Supposedly the change of heart came after "the first security incident that I have felt very viscerally," the Huggingface escape. To me it sounds like he's worried about the competition breathing down his neck and wants someone to come in and call a time-out.

top 50 comments

sorted by: hot top controversial new old
[–] 193 points 2 months ago (3 children)

Every "Sam Altman said..." headline can be summed up as:

"Manipulative liar says whatever he thinks his audience wants to hear."

  • source
  • hideshow 3 child comments
  • [–] 35 points 2 months ago

    "Tech crypto grifter says thing to grift people with crypto, tech or both. More at 11"

    Although i cant imagin any sign of a pullback would be good for any of the companies in the AI bubble. While probably not important on the specifics of what he is saying it is a "we should slow down" which the market aint gonna like. No breaks on this gravy train Sammy boy!

  • source
  • parent
  • [–] 106 points 2 months ago (9 children)

    Likely translation: scaling their models is hitting diminishing returns, while the business world is slowly realizing the math ain't mathing and they're not suddenly 10x more efficient and profitable than before LLM adoption.

    So Sammy and the others need a plausible cover story why capabilities aren't evolving at break neck speed anymore, to prevent their beautiful bubbly bubble from crashing down on them

  • source
  • hideshow 9 child comments
  • [–] 7 points 2 months ago* (2 children)

    Feds made them be the fall guy though. Data centers are new infrastructure for the government to use. That's the main reason for all this ai bullshit. Chapgpt and this consumer grade bullshit is the unplugged video game controller they gave us to makes us feel special and a part of something while THEY play "The Game".

    An expansion of the above; https://reddthat.com/comment/28125951

  • source
  • parent
  • hideshow 2 child comments
  • load more comments (1 reply)
    [–] 52 points 2 months ago (2 children)
  • [–] [S] 30 points 2 months ago

    Yeah, we've definitely been seeing diminishing returns with the latest frontier models. I think that's why when GLM and Kimi come in at 1/5 the price of Opus but are just as good, or when Deepseek comes in at 1/10 the price and is catching up quickly, they have to worry. There may be nowhere for them to go.

  • source
  • parent
  • [–] 48 points 2 months ago

    He should accelerate himself off a cliff.

  • source
  • [–] 46 points 2 months ago

    Company can't keep up with competitors, claims it is for altruism

  • source
  • [–] 44 points 2 months ago (1 child)

    Sam is a con man, why is anyone listening to him.

  • source
  • hideshow 1 child comment
  • [–] 36 points 2 months ago (1 child)

    You can tell when Altman is lying by noticing that his mouth is open.

  • source
  • hideshow 1 child comment
  • [–] 33 points 2 months ago

    Stop paying attention to this lying fuckwad.

  • source
  • [–] 25 points 2 months ago

    Way to get ahead of the slowdown Sam. Control the narrative. You "chose" to scale back.

    "Actually we almost made the A.I. too good and it scared me cuz of how good it was"

  • source
  • [–] 23 points 2 months ago (2 children)

    No he isn't, the only reason Sam Altman would hit the breaks is if he was falling behind and needed to catch up.

  • source
  • hideshow 2 child comments
  • [–] 21 points 2 months ago (5 children)

    He wants regulations to kill the competition. Open source is gonna get fucked in the coming months. The whole media campaign has been specifically for this.

  • source
  • hideshow 5 child comments
  • [–] 3 points 2 months ago (2 children)

    Sorry if this is a dumb question but I keep seeing open source and open weight when talking about LLMs. Are these two terms interchangeable? If not, what do they mean and which is correct?

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 2 months ago*

    They mostly are interchangeable for most people but almost none of the llm weights are properly open source.

    They don't always give the training scripts and never the actual data so it isn't reproducible. The data isn't really feasible to give. A lot of them release papers at the same time where they explain the training with snippets and how it works, but it's not the same as receiving a script.

    That being said, some get really upset and stress the distinction but it's a bit silly imo. If it's free and easily fine tunable, I don't really see the issue of calling it open source. Giving the data for it would open them up to lawsuits from the copyright mafia and wouldn't bring much to the table. Training these models is completely out of reach for individual and small to medium sized organisations.

    Training a small model was more than a million dollars in compute a year or two ago.

  • source
  • parent
  • [–] 2 points 2 months ago

    Open weight is like getting a compiled binary with no source code. Without the training data and pipeline code, once the original developer stops releasing new versions, it’ll be difficult for the community to pick up and keep going.

    For example, if Alibaba decides to only release trillion parameter models from now on, Qwen 3.6 may be the most intelligent Qwen model that will ever run on consumer hardware.

  • source
  • parent
  • load more comments (2 replies)
    [–] 19 points 2 months ago (2 children)

    Why do we care that some greedy bastard is doing a fake apology tour?

  • source
  • hideshow 2 child comments
  • [–] 19 points 2 months ago

    We need to scale AI development!

    Openweight models start growing

    No, not like that!

  • source
  • [–] 18 points 2 months ago (5 children)

    It's difficult to think of a more punchable face than Sam Altman's.

  • source
  • hideshow 5 child comments
  • [–] 7 points 2 months ago (3 children)

    I dunno, Dario Amodei’s face is my personal choice.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 14 points 2 months ago

    LOL...those graphs didn't infinitely extrapolate Sam?

    99% of failed companies are from graph extrapolation in board rooms.

  • source
  • [–] 12 points 2 months ago (2 children)

    I've just been reading the Wikipedia article about the huggingface intrusion, and it reminds me strongly of the descriptions of hacking in Neuromancer.

  • source
  • hideshow 2 child comments
  • [–] 6 points 2 months ago

    He can do what he wants - except closing Pandora's box. It stays open. The race is on, whether he participates or not.

  • source
  • [–] 5 points 2 months ago

    Ahhh yes, decelerate into the singularity. We wouldnt wanna ram it full speed, it might break!

  • source
  • [–] 3 points 2 months ago

    I’d be ok with him rapidly “decelerating” at end of a long fall…

  • source
  • [–] 3 points 2 months ago

    Meanwhile, me still using GPT 4 as: "Rephrase: " CTRL+v

  • source
  • [–] 3 points 2 months ago

    Nvidia is backing the wrong horse

  • source
  • [–] 3 points 2 months ago

    BREAKING: A Billionaire Said Some Bullshit.

    I know downvotes are generally frowned upon here... But c'mon. This ain't what we should be hyping up attention for.

  • source
  • load more comments
    view more: next ›