Dean W. Ball argues the US AI industry has frozen public releases until the Claude Fable 5 impasse is resolved

I don't think there's a coherent plan or single motivation here, but this does seem to be one effect

@Afinetheorem correct. if anything development can maybe be faster

@TheZvi I was short this starting Thursday of last week. I wrote up my model in https://www.lesswrong.com/posts/zhRe3tdBpsZbGCdDK/world-modeling-the-us-vs-anthropic-standoff-on-claude-fable. It put a lot of weight on your "it's politics" view from mid last week, but also put weight on other scenarios that pushed out the timelines.

@akarlin I think it's definitely not 'just' that.

@deanwball Frozen from releases but not from development, of course!

@luke_metro no the gov is holding back fable until Trump gets handed a large quantity of Anthropic stock for free

@deanwball It’s pretty frustrating and they need to resolve it asap

@TheZvi

@deanwball @Afinetheorem - In favour of faster: more compute applied to internal R&D - In favour of slower: uncertainty about whether non-national researchers are allowed to use/work on the models, unless this has been resolved?

@deanwball Damn does that mean no GPT-5.6 this week?

@S_OhEigeartaigh @Afinetheorem I overrated the latter concern in my initial analysis of this issue. I don’t think it’s a real issue though. Ironically it is a perfect illustration of how model-based regulation is dumb; there are many many models within Anthropic that are not technically “Mythos” or “fable”

@Miles_Brundage In some way, yes, but also this probably gives Anthropic more training compute to advance faster towards the next gen without losing investor confidence of being at the absolute frontier

@luke_metro It's just a matter of time until a Chinese model drops with the same capabilities as fable.

@luke_metro OpenAI bought an EZ Pass.

Anthropic did not.

The admin extracts its revenge and helps out its friends.

https://www.sfgate.com/tech/article/brockman-openai-top-trump-donor-21273419.php

@zephyr_z9 My internal plan is as follows

⬇️✅Details are as follows

@deanwball Gov: 'This chat model is not safe for the public. The public are irresponsible.' Also Gov: 'Where did I leave my gain of function virus bioweapon nuclear warhead? Has anybody seen it??'

is the gov just holding back Fable until OpenAI catches up

@deanwball I doubt OpenAI would have any difficulty releasing a GPT 5.6.

Mostly regardless of its capabilities. As long as they called it “5.6” and didn’t release blog posts about how dangerous it is.

It’s easy to underestimate how much Anthropic’s difficulties derived from comms.

@JacquesThibs @Miles_Brundage and dean is obviously in the know:

@zephyr_z9 My investment strategy.. ⬇️ ⬇️