Welcome to our next ‘Mini Minds’ installment!
As I mentioned before, we’re still figuring out what we want this format to be, so you can expect a bit of tinkering over the coming months. Our first mini was a short audio blogpost of sorts, and today’s mini is a mini interview. I chatted with Matt Crosby. He’s a postdoc at Imperial College London and has been spearheading a super cool project called the Animal AI Olympics. (If you recall, this is something that Marta Halina and I touched on briefly in our last episode, but it seemed intriguing enough to merit a longer look.) The basic idea behind the project, as Matt and I discuss, was to have a bunch of artificial agents—submitted by teams from around the world—compete in a gauntlet of tests. The wrinkle was that these weren’t the kinds of tests usually given to artificial systems. They were tests usually given to animals: some involved avoiding obstacles; others involved remembering locations; still others involved solving physical puzzles, like the celebrated “trap tube” task. This all sounds like good fun—and no doubt it was. But there were also deeper motivations for the project. Matt and I talk about those. We also talk a bit about how the contestants performed, about whether he was impressed, and about his team’s plans for the next iteration of the Olympics, to be held in 2021.
Enjoy the mini—and thanks for listening!