Now this was fun

Project information

The culmination of a semester's worth of knowledge-based AI methods working on a single goal: perform better than most children on a standardized intelligence test.
The agent itself uses a mixture of trend detection, limited visuospatial reasoning, recorded cases, and heuristic learning to solve highly varied RPM-derived problems ranging from 2x2 to 3x3 grids.

One way I'd like to improve this is to allow the agent to recognize new trends on its own and build knowledge on how those trends act without manual intervention. Maybe a good topic for a PhD?