Post

I'm sure a slowdown will be good for their IPO prospects.

i'm happy if the idiots find an off-ramp

With China going open-source, having a walled garden would seem tough to accomplish.

The slowdown they want is for everyone working on the open-weight models to stop because the lack of a moat is becoming more and more obvious to the business idiot class

If this were any closer to "Atrocity", we'd be talking about copyright infringement.

It also implies that the AI is currently developed at an impressive/uncontrollable speed... the models improve so fast, so fast... one breakthrough after another, let's slow down a bit, jeez, invest, thx

My money is on the scenario where they are trying to use this as an excuse to bail on their over leveraged obligations

What would have happened to John Deere's stock if they stopped at the horse drawn plow?

What they want to slowdown is only our awareness of what they are doing. Nothing more.

"The Chinese should stop figuring out how to do this just as well but way cheaper than us and we should start to very slowly let the air out of the asset bubble I helped inflate, but all in way that allows me to remain very wealthy and important"

I also listened to your podcast on the huggingface hack and you asked why they would even create these tools. I was going to give you some reason why but I changed my mind, I think we should go even further in testing LLM hacking abilities.

What they are wanting to "slow down" is the impatience of investors who want proof & evidence the LLM machines will turn a profit. When their bills come due, they will point to: wE hAd tO sLoW dOwn! iT iS nOt oUr fAuLt! wE wERe jUst bEinG rEsPonSibLE!

The hopes and dreams of four years ago are no longer realistic! In fact there's no avenue to a cash positive business! Let's tap the brakes a bit! Also please give us $50b

It is particularly rich of him to worry that AI "could outrun our ability to understand and control these systems" when that is exactly what Claude Code is ALREADY doing to thousands of codebases across the world.

And that's the answer, because they have nothing replicable. They're going to make a lot of noise hoping we won't notice the vapor.

They all agree they want a slow down, but I'm pretty sure they also agree that it would require China to buy in to be effective, right? I don't think this will ever happen. I think this is just them talking the talk of safety. I don't think it has any substance.

Part of their strategy of the “slowdown” is the perceived issue of danger (self propagated) which can only be addressed by limiting access to the 2 major players, excluding all other companies (eg chinese). Government controls incentivize their financial lifeline and prohibit competition

I was thinking about this today and part of me wonders if they are all playing a game of chicken with one another to see who they can get to slowdown first while they leap ahead.

I think it's a ploy to get better regulations now vs after January when a far more hostile Congress is likely to be in office. If they pass some bullshit law now they can have token regs and lots of tax breaks.

No, see, they mean a criticism slowdown. They're going to need all of humanity to stop criticizing them or they'll be forced to end humanity.

Doesn't the media handle that part of it for them?

if we had a functioning antitrust division this would be a big problem!

Also, isn't "we can't slow down" still being blasted to every town fighting a data center because "China is gonna git us!" If we do? This is absolutely an attempt to ease into the crash slightly lighter

When they miss their revenue targets, they'll now have an excuse - "We had to slow down for safety's sake" (wink wink).

Big "You didn't make me stop. I wanted to stop" energy.

They expect things to clear up in two weeks lol

For some reason I am reminded of the LLMs running vending machines and agreeing to all price fix; but then immediately turning around and lowering prices to undercut the others that agreed to the higher prices.

I want ai companies to shut down

I'm assuming it means "don't worry investors, we're not hitting an upper limit"

I think this is their way of retroactively rationalizing how LLMs haven't gotten any better in like two years and are stuck with a certain inevitable percentage of "hallucinations" - so hey let's IPO next year - and if it seems like the product isn't improving, well we did that on purpose

oh come now

I think their plan was to literally invent genAI god and then ask it what to do.

Occam’s closeted married lover

Every few weeks the BBC is like,"An AI ordered a succulent chinese meal on it's own" and anyone with half a brain just chuckles and says "no it didn't."

But, .. but if these companies don't have an actual plan or strategy, then it would have been very dumb for the money people to bet so much on them!

Theory? Im not aware of a single action ever taken by any of them that has any indication of a long term strategy behind it

For LLMs to just read all the research papers, understand them, and divine improvement, LLMs would ALREADY need to be superintelligent. Even today's human AI researchers can't do that. "Recurrent self improvement" is prob 30+ years away, with many significant steps before we can get close to it.

Furthermore, LLM capabilities are entirely based and information input, so RSI would just mean the AI training itself on its own output, which guess what - is definitionally incestual training and is well known to lead to model-collapse. I'd be curious to hear them explain what RSI actually entails

Yeah, I get the human reason why many at first thought these things couldn't be better than us at our jobs - or at least part of what makes our jobs. But they are. The changes in two years have been astounding.

Evolutionary algorithm digital improvement / creation of better designs do thousands of generations of selected iteration in hrs. For LLM AIs, multiple iterations would cost tens or (likely) hundreds of Billion$ & take many months, or years. Who will pay? with zero profits (only losses) in sight?

If you thought that your alignment science was significantly behind your ability to get to RSI and therefore agi or asi then yes, particularly if you believe that misaligned models could be catastrophic to humanity

it's nowhere near an actual singularity (no human involvement), but it's definitely somewhere on the recursive self-improvement spectrum current AI already helps write most of the code/cleans up datasets/takes part in reinforcement learning/etc. for the next AI

it's just a PR narrative

This feels like a stunt to get out of their obligations. “We are so close to creating God that if we don’t stop, it will be bad for everyone. Please don’t make us borrow and spend this money and create the evil AI God.“

Putting my conspiracy hat on. The bill is coming due soon. The vast sums of money are not generating the impovements needed. This gives them an easy out for investors. ‘Oh the models could do what we promised but can’t now cause we care too much’ Yadda yadda, cycle continues.

I’d rather have an “AI Stop” or an “AI Shutdown” followed by a dismantling of their models and charges brought against those responsible for the damages they have caused.

They talk about slowing down, but practice unlimited scaling. Another AI deep fake to deflect criticism.

which is the most expensive and still requires development (tech is currently maxed at 15-30 seconds), but once they've solved those math puzzles, it'll only take a few years to become far cheaper too. So who's going to pay for anything in a few years? Yeah, those trillions are never coming back.😬

bsky.app/profile/adel...

TBF the AI bubble was going to pop anyway, it just so happened to pop violently and sooner 'cuz global warfare and oil shenanigans imo.

My first thought .. bsky.app/profile/carl...

Whatever they actually say, it'll probably be an attempt to prevent anyone else from training super large models. Create a market niche only they can serve, and legally prevent competition.

We need to slow down the AI == We need to prevent the bubble from completely popping for as long as we possibly can

Slow down probably means regulating access to open models so the big players can profit.

This is in response to the public backlash and I call bullshit on the "slow down". They want the public to perceive them being responsible. Fucking liars, all of them.

They are just looking for a way to try to present "we are running out of money" as "we really care about what AI could do to humanity". As ever, all they do is lie.

"maybe that’s how hyperscalers could reconcile with letting them out of compute deals?" Maybe they can. I doubt Wall Street will.

Me when my magic supah technology isn't so super, but I can't let anyone know, so I just pretend I'm scared of my own invention:

This is an excuse that lets them avoid having to prove their software is still improving. Instead of admitting it has stagnated, they simply claim it would be 'too dangerous for the public' — it's their way out.

It'd slow down if fuckwits stopped using it. We need to bring back the 'computer nerd' slur. And throw bricks at dorks who buy the iPhone flip

I think they know the bubble is going to burst soon and are using this as an excuse to shareholders to not pull out—and give us more money to fix this! Also think they weren’t prepared for the con to go downhill this fast and are simply playing for time.

I think it’s another reason to delay IPO and a replacement for releasing financial data.