Discussing the limits of artificial intelligence

throwing more data at problems won’t necessarily lead to progress... us to artificial general intelligence... there’s low-hanging fruit that deep learning’s good at  —  specific narrow tasks ... we’re not making progress on cognition and strong AI....
we can’t well interpret what deep learning is doing... it isn’t literally exactly memorization but it’s a little bit like that. ... if you build your system entirely with deep learning... and something goes wrong, it’s hard to know what’s going on and that makes it hard to debug...
a better paradigm, taking inspiration from cognitive psychology... algorithms perform better, using smaller amounts of data, often called sparse data... in most domains, there’s a long tail of things where there isn’t a lot of data.