3 ms·
Why Effective Altruists Fear the AI Apocalypse
- mike_hearn 4y agoThe interviewee says: Machine-learning experts say that there is a greater than 50 percent chance of human-level artificial intelligence in 37 years and that the chance of catastrophe as bad as extinction would be like 5 percent conditional on that I'm curious which experts say this, as it sounds like the sort of prediction that has damaged the reputation of expert forecasting so much. I suspect that maybe there actually aren't any. If there are then where did these numbers come from, why are they so bizarrely precise yet also so vague, why doesn't the unfalsifiable nature of the prediction bother them, what does "human level" even mean etc. You can also look at the trend in computing power over time, where the largest machine-learning models at the moment have the computing power of about an insect brain. Given what we know from neuroscience, it seems like in about ten years AI models will have computing power on the order of a human brain. Deep neural nets are somewhat inspired by brains, but it seems like a stretch to try and make direct comparisons between the "computing power" of machines designed for matrix multiplication vs the brain. If we just say, “Oh, there’s a fair chance of this happening” or “a significant chance,” that’s extremely vague. Whereas if I say, “Look, I give it 5 percent credence,” then at least I’m making my beliefs about the world clear. In this context saying "I give it a 5% chance" seems pretty similar to saying "a slim chance". Assigning predicted probabilities to events that are almost by definition one-offs sounds scientific but you can't do an experiment to determine how likely it actually is, so a prediction about it with a number in isn't any more testable than using English words.