Mastodon Feed
jonny@neuromatch.social ("jonny (nonvenomous)") wrote:
At the MIT museum. "Can we trust an AI pilot to fly in aerial combat" the text frames the problem as a technical challenge like any other. Predicting a high dimensional state space of pilot actions. An interactive exhibit, but broken. Unresponsive touchscreen. The voice of an AI assistant inviting to start the display chatters helplessly. The horror of algorithmic war needs tech support. We can't trust AI to run a museum exhibit, but we've already given it missiles.