Part 3/9:
The speaker commenced their PhD in 2012 with an ambition to cultivate AI that could autonomously master the intricacies of poker. Unlike many assume, poker transcends mere chance; it requires strategic thinking akin to chess but with additional complexity. Early in their research, conventional wisdom within the academic community indicated that merely scaling models would suffice to yield improvements. By 2015, the models had advanced to the extent that they could challenge top human players in a rigorous 880,000-hand poker competition.