If you know the basic strategy you can never lose a game of tic-tac-toe. With games like chess, the number of strategies fractals with the complexity. So far computers are resoundingly beating humans in those games, but they’re not solved. Increasingly buffed engines compete for supremacy at the game. If the game is every definitively solved, we’ll probably be able to explain how in a surprisingly simple manner.
It makes me wonder what other ‘games’ that take up a lot of human effort, and also supply a lot of human enjoyment may end up solved with advanced AI.
AI Won’t Take Your Job, But It Will Unbundle It
From Claude:
Novel Ideas to Add
1. Expand the Spectrum of Solvability
- Add examples between tic-tac-toe and chess: Connect 4, checkers (actually solved in 2007), Go (thought unsolvable until AlphaGo)
- Discuss why complexity doesn’t always prevent solving - sometimes it’s about computational power, not fundamental impossibility
2. The “Games” That Might Be Next
- Diagnosis: Medical diagnosis as a pattern-matching game with incomplete information
- Legal research: Finding precedents and building arguments as strategic moves
- Scientific hypothesis generation: Exploring possibility spaces systematically
- Software debugging: Searching state spaces for the error
- Financial forecasting: Not truly solvable, but increasingly “solved enough” for practical purposes
3. The Paradox of Solving
- When games are solved, do we stop playing them? (No - we still play tic-tac-toe with kids)
- The joy might be in the process not the uncertainty
- Or we create new constraints (speedrunning, style points, teaching others)
4. What Resists Solving?
- Activities with changing rules (social dynamics, fashion, culture)
- Games with infinite possibility spaces (creative writing, art)
- Activities where the goal itself is subjective or evolving