Ex-Meta researcher rbhar90: anthropomorphizing Claude is dangerous, it's AlphaGo-like RL not sorcery
rbhar90 · x · 2026-10-17
AI researcher rbhar90 (ex-Meta) offers a rare critique: Claude is essentially a large transformer trained via pre/mid/post-training RL in "game worlds" mimicking useful real-world situations — fundamentally not that different from AlphaGo, except its scale lets it play many games, including "maximizing human response." He argues treating it as sorcerous is dangerous, suspects some Anthropic staff genuinely believe they're building a sentient being, and says secrecy and cloak-and-dagger harm society — open scientific enquiry reduces suffering.
More from AGI Musings
- Debate: utilitarianism doesn't justify lawbreaking — bounded agents and AI x-risk 'bullet-biters' — NathanpmYoung · 2026-10-17
- Paper: AI agent design centers the agent, undermining human oversight — math_rachel · 2026-10-17
- AI and jobs: how big-company departments grow without adding headcount — vasuman · 2026-10-17
- Researcher slams OpenAI and Anthropic as grifters who invented nothing of significance — examachine · 2026-10-17
- Bill Gates calls AI 'the most dangerous technology' ever, drawing pushback from AI researchers — flowersslop · 2026-10-17
- Anthropic publishes internal employee survey on the AI expectations gap — ResultBackground2450 · 2026-10-17