@vvvincent_c@KaivuHariharan@polynoamial So call it 1000 attempts, and not just $200 per problem, that's the cost of the proof once you know you can solve it/have a solution. So $1M budget?
@allTheYud@fchollet Seems like this disagreement is about whether to classify raw compute (and time) as intelligence. We could have <50% algorithmic gains to be made in Go but eg. require 10,000x more compute to get anywhere near optimal play
@ptrschmdtnlsn For a rigid beam with a single spring at the wall, deflection is proportional to L^2 - one for torque, one just because at a fixed angle x ~ L. Then in the beam case you get more springs in series the longer L is, making L^3