4 ms·
AI safety is a genuinely hard problem.
by jsnider3 1y ago
AI safety is a genuinely hard problem.
- moralestapia 1y agoIndeed. I just can't wrap my head about what the actual product/service is. Let alone something that could be sold for billions. "Safe AI" is very ambiguous in terms of product.
- jsnider3 1y agoIf you have a Safe AI, then becoming a billionaire is being an underachiever.
- moralestapia 1y agoSure, but again, define "Safe AI" in terms of a product. What exactly am I buying? How much I'm paying for it? That's the thing I don't see. Is it a model? `gpt-3.5-turbo-safe`?
- kbelder 1y agoWouldn't all the money go to the unsafe AI, since it does more?
- jsnider3 1y agoIf someone invents an unsafe AI capable of making a billion dollars, then we will probably all die, which is why we should make safe AI instead.