Quote:
Maybe I'm biased because I have friends who work at Tower, but Sam seems like a pretty cool guy, and I hope his research and PhD is successful.
But the professor on the other hand, just wow. I honestly don't know where to begin. When WCG was offering insight into certain spots where the bot was not playing well, he refused to even listen, he would immediately try to defend that Claudico was perfect. I get that he's really prideful in this bot and Tartanian7
Same thoughts here.
Sam was great on the stream, answered the questions straight and is genuinely interested in what is going on and what humans think about the bot as it's clear as sky it's not perfect and there is something to be learnt.
The professors though sounds like bigger and bigger bs'er with every appearance he makes.
Quote:
, but isn't it possible that the simplifications made by the abstraction algorithm allow for error (or at least, exploitability)?
Again, if they knew what they are doing they would measure it or at least approximate it. You can't optimize for something you can't measure. When there is a claim about Nash equilibrium or optimal play then "give me the number" follows. If the numbers are not given the claim is worthless. It's as simple as that really.
You can have all your academic papers and all your magical theories, just write some code to measure the very thing you try to optimize for.