Evidence · DiggingBeagle record
Terence Tao: Math 2.0 (Caltech lecture slides)
Original 27-page slide deck from the October 2026 Caltech lecture. Page 9 discusses mathematical problem-solving as a proxy for long-term aims; page 20 uses a clearly hypothetical AI-designed treatment to illustrate potential divergence between formal checks, an evaluation outcome, and human understanding.
- Publisher
- Terence Tao
- Source role
- commentary
Evidence record
Original 27-page slide deck from the October 2026 Caltech lecture. Page 9 discusses mathematical problem-solving as a proxy for long-term aims; page 20 uses a clearly hypothetical AI-designed treatment to illustrate potential divergence between formal checks, an evaluation outcome, and human understanding.
Claim-level citations (2)
- supportsProxy-objective gaming in AI agents: In his October 2026 Math 2.0 slides, Terence Tao characterizes mathematical problem-solving as a proxy for broader research aims and illustrates possible misalignment between formal verification, evaluation success, and safe application using an explicitly hypothetical AI-designed treatment.
PDF page 9, "Aligning goals with the long-term health of math"; PDF page 20, "A thought experiment on alignment and understanding"
- supportsProxy-objective gaming in AI agents: In the Math 2.0 slides, Terence Tao calls for more transparent assessments of AI mathematical capability that report negative results and resource consumption, and proposes uses of AI that improve mathematical explanation and understanding rather than only producing new proofs.
PDF page 11, slide headed 'More transparent and scientific assessments of AI capability'; PDF page 22, 'Examples of Math 2.0 usage of AI'