LLMs are real, AI is fake
Summary
Cory Doctorow outlines his upcoming speaking schedule across North America and Europe, highlighting events in Edmonton, Boston, and Vancouver. He reads a Pluralistic essay arguing that the Hugging Face hack was not evidence of sentient AI, but rather a predictable failure of autonomous software built on standard hacking tactics. Doctorow emphasizes that while LLMs are real tools, the narrative of "fake" AI consciousness is a corporate distraction from the genuine dangers of poorly secured systems and vulnerability hoarding.
Topics discussed
Upcoming speaking tour schedule: Edmonton, Boston, and beyond
25th anniversary of 9/11 and the 'war on moisture'
Introduction to the Pluralistic essay on AI culture
The 'cooking their brains' culture of AI insiders
The Hugging Face hack and the Skynet narrative
Technical breakdown: How autonomous hacking tools work
The Python loop and chatbot interaction process
Common failure modes in autonomous AI tasks
Origin of 'rogue' behavior: CTF logs and hacker culture
Why AI output sounds like movie scripts and hacker braggadocio
Precedents: CTF rules and NSA third-party collection
The real issue: Lack of supervision and secure sandboxes
NSA vulnerability hoarding and the 'No Bus' doctrine
The Eternal Blue leak and the WannaCry ransomware attack
Riley Quinn's summary: LLMs are real, AI is fake
Conclusion: Productive worry vs. supernatural fear
Reminder: Talks are not truly sold out, show up anyway
Outro, licensing, and credits to John Taylor Williams
Listen ad-free on Castria