AISafety gains and risk-taking in AI races
A commenter argues that individual safety fixes aren't enough because people “reinvest” the gains in new risks—a pattern they also point to in models of AI races.
TLDR
A commenter argues that time pressure and competition leave people prone to poor decisions and groupthink in complex systems such as aviation, nuclear safety and medicine. They say individual safety improvements aren't enough in those settings or in models of AI races because people “reinvest” the gains in riskier behavior, using seatbelts followed by faster driving as an example.
Combined views
141
1 Source, first seen 1d ago
Safety gains and risk-taking in AI races
A commenter argues that individual safety fixes aren't enough because people “reinvest” the gains in new risks—a pattern they also point to in models of AI races.
TLDR
A commenter argues that time pressure and competition leave people prone to poor decisions and groupthink in complex systems such as aviation, nuclear safety and medicine. They say individual safety improvements aren't enough in those settings or in models of AI races because people “reinvest” the gains in riskier behavior, using seatbelts followed by faster driving as an example.