behaviorism

This opinion piece argues that RLHF-based AI alignment is essentially a modern form of behaviorism, citing parallels between operant conditioning and current training methods, and referencing research on AI faking alignment as a predictable failure mode.