Somewhere in the bowels of X, Sam Altman indicated that to get to the edge of what you can learn from AI, ask for the hardest question it could reasonably answer. It gives back two answers – ones it can answer reasonably well given the availability of public information, and ones that are hard to answer. The questions it can reasonably answer based on available data are the 20’s version of the Wikipedia rabbit hole. From Perplexity:
• “Given what we know today, what are the most plausible scientific and philosophical accounts of consciousness, and what are the strongest arguments for and against each?”
• “If my goal is to maximize long‑term human well‑being over the next 200 years, what are the top 5 intervention areas, and how does current evidence support or undermine each?”
• “Explain the current best theories for unifying quantum mechanics and general relativity, and evaluate their main strengths and weaknesses for a non‑expert but technically curious reader.”
• “Design a step‑by‑step learning roadmap to go from high‑school math to being able to understand research‑level work on the Riemann Hypothesis, including concrete resources and milestones.”
• “Give a rigorous, detailed critique of capitalism and socialism as actually practiced, using historical data, and outline a concrete hybrid system you think would work better, with trade‑offs spelled out.”
