PERSON DIRECTORY
Buck Shlegeris
Buck Shlegeris appears in 1 indexed conversation across The Cognitive Revolution. This directory brings every appearance, source, TL;DR, digest, and transcript into one searchable feed.
AI Control: Using Untrusted Systems Safely with Buck Shlegeris, Redwood Research (80,000 Hours Pod)
AI control offers a deployment thesis for extracting useful work from potentially misaligned systems without first making them trustworthy.Cheap monitors, suspicious-action replacement, resampling, and escalating audits can improve safety within roughly 5% budget or productivity constraints, though stronger scaffolds require organizational buy-in.The binding risk is execution: frontier labs may have a manageable list of controls but sharply lower appetite to implement them as capabilities accelerate.