By @NatGeo
Publication Date: 2025-12-31 00:00:00
There are areas in which machine cognition still lags far behind that of humans. The planning ability of AI systems is currently “at the level of a child,” says Bengio, although he notes that frontier models are making rapid progress in this area. Because they are trained primarily on text and images, their spatial thinking skills are also poor, he says.
A big question in AI research today is how far further models can be advanced towards human or superhuman intelligence. While companies have made tremendous progress by putting enormous amounts of computing power into their models, it is not clear that ever more powerful computers will lead to ever more intelligent machines. At some point something else might be required.
Mitchell points out that while babies learn by interacting with the world around them, AI systems are trained passively by feeding them massive amounts of information. This could explain why chatbots tend to lie: since they lack real feedback,…
