Anthropic Urges Global Pause in AI Development, Flags 'Self-Improvement' Risk

wsj.com

23 points·joebuckwilliams··7 comments

Hopefully it can self improve to not writing a 40 page essay that restates the initial question and then you have to spend another $.50 asking it to use a web search instead of just yapping.

The models 3 months ago were way more scary for society because they actually worked and were helpful. Now they maximize token spend on basic tasks and get into recursive loops of forgetfulness and complexity.

Yeah, that's not gonna happen. I'm sure Anthropic isn't pausing their AI development, why do they expect anyone else to?

A US tech company is probably not trying to limit themselves by arming government to declare their product a munition but they do want to limit Chinese access to technology and markets.

Also in this post from a few minutes earlier, a copy-paste of the article text: https://news.ycombinator.com/item?id=48403770

[dupe] Discussion on source: https://news.ycombinator.com/item?id=48400842

Does this mean that Anthropic is having trouble raising cash so they want all of their competitors to pause for a while and give them a chance to catch up?

Opus 4.8's critical assessment of Anthropic's "When AI builds itself" [0][1]. Because, why not?

[0] https://pastebin.com/Vc5Yq9Ai [1] https://www.anthropic.com/institute/recursive-self-improveme...

Reply on news.ycombinator.com