Shocking MIT Discovery: AI Doesn’t Understand the World
In a move that went largely unnoticed by the media, the Massachusetts Institute of Technology (MIT) just landed a massive blow to the giants of artificial intelligence.
A team of MIT researchers developed a groundbreaking new benchmark called WorldTest, designed specifically to reveal whether AI models genuinely comprehend the world or are merely imitating learned patterns. The outcome? A total disappointment.
During the evaluation, even the most advanced systems on the planet—Claude, Gemini 2.5 Pro, and OpenAI o3—were thoroughly defeated when pitted against basic human reasoning.
This isn’t like the usual tests that reward AI for guessing the next word in a sentence or the next frame in a video. WorldTest goes deeper: it demands that AI construct an internal representation of reality and then use it to respond intelligently to completely unfamiliar scenarios.