AI safety, regulation and competitive pressure
One thread argues that even a hypothetical perfect safety recipe for one level of AI would leave companies tempted to cut corners as they compete to build the next.
TLDR
A thread criticizes opponents of regulation for framing AI safety as purely technical, arguing that this makes safety seem easily solvable and industry harder to criticize. The author offers a hypothetical: even with a perfect recipe for making Astra/Fable-level models safe, secure and aligned, companies would still pursue the next level and be tempted to cut corners because of competition. The author stresses that no such perfect recipe currently exists.
Combined views
486
1 Source, first seen ago