The current "approaches" will not give any scalable results as they ignore basic physics. It's at best a very power-hungry gambling with database entries, at worst a noisy and expensive waste of time. I think that it's fitting to call users of these things for gamblers, and the tokens they use and pay for are the same as in gambling, and they are all gambling for a good result. In the end, the house always wins.

Everybody else to pause so they can dominate

Clear nonsense for many reasons for anyone in the business and anyone who knows how computers and/or business works, even if they're not turning redbulls into vibecoded goodness themselves.

Politicians?

Theologians?

Landscapers?

Who's the indented recipient of this propaganda?

They are finding a plateau with where the LLMs can go and could use the narrative of a "pause" to explain why capabilities are going to iterate in a more 'evolutionary' way instead of the revolutionary way folks are expecting.

There isn't to my knowledge a mechanism for the models to "self-improve", whatever one may think, at least the output doesn't have access to change the model in any way. The narrative of "oops the AI started evolving itself on accident" doesn't have a way to happen.

Considering that even the vaunted Opus 4.8 can't always develop mundane traditional software beginning, it's hard to imagine it could rework the model itself even if it had such access.

beating them on all the benchmarks? And why isnt self improving AI a good thing? You're testing them for alignment right?

what bad actors are out there?

what else happened recently? Maybe this: [1]https://yro.slashdot.org/story... [slashdot.org]

Yes Pause! Pause!

[1] https://yro.slashdot.org/story/26/06/02/1658231/trump-signs-ai-executive-order-asking-companies-to-give-government-early-access-to-models

They are all scum.

But.

That does not change the fact that as soon as one develops a self-improving AI, the world will become a very different and probably very awful place.

So stopping one and trying to stop the others is the best we can hope for.

It does not matter if China or the US or Timbuctoo does this first.

The world is fucked whoever gets there.

So please let us try and stop this now, in any way we can.

Isn't that AI will develop some sort of "super-intelligence." It's that AI already is and will continue to be integrated more and more into daily life, and as it becomes more complex, there will be less and less visibility into how it's arriving at whatever response or behavior it exhibits. We'll lose control, not because AI is taking control, but because there won't be any controls.

One of the problems AI companies have been persistently crying wolf with headlines about the dangers of the latest incremental release with similar calls for regulation and slowing it down for the last several years already. Assertions with the benefit of hindsight have proven as ridiculous and outdated as the baseless 10^26 idiocy having already weaved its way into legislation.

Anthropic's statements coincide with an impending Anthropic IPO making them impossible to take seriously as anything other than a marketing ploy.

Any firm genuinely afraid of AI wanting to act responsibly and slow it down can always elect to stop funneling hundreds of billions of dollars into the enabling knowledge and industrial base by working on something other than AI. Nobody will ever do this because it isn't in their best interests. When push comes to shove self interest is all anyone cares about.

"If it were possible to effectively slow the development of this technology to give ourselves more time to deal with its immense implications, we think that would likely be a good thing. But if a slowdown simply lets the least cautious actors catch up technologically, it could leave everyone less safe."

Note Anthropic is not saying they will be acting responsibly. They want everyone to agree to collude and nerf themselves. If such an agreement can't be reached then screw it full steam ahead we can't afford to fall behind ... for ... drum roll... your . .. SAFETY. I give people more credit than to take this nonsense being peddled by Misanthropic at face value.

This feel like they want to slow customer expectations in a coordinated but public move across companies to save money. Also feels like marketing to excite VCâ(TM)s with âoeso good weâ(TM)re scared- it would be a shame if brave VCs invested in usâ¦â

The shit we were gonna sell you? It's too fucking lit. It's so good that we're rally worried, like, for your own safety, dude. We should probably think about taking a break and taking it easy so this good shit doesn't become just dangerously good, like, on it's own and shit. Then we couldn't stop it from becoming always better forever and that would be bad! I mean, it would be good but too good to the point of badness. So we propose everyone takes a chill pill for a little bit okay? Just for a moment, everyone, chill.

Anyway the IPO filings are going great and you'll be able to buy in on our good shit soon enough. But don't worry, we're gonna keep it on the DL so it's just good and not too good. We got you, dude.

in their daily lives, the more dependent, helpless and dumber they will become. But then that might be the goal.

This is just "please stop our competition" and "please believe our unhinged claims about what this tech can do".

Paging Dr. Forbin.

Our competitors have done something that we can't compete with yet. Ask them to slow down and wait for us.

I had to laugh at this, as that ship has sailed. There is no way they will put that genie back in the bottle; furthermore, we've had AI application in classified environments for a long time. I feel like this is just another move to establish control. Remember back when Microsoft, OpenAI, et al, moved to create some oversight mechanism for AI? That kind of power over society is unacceptable. But here we go...

...in Berne.

1) We do NOT have AI. We have Large Language Models and similar predictive software.

2) When it attempts recursive improvements we get recursive deterioration. LLM fed the output of other LLM get worse, not better. This will NOT change. The best they can obtain is 0 deterioration. Why? Because prediction needs good data. Predictions based on other predictions is like making a copy of a copy. The best it can do is stay even. The LLM did the best possible prediction in the first round, using it again without more data does NOT work.

3) We already are seeing this problem as so much of the internet has become AI slop that it is feeding AI slop as input for other AI, resulting in worse slop. Example:

Not a robot, not an android, but instead a Large Language Module -> Not a machine, not a phone, but instead a huge English component.

4) AI can be helpful for a lot of tedious work, such as going through theoretical chemicals looking for possible medicinal drugs. But the idea that it is actually becoming intelligent leads people to over-estimate it's capabilities and it's fears.

5) An AI that tries to take over the world is likely to threaten us with an anti-matter bomb it swears it made using a Mr. Fusion machine, a bannana peel and 12 oz of beer. Do NOT fall for it.

If we let the less cautious actors catch up...

Hahahahahahaha.... [breathes]... hahahahahahahaha

An here I was thinking that enshittification was the bane of my existence. Prepare yourselves, fine folk, for the oncoming ENSLOPIFICATION!

If they had some super powerful analysis and synthesis computer that solves problems for them, would it be able to sort this out?

Anthropic file confidential ipo information. Knowing that, take everything else coming out of the company accordingly.

First AGI, now this. We're at a point were the realities of the tech are going to take over. If investors where figure out that the spend on data centers and energy will eclipse revenue and not be recouped, AIs boom is over. There are already cracks in the wall, company spend on AI in some companies is being rolled back. The next 3 to 6 months will show the real data from productivity studies and revenue.

It seems like the most thoughtful AI company of social values, and the most useful use of AI's. I subscribed, let my mind experiment with the AI, and I was inspired. In my human mind, I think it the best, but I would like to hear from other slashdotters what you like, and why? It seems so "wild wild" west to me like nothing I have ever experienced before. I seems most certainly more "wild wild west" than when the internet came online.