Respectability

Key Takeaways

this is a video I should probably have made a while ago but better late than never I've been talking about AI safety for quite a while now and in a very public way on the computer file Channel since about 2014 from the beginning a lot of AI safety advocacy has had a certain quality to it which to be honest I'm sure I've contributed to a little bit and it's true that might be the appropriate response to the problem but if you want to warn people about something it matters how you phrase things and how you frame things ideally the argument would stand on their own merits people should be able to look at the facts and decide for themselves and one of the things I'm trying to do with this channel is to give people the information they need to form their own informed opinions about this kind of thing but at the end of the day most people don't have the time to put in to get the required level of understanding and they shouldn't have to I mean it's not actually possible to study everything in enough detail to understand it there's just too much to know so we have experts but this kind of sucks because then how do you choose your experts it turns clear objective question questions about science and facts into these messy ambiguous questions about status and respectability and a lot of the time scientists don't want to lower themselves to playing that kind of game so they don't play at all or play half-heartedly but uh this is one of those games you can't win if you don't play and we do need to win there are problems with having to trust experts but ultimately specialization is how we manage to swap sucky problems like we keep being eaten by Predators for really cool problems like what if our technology becomes too powerful and solves our problems too effectively that's a good problem to have yeah sorry hin insects are one of the most successful groups of animals on Earth so a couple of years ago people had a go at making AI safety more respectable with this open letter which was often reported something like Steven Hawking Elon Musk and Bill Gates warn about artificial intelligence usually with a picture of the Terminator also surprisingly often Steven Hawkings why does this keep happening is he secretly a team of clones anyway the letter itself isn't very long and basically just says AI is advancing very rapidly and having more and more impact so we need to be thinking about ways to make sure that impact is beneficial it says we need to do more research on this and it links to a document that gives more detail about what that research might look like so the content of the letter itself isn't much to talk about but the core message I think is that this stuff isn't just science fiction and it's not just futurologists who are talking about it real serious people are concerned this was good for respectability with the general public because everyone's heard of these people and knows them to be respectable smart people but I've seen people who are slightly more sophisticated noticing that none of those people are AI researchers Professor Hawking is a physicist Bill Gates is a software developer but Microsoft at the time he was working there was never really an AI company doesn't count and Elon Musk though seriously overpowered is more of a business person and an engineer so why do these people know anything in particular about a but there are actually more than 8,000 signatures on this letter top of the list here is Stuart Russell he's actually currently working on AI safety and I plan to make some videos about some of his work later if you've never studied AI you may not know who he is but uh he's a pretty big name I'm not going to say he wrote the book on artificial intelligence but uh I I am going to imply it pretty heavily and of course his co-author on that book Peter norvi he's also on the list what's he up to these days director of research at Google the signatories of this open letter are not AI lightweights who else have we got Demis hassabis and just about everyone a deep mind uh Yan Lun head of AI at Facebook Michael wridge head of the computer science department at Oxford uh I mean I'm skipping over big people but uh yeah Tom Mitchell I know him from a little book I used as a PhD student you know this is an expert cuz the name of his book is just the name of the subject that's not a totally reliable heuristic though anyway my point in this video is just if you're talking to people about AI safety it's not cheating to say oh these high status respectable people agree with me but if you're going to do that pay attention to who you're talking to if it's someone who's heard of Russell or norvig they're likely to find that much more convincing than Elon Musk and Steven Hawking and don't use me for this I am just a guy on YouTube [Music] I just want to thank my amazing patreon supporters uh and in particular Ichiro DOI who skips the queue by sponsoring me $20 a month I don't even have a $20 a month reward level now I've got to make one that's a good problem to have anyway thank you so much you may have noticed that the gap between this video and the previous one is shorter than usual and it's largely thanks to my patreon supporters that I made ble to do that so thanks again and I'll see you next time