George Hotz: Hacking the Simulation & Learning to Drive with Neural Nets | Lex Fridman Podcast #132

Congratulations, you made me come out of retirement purely to disagree with you. No chain more richly deserves death than BTC and BCH is one of the best chains in existence. I'm prepared to sign this message with a Bitcoin key from 2011 if you doubt the fact I've been active that long. It's the unfortunate and poorly informed post 2017 split BTC adherents who have no idea what's going on. I watched the entire thing unfold first hand and the BCH camp simply has it correct. The fact that the market is largely completely ignorant of this is one of the most glaring indications of just how irrational it presently is. Until BTC dies, the entire cryptosphere is little more than a joke. Witness the tone very technically proficient but not "in the cult" people like George Hotz have when they're discussing the block size debate; there's no doubt whatsoever that core is utterly wrong, to the point that any suggestion to the contrary is nothing more than a joke. Deep link to the exact part in a recent interview he addresses this [this video]

It was a discussion about how such a system must function with all the different inputs vs the vision only model. If you have radar and lidar and vision, then you have at least three different specialist machine learning models running, and then another model running that takes their outputs and decides what the car is going to do. You may have even more than that, some doing specific tasks like localization. Neural nets and vision only is a more difficult but in the long run straightforward solution. The example he brought up was alphazero vs some other chess engine that has a rook engine, and a knight engine, etc. Basically he's backing the end to end neural network back approach over some kind of multi-sensor fusion. [this video]