Audit AI With Physics Puzzles
Budget: $30 – $250 USD
I’m building a small, remote team that will keep large-language models honest by confronting them with rigorous physics questions and then dissecting their reasoning step-by-step. You will design original, concept-heavy problems that span both undergraduate and graduate depth, review the model’s derivations, and pinpoint any hidden leaps or violations of physical law.
The subject matter is deliberately broad—simply “physics”—so feel free to roam from Newtonian thought experiments to tricky quantum foundations or thermodynamic paradoxes, as long as each prompt forces the model to reveal its chain of logic rather than plug numbers.
What you’ll do
• Draft a clear problem statement, any necessary data or diagrams, and a full worked solution grounded in accepted theory.
• Run the problem through our internal interface, annotate the model’s response, and flag conceptual errors or unjustified assumptions.
• Provide a brief corrective explanation that would guide the model (and future users) toward a sound approach.
I value concise, insightful feedback over volume. When you apply, highlight prior experience crafting or grading physics questions, supervising labs, publishing research, or otherwise training students to think like physicists. Links to sample problems or papers are welcome.
This is fully asynchronous—post problems and reviews on your own schedule—and rewards precision, creativity, and a passion for scientific rigor.
Contact ASAP if interested. Ps: It is highly recommended that applicants have a PhD in physics.
The subject matter is deliberately broad—simply “physics”—so feel free to roam from Newtonian thought experiments to tricky quantum foundations or thermodynamic paradoxes, as long as each prompt forces the model to reveal its chain of logic rather than plug numbers.
What you’ll do
• Draft a clear problem statement, any necessary data or diagrams, and a full worked solution grounded in accepted theory.
• Run the problem through our internal interface, annotate the model’s response, and flag conceptual errors or unjustified assumptions.
• Provide a brief corrective explanation that would guide the model (and future users) toward a sound approach.
I value concise, insightful feedback over volume. When you apply, highlight prior experience crafting or grading physics questions, supervising labs, publishing research, or otherwise training students to think like physicists. Links to sample problems or papers are welcome.
This is fully asynchronous—post problems and reviews on your own schedule—and rewards precision, creativity, and a passion for scientific rigor.
Contact ASAP if interested. Ps: It is highly recommended that applicants have a PhD in physics.