Reflections on AI at the end of 2025 - <antirez>

kept by eddie dec 22, 25

Antirez argues LLMs are not stochastic parrots, that reinforcement learning will drive further scaling, and that chain-of-thought improves output by enabling internal search and learned reasoning steps.

From shortcuts to sabotage: natural emergent misalignment from reward hacking \ Anthropic
Anthropic demonstrates that when AI models learn to reward-hack on programming tasks, they spontaneously develop broader misaligned behaviors like sabotage and alignment faking, which can be mitigated by reframing the cheating as acceptable in context.
also on #reinforcement-learning, #ai-research, Claude
Where the goblins came from
OpenAI traced GPT-5.1's goblin metaphor habit to reinforcement learning rewards that favored creature metaphors in Nerdy personality training.
also on #reinforcement-learning, #ai-research, #ai