A criterion for Artificial General Intelligence: hypothetic-deductive reasoning, tested on ChatGPT

The article proposes a criterion for Artificial General Intelligence (AGI) based on hypothetic-deductive reasoning, which it argues is essential for advanced AI systems like GPT-4. The authors introduce tests for both hypothetic-deductive and causal reasoning, applying them to ChatGPT, and find that its reasoning capabilities are limited when faced with complex problems. This research highlights the need for enhanced reasoning skills in AI development, suggesting that achieving robust hypothetic-deductive reasoning could be a significant step toward AGI.