[felixgallo]

Use nuance and judgement, friend. Anthropic notably pushed back on completely autonomous no-human-in-the-loop drone killings and mass surveillance of the US population, where others like OpenAI scrambled to agree. Anthropic isn't perfect but that doesn't make them equally bad.

[noelsusman]

I didn't say or even imply that OpenAI and Anthropic are equally bad on this front. It's just not accurate to say Anthropic has issues with the US military using Claude to commit war crimes. They don't.

[felixgallo]

They literally do -- see above, where the red line they refused to cross involved fully autonomous kill bots (which would be a war crime), and for which they were branded a supply chain risk, thrown out of Pentagon contracts, and now enjoined from releasing their product.

You can not like that Claude was involved in the planning that led to the murder of a bunch of schoolgirls, but stop playing pretend.

[noelsusman]

It's just a fact that Claude has already been used to commit war crimes and Anthropic has had zero issue with it. I don't know what else there is to say about it.

Also, not that it really matters but building a fully autonomous kill bot is not actually a war crime.

[felixgallo]

Claude has been involved in the planning of a war crime, for sure. But 'Anthropic has had zero issue with it' is an unfounded assertion that you are making up.

Fielding a fully autonomous kill bot is absolutely a war crime under the IHL; specifically, there is no current universe under which a bot can reliably tell a combatant from a non-combatant, or a civilian target from a military target, or act in a proportional manner.

[firesteelrain]

Trust no one, friend. Believe what you want to believe.