Post #2841653
2026-05-05 22:04 UTC
RE: https://mastodon.social/@craiggrannell/116523279778302330
Same. Hannah Fry is great but I took issue with the same two things:
1. Even if AI is “getting better fast,” that’s more scary, not less. See the claims about Mythos, which are terrifying (if true).
2. Personifying the agent, which I think is dangerous because it lets users shirk their responsibility for its actions. (And it’s easier for us to get fooled if we start thinking about agents and tools as human with human-level judgement and intelligence.)
Replies (1)
-
@craiggrannell@mastodon.social 2026-05-06 09:24
@json@micro.sadlerjw.com Yeah. The ‘better’ statement doesn’t really hold up. It’s becoming more capable, but the basic failures are just happening faster and there are more things that can go wrong. Personification is an ongoing problem, and that was clearly intentional by LLM/agent designers. They want you to get sucked in by the perceived ‘humanity’ of something that is a tool. But these things are no more alive than autocorrect. As for Fry, I do also worry about her AI ads if if that colours things.