Why does artificial intelligence play some video games well, and others poorly?
with Adam Stooke
Friday, July 20th, 2018 at 6:30 PM at Scarlet City Espresso Bar
Scientists use video games to build better artificial intelligence
In the past few years, deep neural networks have burst onto the scene of artificial intelligence. Scientists commonly test these networks by teaching them to play video games, helping them develop better algorithms to accomplish more serious tasks. Adam will describe why neural networks are often able to outperform humans in many Atari video games, even when the network can only use screen images as input. He’ll also explore what challenges lay ahead in the field of artificial intelligence.
Adam Stooke is a PhD student working in computer science, researching learning algorithms for robotics. His recent work has focused on harnessing parallel computing resources, including GPUs, to run simulation-based experiments faster and at larger scale.