> there are signs that extremely powerful things are possible right now in the narrow machine-learning stuff, it’s not clear that those are the necessary steps

> there are signs that extremely powerful things are possible right now in the narrow machine-learning stuff, it’s not clear that those are the necessary steps to get all the way to artificial general intelligence This is my main takeaway from the interview, as it suits my beliefs. Most people seem to think that if we develop ML further we will go all the way to AGI, I think this is just mimicry step similar to how initial attempts to flight had flapping wings. I do think it is mandatory to explore in all directions but at this point this does not seem to be the one to lead all the way up to AGI.