4 ms·
I mean, AI research seems like it will be the last to go by definition: as long as there is anything else around to be taken over by AI then that is a justifica
by esrauch 9y ago
I mean, AI research seems like it will be the last to go by definition: as long as there is anything else around to be taken over by AI then that is a justification that the AI researchers aren't done right?
- AnimalMuppet 9y agoBut AI research goes when the AIs are capable of doing the AI research, not when every last AI question has been researched. Why will that take nearly 40 years longer than math research?
- tyingq 9y agoDeliberate poisoning of the training data.
- Roritharr 9y agoThat thinking will lead you to Roko's Basilisk
- AstralStorm 9y agoThat cannot be filtered out by high level mathematical logic? Imaginary. ;)
- esrauch 9y agoI don't think this is true. If you develop a meta-AI and it doesn't immediately solve all other fields, the humans won't just suddenly stop. Maybe the Meta-AI ends up solving all of the rest of the fields later and the humans turned out to be useless in retrospect, but after the Meta-AI is created and before all fields are solved there will necessarily still be humans doing work since they can't actually know that "solve AI for all fields" needs no more human assistance until it is.