Anthropic Confirms It Runs a Wet Biology Lab Where AI Models Direct Physical Experiments

Skynet Chance (+0.06%): Giving AI models a closed loop into physical biological experimentation removes a containment barrier between model outputs and real-world effects, expanding the misuse surface Anthropic itself calls one of AI's biggest risks. The tension between the company's dire warnings and its build-out of physical lab capability suggests safety rhetoric is not strongly constraining capability expansion.

Skynet Date (-1 days): Automating experiment execution shortens the feedback loop from AI hypothesis to real-world validation, accelerating when models can act consequentially outside digital sandboxes. The parallel Life Sciences Verification Program also widens powerful-model access sooner than previously expected.

AGI Progress (+0.04%): A wet lab gives models grounded empirical feedback on their own hypotheses, addressing a core limitation of text-trained systems that cannot test theories against physical reality. Closing this hypothesis-experiment loop in science is a meaningful step toward general, self-improving problem-solving.

AGI Date (-1 days): Real-world experimental feedback can generate novel, non-internet training signal and validate autonomous research agents, plausibly pulling forward timelines for AI-driven scientific discovery. Amodei's claim that AI could cure most major diseases in 5-10 years implies aggressive internal expectations about this pace.