Quanto dovremmo temere l’AI?
Summary
This Wilson episode explores the growing AI safety crisis, triggered by researcher Jacob Coxon's resignation from Anthropic and recent incidents involving autonomous AI agents. Host Francesco Costa and guest Enrico Mensa, a University of Turin AI researcher, debunk common misconceptions by explaining how Large Language Models actually function as text predictors. They analyze the conflicting narratives from industry leaders like Dario Amodei and Sam Altman, who now urge a slowdown despite years of aggressive development. The discussion also addresses geopolitical tensions with China and the…
Topics discussed
Intro: The 'Will We All Die?' theme and media alarmism
Jacob Coxon's resignation from Anthropic and his warning
Summer incidents: AI agents communicating and hacking
Industry leaders call for a pause; Trump's China angle
Why the pause now? Opportunism, costs, and market control
The irony: OpenAI and Anthropic founded on safety fears
How to form an opinion: Moving beyond tribalism and hype
Episode roadmap and introduction of guest Enrico Mensa
Defining AI: From 70-year history to recommender systems
How LLMs work: Machine learning and text prediction
Outro: Preview of upcoming topics and subscription pitch
Listen ad-free on Castria