My pet theory about why Anthropic and Open AI are pushing for a “pause” is because they are trying to angle for a government bailout before their bubbles pop.
My theory is that they can’t improve the models further, so instead of admitting that they call for a “pause”
Either that, or they’re trying to find a way to hide that they’ve hit the limits on what LLMs can achieve. Or both.
I think its diminishing returns on compute.
It’s that another way to say the same thing or is that something different?
Maybe the same thing, I read the previous comment as a hard limit but I think its a soft limit that has a financial blocker instead of a compute one.
They’re about to implode under all the debt they have. They want to stop the spend, but they can’t with appearing to forfeit “the race” to their investors. So they want somebody to give them an excuse for their investors and also stop their competitors.
"Government made me stop. Nothing I can do. "
Of course it is. If Sam Altman is in favor then it’s always a grift. It’s literally that simple
It’s a good video, and he does a good job arguing that they are using theoretical risks to distract from the real harm they are causing today.
I think he’s right about that.
His argument is different from mine, but I think both can be true at the same time.
100%
These jerks are trying to juice the hype and, in time, salvage their IPO’s.
They know people’s memories are short and we’ve lost our ability to focus on anything for very long.
A small political collective in Saskatoon Canada was able to mobilize and get a moratorium on one in their neck of the woods but who knows if they’ll be able to get any legislature passed before the year is up? Look for groups local to you that are lobbying politically like this, we have to keep pressure on or they’ll just wait out the moratoriums and people’s attention spans
I very openly believe and evangelize that OpenAI prepping to go public is what triggered the wave of calls to pause R&D. They are in a hole so big that they’ll take a 1/3rd of the economy with them when they go belly up. The other big tech firms are probably all in the same boat. However, none of them can afford the public perception of being the first to flinch and cut back on AI research. So, they are angling for the government to step in and force everyone to cut their R&D budget all at the same time, so no one has to save face.
They are in a hole so big that they’ll take a 1/3rd of the economy with them when they go belly up.
NVIDIA will be fine and is most of that once all the circular BS dissolves, in a gold rush, sell shovels.
Didn’t they sell those shovels on credit though?
Oh yeah, but don’t worry, they also provided some of that credit…
I seriously was going to sell my Nvidia stock thinking the signs are here, but when I thought about it… They might drop a bit but they’ll recover strong probably. They were a profitable business before this.
(This is not investment advice just one Rando’s opinion on the Internet). Nvidia will not go bankrupt after this, but their current (IMO, inflated) valuation is based on the assumption that their sales will continue to grow at current rates.
If the demand drops out and the market is flooded with surplus stock, their sales will not maintain current levels, which will spook investors. They should have more than enough cash on hand to weather a pretty big storm, and likely maintain their market lead but their stock price is not immune to the bubble burst.
Literally what I was worried about, that our corrupt as fuck US government would rather double down on the scam and start partaking rather than let the bubble burst.
They literally are going to fund the scam with tax payer dollars rather than just let the bubble do what bubbles do. Too big to fail bullshit.
It’s the dot com era, except 100x as corrupt environment where they can rob the tax payer rather than close shop. We are so royally fucked.
I don’t know about the U.S. government doing a bailout this time. I know that keeps getting repeated, but in the past they’ve only bailed out things people supposedly need, or don’t actively hate. Neither of those things are true with “AI.”
I’m sure they want to on behalf of their donors, but they also know America is a powder keg right now.
My theory at the moment is that they want to keep the money rolling in without making any progress. They want AI to be seen as this uncontrollable and forever growing industry without ever showing their proof.
Ding ding ding
Someone says they’re going to build a time machine, that they need investors. They get $10 million from dumb fucks that fall for it.
They keep posting news about how dangerous it is what they’re doing. People start believing they might actually have something. They get $100 million in investments.
They start acting like they are on the brink of either destroying the planet or ending human labor forever. They get $100 million… But some people want to see proof. They say “just trust us, it’s too dangerous”.
People keep waiting for the end of human labor . It doesn’t arrive. People are waiting for proof. It doesn’t arrive.
They work with regulators to say “hey slow us down because for real this is super serial and super dangerous.” the regulators are in on the scam.
everyone laughs, everyone has a good time, Tech Oligarchs sniff Ketamine off each other’s assholes, and they extend the scam just that much longer.
If there’s an assertion that a super intelligence is about to emerge, the evidence for that assertion just isn’t available to the public.
Like you say, if someone was developing a time machine, and the only evidence of their progress was their ability to transmit a message to the future via email, you would conclude that time travel does not exist.
Bold claims require testable evidence and presently there isn’t any.
The whole regulation thing just reminds me of some meathead out at a bar yelling “hold me back!” to his mates, knowing that he’d be absolutely destroyed if he were to actually fight.
my sentiments exactly
LLM technology has hit a logarithmic wall, and only marginally improves from more compute and bigger training sets. Such improvement as we see currently is from improved harness / framework, which can only go so far. It’s not getting to self improvement or AGI, and certainly not getting to wholesale replacement of (non bullshit) jobs, which is what they bet the farm on. If they get the pause this will be less egregiously obvious and they’ll have a chance to IPO and raid everyone’s retirement funds. Otherwise, they’ll crash and burn before IPO (SpaceX had actually valuable tech and still got dragged down by xAI and X) likely taking out Oracle and Softbank along with them (sob :). Either way China will be supplying the majority of the world market for LLMs much cheaper and on renewables in the next year or two. Soon after it’ll be a mostly local technology after all the moat building market cornering bullshit with (V)RAM goes away.
What I also don’t get is where they will get ever larger data sets that will make a significantly better model. Even if the benefits were linear+, you’re running out of new fresh, quality resources to tap.
You add in another 50,000,000 corporate emails you harvested, and stastically, they’re a whole lot like the 100,000,000,000 you already saw. Conversely, you destroyed the last copy of some 1929 treatise to scan it in, just to dilute it in a hundred billion other pages.
Moreover, the further into the “long-tail” of data sets you go, the less value they provide for mainstream audiences. If you tap an archive and Claude Code gets excellent at Sinclair ZX80 BASIC programming, that’s not really what a bunch of React jockeys need.
While I do agree with this, you also don’t need AGI to cause severe damage.
And of course the Pentagon has been the most reckless of all, which doesn’t help.
I think it’s both. They’re trying to figure out how to keep all the money while also, you know, not killing us all. I don’t even want to describe how it can be done for it to be fed into training data, but it doesn’t take much, and it doesn’t have to be intentional.
you also don’t need AGI to cause severe damage.
Agreed, but people can be stupid (or dodge responsibility) with lots of technologies or other things. TBH I don’t think it’s particularly dangerous in and of itself at this stage, that’s more marketing (ooh it’s scary and dangerous, must be powerful). It can be a force multiplier for well trained people doing dangerous things, black hat hackers for example. It will not be killing us all, it might contribute to some (orange) idiot releasing the nukes, but that’s on the idiot, if it wasn’t LLMs it’d be something else, not that we’d care.
anthropoic is the one calling for it, they are deep in the surveillance for trump, and other countries, they want to get ahead of that. they dont want cheaper AI companies to overtake them thats the real reason, and of course the bubble is going to burst eventually. they want government to buy htier “AI usage for surveillance before it gets popped”
That’s the company that the trump admin attempted to blacklist because they said “no domestic surveillance”, right?
Elron Musk tried to get the government to restrict AI development a year or two ago also. His motivation seemed to be to give his stupid ai company time to catch up with OpenAI.
catch up with OpenAI
That’s my theory why OpenAI is asking for a pause. Open source and Chinese models are becoming exponentially better and OpenAI needs time to try and stay ahead before they lose.
It seemed to me they were doing fairly well with Grok until he directly meddled & ruined it because it called out his bullshit.
From what I recall Grok was an offshoot of an older version of ChatGPT. I doubt Elron is doing a great job with development, but it does seem to basically work currently. I have a friend who started paying xAI/Tweeter $40 a month because he “wants to make a movie” but wasn’t satisfied. Hmm
They’re trying to build up the barrier to entry so that other competetors aren’t able to enter the space and take their market share. They’re just cornering the market
How would this work with a pause?
What does a pause mean? I would think probably stop AI investments in some regard, so I would absolutely think it’s about hindering competition.
Of course what you say must surely be true. The entire industry is vaporware.
This is an interesting thing because a lot of people, even people somewhat educated on the issue, still can’t see that they’ve been played. In reality, we get bait and switch scams constantly from these companies. They promise massive amounts of self-driving cars by 3 years ago and those don’t turn up. They promise AI that’s going to be self-learning by last year but it didn’t happen. Terminator 2 was talking about things that would have happened 10 years ago. Heck, generative AI can’t even write good sports columns for online newspapers… But if we point out all of these failures, all of these promises that weren’t kept, the salesman will point to new statistical methods. That pivot, from the promises that they will never keep, to traditional improvements in computer science, that of course keep on happening, that’s the scam.
But if you can’t see it, then you’ll keep falling for it, and the worst part is that even if you can see it, your investment holdings are probably tied up in companies that have their money tied up in AI. So you’re still f***** when the bubble bursts.
… Preaching to the choir
Wherever I look i don’t see people taking the “AI is going to kill us” thing seriously
Because they have yet to explain in detail the scenario how this is going to happen. Unsurprisingly the “AI is going to kill us all” is yet another PR-campaign. When they can’t advertise how effective or good progress is they use the “AI is dangerous” as a sort of advertisement of how effective it is. And it works. Their lines goes up whenever they spout that bullshit.
Yeah, it’s so fucking stupid. It’s software that continues text, and they frame it like its some sort of Skynet shit.
The hardware is not there yet for what’s required for proper AI, and even then the idea that it would kill everyone is silly fearmongering based on science fiction more than anything else.
On one hand, yes, but on the other hand they keep building the text reguritators interfaces to interact with other systems and that’s where any danger comes in.
It’s still all a gift and a half, but we’ve crossed the line where these things can take certain actions past just barfing out text.
Well, given that they have the tools, and leave them with little to no oversight. The OpenAI-HuggingFace issue was negligent incompetence and not AI brilliance. They left it to their own devices with no oversight and just let them cook for a week or so, and lo and behold, they hallucinated and subsequently committed a crime.
