Using a variety of analogy puzzles, SFI researchers have shown that the reasoning abilities of OpenAI’s GPT-4 model fall ...