It sure looks to me like this new development is hailed as a leap, rather than incremental improvement. Although I'm judging just by reading the comments in the

It sure looks to me like this new development is hailed as a leap, rather than incremental improvement. Although I'm judging just by reading the comments in the HN bubble. Either way, I think the current "AI" is fundamentally limited by the statistical approach to problem solving. Without any reasoning capabilities, no amount of incremental improvements will change the fact that neural networks are simply making guesses based on existing data sets. Nothing magical or revolutionary, it's the same thing we've known for many decades.