@tonywalsh Well, it would provide the motivation to not cheat at the ML. But the socks would wear out faster unless we engineered better feet for the robots. ML for robot foot design... But that would be robot exploitation. All roads seem to lead to Skynet...
@mikamckinnon The Porter flight to Toronto from Newark is shorter than the NJ train ride to Newark. A cab to the ROM from Billy Bishop Island airport would be about $20CDN. Or you could take the shuttle to Union Station and take the subway North in the University line to Museum Station.
@Beetlenaut@spookysquid Especially if you train the nets with behaviour patterns from real world players. Untrained bots can be let loose for extents testing and ruining speedrunners fun.
@spookysquid@Beetlenaut I was thinking about scene detection, colour palate, motion vector heuristic analysis, overall loudness and stddeviation, natural language processing of dialogue for tone... And of course extracting score and sync music and running that through the actually-useful pref algos...