Graycen and Alison talk about artificial intelligence playing Starcraft 2 (the trilogy), artificial intelligence learning handwriting, early career failure, and nonradioactive metals. No interview this week, but it’s somehow just as long as usual. Listen here or on Apple Podcasts!
StarCraft is beloved to many but rather mysterious to your friendly podcast hosts. What we lack in knowledge, we attempt to make up for in enthusiasm. But you know where you can find both knowledge and enthusiasm in excess? At DeepMind, where Oriol Vinyals and colleagues built an artificial intelligence agent that’s pretty good at StarCraft! Special thanks to Andrew Morgenthaler, who recommended this story.
- Research paper: Oriol Vinyals et al., Nature
- Article: Dan Garisto, Nature
- DeepMind press release: The AlphaStar Team, DeepMind.com
- Wikipedia article for StarCraft
- StarCraft Wiki
Failure is kind of the opposite of StarCraft — beloved to no one and very familiar to your friendly podcast hosts. But it turns out that early career failure might actually be helpful for early career scientists, according to a study by Yang Wang and colleagues. Even more special thanks to Andrew Morgenthaler, who recommended this story to us, possibly because he knew how dumb we’d feel after trying to explain StarCraft.
Nonradioactive metals are extremely difficult to find ever since we started introducing lots radioactivity to the atmosphere in 1945. But scientists need these low-background metals for experiments to detect tiny amounts of energy, so we have to find it somewhere.
- Article: Robin George Andrews, The Atlantic
An artificial learning agent is being trained to use a very specific kind of mindreading that could help people write using only their brains.