This study on older AI showing signs of "cognitive decline" is quite interesting, but at the same time, somewhat confusing. Some people believe that older models are struggling with complex tasks, such as remembering things or processing visual information. However, others are questioning the validity of these tests, as they were designed for humans, not AI.

Can this be "cured"? Maybe it's a matter of improving model training or adjusting the tests to better align with what AIs actually do. It remains an open question, right?