Amanda Askell Says Claude Could Replace Her Role

The Observer reports that Anthropic philosopher Amanda Askell, speaking at the Bloomberg Tech Summit, said: "Eventually, Claude is going to be a much better philosopher than I am, and probably be much better at every aspect of my job than I am." She added, per the Observer, "Human input is going to be rarer and rarer. That's the thing that we need to prepare models for." According to the Observer, Askell joined Anthropic in 2021 after earlier safety and alignment work at OpenAI, co-authored what the article calls Claude's "soul doc," leads Anthropic's personality alignment team, and was named to the TIME 100 AI list in 2024.

Askell's remarks are notable less for any announcement than for who is making them: a senior alignment researcher at a frontier lab publicly framing advanced models as candidates to outperform experts at qualitative, judgment-heavy work, and arguing that the field should prepare models for a future with less human input. The comments feed ongoing debates about how labs will supervise and evaluate systems whose outputs are increasingly hard for humans to check.

As a generic industry pattern rather than a statement of Anthropic's specific plans, growing model competence at nuanced reasoning tends to turn human-rater evaluation into a scaling bottleneck. Teams working on alignment and evaluation increasingly lean on automated and hybrid evaluation suites, adversarial and long-horizon testing, interpretability, and scalable-oversight techniques to assess behavior beyond single-label accuracy. Askell's "human input ... rarer" framing aligns with that broader shift from human-in-the-loop checking toward scalable oversight, though how to do it reliably remains unsettled.

• •Whether labs publish reproducible evaluation suites for qualitative and value-laden tasks.
• •Adoption of automated or hybrid human-machine evaluation pipelines in production.
• •Research connecting capability gains to specific alignment and safety failure modes.