Can AI make advancements in moral philosophy by writing proofs?

In a thought-provoking experiment, researcher Michael Dickens tested whether current AI systems like Anthropic's Claude 3.5 can contribute to moral philosophy by generating formal proofs. The test produced three novel impossibility proofs, including one showing that introducing moral uncertainty cannot escape Arrhenius' famous population ethics impossibility result. While these proofs aren't mathematically complex—comparable to undergraduate-level work—they demonstrate AI's ability to systematically explore philosophical axioms and produce verifiable arguments.

Dickens argues that philosophy represents the "set of problems left over" after well-understood problems become formalized fields like logic or physics. Since AI struggles with ill-defined problems but excels in formal systems, philosophical proofs represent a measurable frontier where AI could contribute meaningfully. The experiment suggests that even non-superintelligent AI could help philosophers by rapidly exploring logical consequences of different axiom sets, potentially uncovering contradictions or implications humans might miss.

The key insight is that AI doesn't need to solve philosophical debates to advance the field—it can generate proofs that force clearer thinking. As Dickens notes, "Having that sort of proof doesn't tell you whether you ought to reject one of the axioms or accept the conclusion, but it does tell you that you have to do one of those things." This positions AI as a discovery tool rather than an arbiter of truth in philosophical discourse.