‘Too little, too late’: critics perplexed and suspicious of AI leaders’ call for a slowdown
the Guardian
‘Too little, too late’: critics perplexed and suspicious of AI leaders’ call for a slowdown
From the Trump administration to AI experts, plans by the Anthropic boss to boost safety have spawned a largely negative response
I’m not perplexed. When in doubt,look at the wallets. They’re out of money. The interest on the loans is starting to bite, and their revenues aren’t going up like they hoped.
With no exponential improvement in efficiency it means they have basically hit the ceiling on LLMs and constsntly increasing compute to make any progress is no longer drawing in massive funding. They just want a reason to maintain the current level while they shift to sucking money out of the people who are stuck with LLM workflows.
Recursive self-improving AI isn’t happening like they expected. AGI is nowhere in sight.
The pace of advancement is slowing down and they need a cover story for why they’re not living up to their own hype.
The pace of advancement is still going strong, especially in math and “reasoning”. For example the frontier math benchmark is showing big strides for the frontier models. A year ago they were only getting <10% of questions, now sol from openai is scoring 89% and they’re reporting the unreleased astra is at 98% .
They even had to create a new frontier math benchmark called erdos that all previous models scored 0% on and astra got 3% , which may not look like much but is a huge relative jump.
I would say that AI is becoming much more effective for the tokens produced. I have been running Qwen 3.8 Flash-Next on my new PC. This AI model weighs in at 120b parameters, with another 50b added via an Engram table that costs about an extra 20% in RAM for its parameter count. This is a major jump in world knowledge, and the model in question has 262k context length.
While experimental and with rough edges, this model has been able to use Lingua Gacha to undertake a game translation project - understanding and converting a Chinese game into English.
Agentic AI has become practical on local hardware over this last year, and there looks to be much more refinement to come.
00000000
What we are seeing with Anthropic and the other American companies, is a scramble to protect themselves from the consequences of badly planned data centers and poor practices in AI development. Their Chinese competition has produced much more efficient models, while delivering almost the same level of quality as American offerings. This means that local AI is becoming “good enough”, which makes data centres and AI subscriptions unnecessary for many businesses if they distrust the cloud.
We use AI as an umbrella term for so many things. Literally the type of machine learning you’re talking about has been in use in science and mathematics for quite some time and unrelated to all this buzz. It is decidedly not what OpenAI, Anthropic and Meta’s product are. Those are LLMs friend. The promised harbinger of AGI via agenetic AI. So far, all it’s been is an embarrassingly expensive and destructive waste of resources.
They must have just learned they’re not getting the government bail-out they were hoping for, so they’re scrambling for Plan B, which is to play the “safety” card to lock out new startups and foreign competitors. The big companies will probably form an AI Safety Consortium that conveniently sets standards in their own favor.
I don’t believe for a second that the same people who were willing to destroy the planet last week are concerned with saving humanity this week.
Supposedly they are already profitable or breaking even on inference. So the money they are burning is mainly for the bigger and better models. If they stop any further devellopment they might actually be quite close to profitability. But then they will be leapfrogged within a few months unless the competitors also stop.
“slowdown” means “prevent competition and maintain the status-quo”
Yep - they’re established, so it’s ladder-pulling time!
While they quietly panic about the constantly rising costs of running their models, the lack of new training data, and the sheer drop investment capital took while they were too busy hemorrhaging money to realize.
Now they'll act like they're scaling back in order to protect the public and hopefully figure out their golden parachute strategy while cutting startups and smaller businesses out of the market completely.
No it’s simple and not confusing at all: The models are well beyond the point of diminishing returns for investment and the AI bosses know they cannot possibly meet the crazy expectations of the stock markets. So instead of them failing to deliver, they are using this excuse to seem like they’re choosing not to deliver for everyone’s good.
It won’t work - the bubble is going to burst regardless.
You can’t break up with me, I’m breaking up with you type of energy.
No, you don’t understand. If we don’t break up I become Hitler and you become Eva Braun.
I agree on the bubble - think less than 2 years before the jig is up.
It’s going to pop before 2027. It’s a pop not a deflate, the money will disappear in a 48 hour period, not that the theoretical money every existed.
The models are still advancing, especially in math and “reasoning”. For example the frontier math benchmark is showing big strides for the frontier models. A year ago they were only getting <10% of questions, now sol from openai is scoring 89% and they’re reporting the unreleased astra is at 98% .
Ah finally, we are here. The diminishing return is here. The trillion dollar bet is finally showing its first cards, and its flopping hard.
We dont even have to wait, LLM bullshittery is dead in less than 10 papers away lol
Can we, maybe, NOT build the big human eating machine?
The Torment Nexus? But it sounded so cool in that classic sci fi novel “Don’t Create the Torment Nexus”
It’s lucky for us that they are wrong. Still, it’s disturbing how eager they are given that belief. I will never be able to understand people trying to end the world.
I mean, I get being nihilistic, but being enthusiasticly so is baffling.
Hey maybe we should put it all back in the box.
Said pandora
It’s a marketing scam.
They will soon launch AI plans with security upgrades, just to charge more than they do now
All those assholes trying to shove it into our faces. Fucking shut all of that down.
Anyone else find it funny that if you skim over “Amodei” too quickly, it looks like it says “AI Model”? Dude was born to be in this industry.
Maybe they are building a moat.
It’s a shame how polarized people have become. No discussion of game theory making it difficult to slow things down even if everyone in America wanted to. No discussion about potential solutions. Just people feeding their confirmation biases, and obliviously living in their echo chambers.