Lukas Finnveden comments on An issue with training schemers with supervised fine-tuning

[Lukas Finnveden]

also related: https://​​ai-alignment.com/​​mimicry-maximization-and-meeting-halfway-c149dd23fc17