I (almost) entirely agree with this. A lot, but not all of the narrative around AI is hype. Caveat: I’m not an AI specialist but like a lot of cyber experts I get annoyed with AI hype distracting people from fixing long-standing problems with the tech we already have. Anyway. A few thoughts: 🧵 1/ So much of the recent hype is around Gen AI. GenAI has been around long enough now for us all to see that it’s (a) a glorified and very, very clever and fluent search engine plus (b) a generator of content which, in the memorable phrase of @meredithmeredith.bsky.social, is “plausible nonsense” 2/ But that’s all Gen AI basically amounts to. So then you come to AGI. What is it really, and is there any sign of it? I think Ms Bender is right to be sceptical here. So far as I can see AGI is people claiming “in x months/years a database is going to be smarter than Einstein & that’s about it 3/ Then you come to AI doing tasks. Here again she is right I think. AI automates lots of lower skilled jobs & also some skilled things like basic translation. But beyond that? My friend Prof Mike Wooldridge, prof of foundational AI at Oxford has a really good test - can a robot do a plumbing job? 4/ He tells me that AI is miles away from competence in plumbing, if it ever gets there. An AI robot plumber is currently worse than useless - its total inability to adapt & respond to the physical human environment means it is far more likely to flood the place than fix the plumbing problem 5/ Every so often, the likes of Elon Musk display a robot butler serving tea or champagne or something. But these appear to be very scripted & tightly controlled demos which cost a fortune - we aren’t close to these things being practically commercial and they can’t really respond dynamically 6/ So there is a huge amount of hype. Ms Bender’s phrase “a fancy wrapper around some spreadsheets” is glorious. But where she even outsceptics me - & I think goes too far - is she doesn’t (in this interview anyway) cite some of the ways these fancy spreadsheets could be genuinely transformative 7/ Healthcare is the obvious and most exciting example where these incredible “spreadsheet” capabilities to analyse patterns in data way beyond human capabilities offers huge potential for progress. There are also more boring but worthy uses of this type of capability Even I am excited about this 8/ But mostly my personal view, for what it’s worth, is that it’s one big bag of hype. My favourite visual depiction of the folly of our age is below, drawn by the brilliant Tom Fishburne This sums it all up better than I can. 9/END
A better quality version of the cartoon