<ul><li>A new study from MIT reveals that large language models (LLMs), such as those used in powerful chatbots, do not understand anything.</li><li>MIT researchers developed new metrics to test AI models, focusing on deterministic finite automations (DFAs).</li><li>The study found that LLMs performed well in providing driving directions in New York City but struggled when faced with detours, revealing a lack of accurate internal models.</li><li>The research emphasizes that LLMs rely on prediction rather than reasoning and understanding, and their accuracy can quickly diminish in real-world scenarios.</li></ul>

New MIT study shows what you already knew about AI: it doesn't actually understand anything

Discover more