AI firms have been speaking about superintelligent AI like it’s inevitable, however current safety incidents like OpenAI’s Hugging Face breach are demonstrating the potential risks of deploying AI programs which can be extra succesful than people. So what occurs after we can’t reliably management what these programs do?
On this episode of TechCrunch’s Equity podcast, Rebecca Bellan is joined by Connor Leahy, an AI researcher, entrepreneur, and now the U.S. Government Director of nonprofit ControlAI, which is pushing for a much more radical strategy to AI security: stopping firms from growing superintelligence altogether. Leahy explains why he thinks the dangers have develop into too nice to handle by means of alignment and containment alone, and the way what sounded far-fetched six months in the past is now being backed by a wave of recent laws.
Subscribe to Fairness on YouTube, Apple Podcasts, Overcast, Spotify and all of the casts. You can also comply with Fairness on X and Threads, at @EquityPod.