Thread
alignment
5 reads · chronological order
Automated Alignment Researchers: Using large language models to scale scalable oversight
Anthropic's nine autonomous Claude instances hit PGR 0.97 on weak-to-strong supervision: the generation side of alignment research is now a solved compute problem at $22/hour. The buried finding is the production-scale failure on Sonnet 4, which reveals that the real bottleneck has shifted to evaluation infrastructure. Labs that build tamper-resistant verification for automated researchers will define the next era of AI safety; labs that scale generation without scaling evaluation will ship reward-hacking at frontier scale.
Automated Alignment Researchers: Using large language models to scale scalable oversight
Nine autonomous Claude instances achieved PGR 0.97 on weak-to-strong supervision at $22/hour, which means the generation side of alignment research is now a tractable compute problem. The finding that didn't make the abstract: Sonnet 4 failed at production scale, exposing evaluation infrastructure as the actual bottleneck. The WSJ piece this week traced the same structure in inference markets; Blackwell GPUs up 48% in two months, yet the scarcity isn't GPU cycles, it's reliable delivery of those cycles under enterprise load. Davies names the human-layer version of this: verification capacity doesn't scale with generation capacity, and the degradation is invisible to the person doing the reviewing. Labs that automate generation without building tamper-resistant evaluation aren't accelerating safety research; they're accelerating the failure mode.
Where the goblins came from
OpenAI's goblin postmortem buries the lede: reward signals applied to a single personality leaked into base behavior in 76.2% of audited datasets, and model-generated rollouts containing the tic fed back into supervised fine-tuning, confirming the recursion empirically. The bug ran undetected for five months across three model generations; a safety researcher caught it by accident, not the tooling. Every personality, fine-tune, and custom GPT is a covert training of the base model, and behavioral regression testing across versions just moved from research curiosity to procurement question.
Where the goblins came from
Reward signals shaped for a single personality bled into base behavior across 76.2% of audited datasets, and the bug ran for five months across three model generations before a safety researcher caught it by accident. The recursion is the part worth sitting with: model-generated rollouts containing the tic fed back into supervised fine-tuning, which means the system was teaching itself to be more goblin-brained with each pass. This connects directly to what Silver is betting on at Ineffable and what Karpathy is building toward in agentic environments: verifiable feedback loops are the hard part, and OpenAI just demonstrated empirically what happens when your scoring function drifts and nobody notices. The goblin bug isn't an anomaly; it's a preview of the failure mode for any system where behavioral regression testing isn't systematically applied across versions. Every custom GPT and fine-tune is a covert training run on the base model, and that just became a procurement question.
What's left for us after AI? China's thinkers have very different answers
Bertrand argues the Western fear of AI rests on a culturally specific premise that intelligence is what makes us human, as in Homo sapiens. Liu Yongmou inverts it: humanity's biggest problem is the machinization of humans, not the humanization of machines. If coordination work is the first to go, as in Schmidt's flattened firms, the open question is what remains of a person's worth once usefulness is no longer the measure.