It turns out that AI may understand humans better than we thought. When 12 AI models were asked for their most controversial views about humans, most answers returned to one theme: people often care more about fitting in than finding the truth. The ModelCompare experiment collected 35 responses, with 29 suggesting that group loyalty and existing beliefs can outweigh facts. Several models also suggested people use intelligence to defend choices they’ve already made rather than question them.
One Claude response offered a warmer view, saying people are often kinder than they realize. But the test doesn’t prove those claims about human nature. It used a small number of answers, a question designed to provoke strong responses, and categories chosen by researchers. The models can repeat ideas learned from human writing without holding personal beliefs, so similar answers may reflect shared training material. Either that, or “pack mentality” rules all; ironically, only humans can decide if it’s true.
Source: Toms Guide
