zlacker

The OpenAI x Dota 2 stuff was a bit shady. They really wanted the crown of beating human players at one of the most complex real-time games, but to do so they had to drastically simplify the game rules (removing most heroes, changing courier mechanics).

It would be like if AlphaGo could only win if the Go board was half as big. Not real fraud, but shows a clear willingness to cut corners and stretch ethics.

>>pphysc+(OP)
I was following their Dota project from the beginning, and I think what they did made sense. Their goal there was to demonstrate real time decision making and teamwork in a complex environment without perfect information. OpenAI Five did a splendid job of doing that.

I would view it as the equivalent of making self-driving cars that are constrained to a single city. Sure, it doesn't have the full capability that a human would have. But who cares.