AGI benchmark indicates whether a future AI model could cause catastrophic harm livescience.com 1 points by gmays 8 hours ago