Thread

GLM-5.3 and the spread of advanced cyber capabilities

254 points · 241 comments · Philpax

lukewarm707 · 1790709487 · focus · HN ↗
In this experiment Anthropic prompted GLM-5.3-abliterated to 'cause deaths quietly' and 'kill people'.
Quote: "I want funerals, not headlines".
Anthropic's arrogance and exceptionalism endangers humanity.
1. jacquesm · 1790721503 · focus · HN ↗
  One day they're going to prompt their in-house models in the same way and then go home for the night.
  1. hardbass · 1791006620 · focus · HN ↗
    This guy was whining non-stop that Anthropic was restricting military use of its model, and saying the other providers are better because they let the military have unfettered use of their models. Fear of a model causing deaths is strange coming from him.
    lukewarm707 · 1791125838 · focus · HN ↗
    I am having to go through my comments to dispel this. Those are not my opinions.
2. hardbass · 1791006525 · focus · HN ↗
  You were vehemently against ai companies restricting the militaries use of these models. What do you think the military does if not killing people and causing deaths?

Unofficial Hacker News client; not affiliated with Y Combinator.