3 ms·
I think it's sleight of hand in that "moral self-correction" is a very complicated phenomenon, and proving that a computing system is doing it would require an
by wpietri 4y ago
I think it's sleight of hand in that "moral self-correction" is a very complicated phenomenon, and proving that a computing system is doing it would require an incredible amount of detailed theoretical and empirical work. Some of which, yes, might include much more careful definitions of morality. Until that work is done, I think it's somewhere between foolish and negligent to anthropomorphize LLMs.
- JieJie 4y agoHumans gave sailboats a gender. Anthropomorphization is our default.
- wpietri 4y agoSure, and when there's no possibility of confusion, I'm all for it. It can be lovely and poetic. Here, though, I think it's dangerous.
- JieJie 4y agoI agree with you that there are outcomes that would be less ideal. If it helps, I refer to them as Intelligent Tools. I do prefer the "tool" metaphor (and so does Bing Chat) and I hope that companies like Microsoft rethink their "copilot" and "assistant" metaphors. I don't think they're "dangerous" per se, I think metaphors matter and we should choose the best ones.