Underfunded safety approaches

Nathan urges funding neglected, out‑of‑the‑box safety ideas like interpretability and self-other overlap research.

Play episode from 02:09:48

chevron_right

Transcript

chevron_right

Transcript

Episode notes

In this AMA-style episode, Nathan takes on listener questions about whether fine-tuning is really on the way out, what emergent misalignment and weird generalization results tell us, and how to think about continual learning. He talks candidly about how he’s personally preparing for AGI—from career choices and investing to what resilience steps he has and hasn’t taken. The discussion also covers timelines for job disruption, whether UBI becomes inevitable, how to talk to kids and “normal people” about AI, and which safety approaches are most neglected.

Sponsors:

Blitzy:

Blitzy is the autonomous code generation platform that ingests millions of lines of code to accelerate enterprise software development by up to 5x with premium, spec-driven output. Schedule a strategy session with their AI solutions consultants at https://blitzy.com

MongoDB:

Tired of database limitations and architectures that break when you scale? MongoDB is the database built for developers, by developers—ACID compliant, enterprise-ready, and fluent in AI—so you can start building faster at https://mongodb.com/build

Serval:

Serval uses AI-powered automations to cut IT help desk tickets by more than 50%, freeing your team from repetitive tasks like password resets and onboarding. Book your free pilot and guarantee 50% help desk automation by week four at https://serval.com/cognitive

Tasklet:

Tasklet is an AI agent that automates your work 24/7; just describe what you want in plain English and it gets the job done. Try it for free and use code COGREV for 50% off your first month at https://tasklet.ai

CHAPTERS:

(00:00) Ernie cancer update

(04:57) Is fine-tuning dead (Part 1)

(12:31) Sponsors: Blitzy | MongoDB

(14:57) Is fine-tuning dead (Part 2) (Part 1)

(26:56) Sponsors: Serval | Tasklet

(29:15) Is fine-tuning dead (Part 2) (Part 2)

(29:16) Continual learning cautions

(34:59) Talking to normal people

(39:30) Personal risk preparation

(49:59) Investing around AI safety

(01:00:39) Early childhood AI literacy