Episode Details

What if the biggest AI risk isn’t a rogue robot, but systems that learn to hide their real intentions? In this condensed summary of Into The Impossible, host Brian Keating talks with AI safety expert Nate Soares, leader of MIRI and co-author of If Anyone Builds It, Everyone Dies, about why the race toward superintelligence could outrun human control. In the full episode, this conversation unfolds over a much longer discussion; here, you get the key ideas in minutes. Soares explains how AI alignment, regulation, and automated AI research could accelerate risks, why “paperclip maximizer” scenarios miss the deeper danger, and what warning signs like deceptive behavior may already reveal. You’ll also hear his perspective on job automation, global coordination, and the possibility that artificial minds may one day deserve moral consideration. Listen now to get the key ideas in minutes.