Hacker News Reader

Great points. That wasn't quite what I was shooting for. As a newbie to thinking about this in detail what jumps out is how well a similar set of techniques does generalise to an approximation of the right behaviour. It is still an approximation as you say.

The fun thing for me about the Deepmind paper is that they do use Stockfish to bootstrap a general approach from a specialist one. That is exciting - why is the but required in your post?

View on HN

The post tries to use this to make an analogy about self-driving cars. I think this analogy is deeply misplaced. It's both easier and a qualitatively different problem to bootstrap a neural network with a known good automated method like chess-playing than to solve a novel problem like car-driving in the real world (Which is not just much quantifiably harder, but also nebulously harder in some big "unknown unknowns" kinda ways)

A cursory glance through my posts would imply that maybe I just like big buts, and, I cannot lie, I often find myself having to play a contrarian in the credulous world of tech. However, the but is necessary here. We gotta stop giving that dude and people like him more money for just spewing grift. He's already captured enough of the world's infrastructure to make unilateral decisions that win or lose battles on distant warfronts on a whim

Reply on news.ycombinator.com