Post

I'll be more impressed when I see what they actually do, not just say what they'll do

This is just spin. Multiple factors beyond their control (energy bottlenecks, chip supply shortages, rising bond rates) are forcing a slow down of the current hyper-aggressive AI build out.

Maybe understand his position. He doesn't want the world to end until AFTER he has gone public? OK, just guessing.

For the companies that invested so much into this technology? Probably a pretty hefty amount of money lit on fire? For regular people? Maybe a hit to their stock portfolios & 401ks?

Serious answer: regulatory capture. This is a preemptive move to hedge against open weight models.

hey I think there's a mistake in your bio—it doesn't say "credulous stenographer for corpo PR", it says some other stuff about journalism?

None is pretty slow. If these guys believed what they were saying, they'd shutter the business. It's horse shit, but they've convinced a lot of gullible press people, so the press people are going to keep uncritically reporting it.

you realise at this point how openly offensive it is to everybody else, to even entertain the possibility these guys are doing anything at all in good faith, yeah?

This sounds to me like a bunch of damage control. "We are slowing things down because AI is becoming too powerful" sounds a whole lot better to investors than "Right now our product is only good for chatbots, deleting files from your databases and hacking companies, so people don't really like it".

bsky.app/profile/numb...

You’d have to be a fool to take them at face value, or an AI evangelist. Not having “safety” was a choice. What this is really is the fear that if one of the top AI companies screws up something & goes down in flames it’ll kill the golden goose (*cough* IPOs). See also: “Don’t regulate me bro.”

I’d love to see what a….oh I don’t know, maybe a CYBERSECURITY researcher has to say about these “AI guru” takes. If I was a pancake researcher who got into the field of making the best damned pancakes ever, I’d think pancakes were powerful too. To a chef, though? They’re alright.

The reasonable response is shutting them all down. Investigating them for all of the fraudulent statements they have made about their technology.

But it's still just complete and utter bullshit, because the entire argument hinges on the expectation that a statistical language model will suddenly achieve "intelligence".

Sad how many journalists are falling for this. The “embedded evaluators” would be selected by the “frontier AI” companies. The “democratic coordination” are … frontier AI companies. Vague words on “global coordination”. No independent oversight. Etc.

I noticed he said “we” without any offer of what he himself will do. Perhaps we can speak to someone in management next?

well, I really doubt the reasons of Amodei and others - when there are other factors to consider, such as availability of AI data centers (they couldn't construct them fast enough AND there is a backlash), resources for both training and inference are clashing, and threat from overseas, among others

This is first mover advantage trying to create barriers to entry to stave off competition. Oldest trick in the book. And people are buying it hook line and sinker. It’s embarrassing.

Ah, the magazine for suckers, you never disappoint