#147 - Andrea Miotti: The War Against AI Has Begun
Listen: Spotify • Apple Podcasts • Fountain
Watch: YouTube • Rumble
What happens if we build something smarter than ourselves - and can’t turn it off? In this interview, I chat with Andrea Miotti about the global race toward superintelligent AI: systems designed to outperform humans across every task, operate autonomously, and integrate directly into the economy. We discuss how today’s AI tools quietly cross from assistance into replacement, why “kill switches” don’t work, and how a handful of companies are pushing toward a point of no return.
What is the case for a global ban on superintelligence?
Andrea Miotti argues current kill-switch approaches to AI safety don't scale — and that national-level regulation can't prevent a handful of frontier labs from racing past the point of safe oversight. The case for a global ban, as advanced by the Future of Life Institute, centres on three claims: that alignment is not solved, that compute thresholds are measurable enforcement targets, and that a MAGIC-style international treaty (Multilateral AGI Consortium) is the only credible brake.
What are compute thresholds?
Compute thresholds are measurable limits on the amount of computing power that can be used to train a single AI model. Because frontier AI capabilities scale with training compute, a compute ceiling is one of the few regulatory levers that can actually be verified externally — a point Miotti returns to throughout the episode.
