These AI models reason better than their open-source peers - but still can't rival humans

Researchers tested AI models' ability to solve abstract visual puzzles designed for human IQ tests. The results were mixed, with most models struggling to understand visuals and interpret patterns. Open-source models performed worse than closed-source models like GPT-4V, though none matched human cognitive abilities. Using "Chain of Thought" prompting improved some models' performance. This research helps identify AI's current limitations in reasoning, guiding future efforts to develop more advanced and capable AI systems.