AI safety, and Anthropic in particular, have been a recurring topic in recent days. AI is the topic of the moment anyway. I see this as a follow-up to both the issue of the Department of Defense putting pressure on Anthropic and the post about Anthropic calling for AI regulations. In that post, many people assured me that Anthropic is not just doing this for regulatory capture, that they are ideological and really believe in it. I want to state here that my cynicism has won out.
Mx_Brightside on
As far as i can tell this is unrelated to the topic of the standoff with the Department of Defence, BTW, which the headline might lead some ~~lazy buggers~~ *distinguished users* who don’t read the article to believe:
> But in recent months the company decided to radically overhaul the RSP. That decision included scrapping the promise to not release AI models if Anthropic can’t guarantee proper risk mitigations in advance.
> “We felt that it wouldn’t actually help anyone for us to stop training AI models,” Anthropic’s chief science officer Jared Kaplan told TIME in an exclusive interview. “We didn’t really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead.”
> The new version of the policy, which TIME reviewed, includes commitments to be more transparent about the safety risks of AI, including making additional disclosures about how Anthropic’s own models fare in safety testing. It commits to matching or surpassing the safety efforts of competitors. And it promises to “delay” Anthropic’s AI development if leaders both consider Anthropic to be leader of the AI race and think the risks of catastrophe to be significant.
Ok_Barracuda_1161 on
In my opinion this is more related to things like Clawdbot (or openclaw or whatever the name is next week). It’s becoming more and more apparent with LLMs that the model itself has little control over safety and it’s more about the access it has and the systems built around it.
neverm0rezz on
Slightly unrelated but I’ve been very annoyed with Anthropics recent moralization over chinese models “distilling” their model, which imo has the serious implication that anything generated by claude belongs to Anthropic in some way. OpenAI is at least honest about what they’re trying to do
4 Comments
AI safety, and Anthropic in particular, have been a recurring topic in recent days. AI is the topic of the moment anyway. I see this as a follow-up to both the issue of the Department of Defense putting pressure on Anthropic and the post about Anthropic calling for AI regulations. In that post, many people assured me that Anthropic is not just doing this for regulatory capture, that they are ideological and really believe in it. I want to state here that my cynicism has won out.
As far as i can tell this is unrelated to the topic of the standoff with the Department of Defence, BTW, which the headline might lead some ~~lazy buggers~~ *distinguished users* who don’t read the article to believe:
> But in recent months the company decided to radically overhaul the RSP. That decision included scrapping the promise to not release AI models if Anthropic can’t guarantee proper risk mitigations in advance.
> “We felt that it wouldn’t actually help anyone for us to stop training AI models,” Anthropic’s chief science officer Jared Kaplan told TIME in an exclusive interview. “We didn’t really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead.”
> The new version of the policy, which TIME reviewed, includes commitments to be more transparent about the safety risks of AI, including making additional disclosures about how Anthropic’s own models fare in safety testing. It commits to matching or surpassing the safety efforts of competitors. And it promises to “delay” Anthropic’s AI development if leaders both consider Anthropic to be leader of the AI race and think the risks of catastrophe to be significant.
In my opinion this is more related to things like Clawdbot (or openclaw or whatever the name is next week). It’s becoming more and more apparent with LLMs that the model itself has little control over safety and it’s more about the access it has and the systems built around it.
Slightly unrelated but I’ve been very annoyed with Anthropics recent moralization over chinese models “distilling” their model, which imo has the serious implication that anything generated by claude belongs to Anthropic in some way. OpenAI is at least honest about what they’re trying to do