Štěpán Los comments on Race to the Top: Benchmarks for AI Safety