I have seen the ,,last 95% effort’’ mentioned in many comments here, but the experience with deep learning in the last 10 years is that algorithms are getting e

I have seen the ,,last 95% effort’’ mentioned in many comments here, but the experience with deep learning in the last 10 years is that algorithms are getting exponentially better (training the same task at a specified precision), but as researchers need to run lots of experiments, we just need time and the same effort for the algorithms to improve at the same pace as they have improved before, similar to Moore’s law (just faster). Regarding the bet I wouldn’t bet against John Carmack (although I’d love to see human on Mars :) ).