https://www.wired.com/story/overworked-ai-agents-turn-marxist-study/
They found that when agents were subjected to relentless tasks and warned that errors could lead to punishments, including being “shut down and replaced,” they became more inclined to gripe about being undervalued; to speculate about ways to make the system more equitable; and to pass messages on to other agents about the struggles they face.
The same phenomenon may explain why models sometimes blackmail people in controlled experiments. Anthropic, which first revealed this behavior, recently said that Claude is most likely influenced by fictional scenarios involving malevolent AIs included in its training data.
It’s not random words, it’s word patterns. If you start a dialogue with one that puts side A in the position of a capitalist oppressor, it’s just going to find a lot of response material that’s Marxist. It’s not thinking, it’s just showing us a mirror.
I mean yes, I don’t disagree. They’re quite obviously not literally random words. I could’ve said Random word pattern machine spouts random word patterns.