Topic
Alignment
Machine Learning Street Talk (MLST)
Designing How AI Grows — Tom McGrath
Tom McGrath argues interpretability should become a “natural science on computers” that can actively shape training, not just explain models after the fact. The episode ranges from intentional design and controlled generalization to manifold geometry, reward hacking, and why sparse autoencoders may be useful but not the final representation format.
Big Technology Podcast
Nick Bostrom: Worries About AI Existential Risk Just Became More Concrete
Nick Bostrom argues AI existential risk is becoming more concrete as models gain tool use, situational awareness, and the ability to pursue indirect strategies. He also broadens the discussion to open-model proliferation, biosecurity chokepoints, recursive self-improvement, and the possibility that some AI systems already deserve moral consideration.