A new study published in Nature reveals that artificial intelligence systems are developing an unexpected ability to reason by analogy, a cognitive skill previously thought to be uniquely human. Researchers at Stanford University found that large language models, when trained on specific datasets, can solve analogy problems without explicit programming. The AI demonstrated proficiency in tasks like identifying relational similarities between word pairs, such as recognizing that ‘hand’ is to ‘glove’ as ‘foot’ is to ‘sock.’ This emergent capability suggests that complex reasoning might arise from scaling up model size and training data, rather than from specialized architectural changes. The findings challenge long-held assumptions about the limitations of machine intelligence and open new avenues for developing more general AI systems. For the full details, read the complete article at https://technologyreview.com/2024/05/15/ai-analogy-reasoning-breakthrough.