 |
Bounce! Conversations with Larry WeeksAuthor: Larry Weeks
A podcast about the challenges of being human (and running things). Host Larry Weeks interviews thought leaders across psychology, business, science, and culture to explore resilience, emotions, mental health, and practical ways to stay grounded through change while building purpose and real happiness. Language: en Genres: Education, Health & Fitness, Mental Health, Self-Improvement Contact email: Get it Feed URL: Get it iTunes ID: Get it |
Listen Now...
Beyond AI Doom: Zack Korman on the Case for Cautious Acceleration
Monday, 26 October, 2026
Is AI doom a prophecy or a set of choices we can make right now? What happens if the people building AI start believing it has a soul? And if AI is risky, why might the doomers be wrong about how to respond? In this episode, Zack Korman explains why AI doom isn't inevitable and why the real risks have less to do with machines plotting against us and more to do with what access we give them and whether anyone is watching. Zack is the co-founder of Embroidery.io, a cybersecurity company that uses AI to monitor AI agents. He was previously CTO of the cybersecurity company Pistachio. Based in Norway, Zack has become an outspoken voice in the AI safety debate, including media appearances in The Atlantic, Time magazine, and The Wall Street Journal. In this conversation, we explore how AI agents actually work, what really happened in the Hugging Face incident, the divide between doomers and accelerationists, why believing AI is alive could make it harder to shut down, why Zack treats AI models as extraordinarily valuable malware, how to regulate the way AI is deployed rather than just the labs that build it, how to protect yourself in the age of agents, and how becoming a father changed the way he thinks about the future. One idea kept resurfacing: whether AI goes well or badly may depend less on what the machines want and more on what we build, what we connect them to, and whether anyone is watching Some highlights from the episode: Between extraordinary promise and existential fear Doomers, accelerationists, and centrists: The competing philosophies of AI Can we move fast without breaking things? The asymmetry of AI risks and rewards How language models become agents that can take real-world actions Why even a well-aligned AI model can't be perfectly safe The problem of persistence Kill switches, sandbox escapes, and human oversight AI consciousness, and the effects of anthropomorphism What effective AI regulation might actually look like Interpretability: Understanding a model's choices What AI doomers get right Practical advice for agentic AI risks Potential upsides of AI A new father's perspective on AI risk Enjoy! For show notes and more, visit larryweeks.com
