not newsworthy when it's spent on existing markets but becomes shock and awe when it is spent in an emerging market.

      What have they spent? $50bn all up. All the while making $120bn a year in revenue and returning $25-3

  • Wait, not like that!
  • Seems like the hype is starting to wear off. Good.

  • Seriously, if you want to push forward on a research topic, you don't want a team of thousands. A team of 600 is already far too big. If they had 600 extra people, well, that's a research team that is going nowhere...

    A core group of 10-15 researchers and engineers. Maybe a few such groups, to chase different ideas. Supporting staff, maybe an equal number. If they had 100 really good people, that would be the max. Add to that the staff required to keep their public offerings up and running - how many is th

    • Re: (Score:2)

      by allo ( 1728082 )

      Don't underestimate the number of people that can be useful for AI training.
      Yes, you probably won't want to have more than max 10 people designing the network structure of a single model. But while they do so, other people need to think about how to prepare data (I don't count the people who actually do the preparation, they are not as well paid as the 600 here), how to augment it, how to optimally utilize your cluster, etc. If you train a model with a trillion parameters, you don't just schedule a job on a

  • Given that AI is a money sink, and revenue increase? Is it actually higher than inflation?

    The AI bubble collapse is coming.

  • The old AI group leaders probably didn't get along with the shiny, new Wang, and so they had to go. When Wang stops being able to sucker Zuckerborg on the whole superintelligence thing, he'll get replaced too. He'll be extremely rich by then so good for him.

Slashdot Top Deals

MESSAGE ACKNOWLEDGED -- The Pershing II missiles have been launched.

Close