Two years have passed since Opena chat, and since then, schoolchildren and researchers have not been interrupted and any person who quickly answers any questions. The famous Great Language Model (LLM) is just one of the many leading programs that intelligently imitate, even bypassing the human mind. Yes, but it seems that the tires may suffer from the same problems we face.
the Science alert According to the article, the Israeli researchers reported that LLMS suffers from people who suffer from age -related decrease. Chatbots was investigated in studies, with version 4 and 4O, two differences in the alphabet and anthropologist Claude 3.5 for the Montreal Cognitive Assessment Test used in the designation of the human cognitive decline.
This test is usually used to detect early Alzheimer's disease and dementia and contains tasks such as the time to determine time; Offer memory exercises.
Chatgpt 4 performance was better, but the oldest Gemini 1.0 LLM was very bad, which means that the oldest artificial intelligence models have a dangerous decrease. While linguistic tests and attention were jumped, they had serious optical/spatial tasks, the solution to the most complex problems, and memory exercises.
This is the level of cognitive decline that resembles the neurological degenerative processes of the brain, and while the authors of the study emphasized that there are major differences between the human brain and artificial intelligence, the results show our weakness. This, in turn, makes its healthy application extensively. Thus, diagnoses, X fees, laboratory results and other medical data cannot be enlarged.
In the future, neurologists and psychiatrists may also need to treat artificial intelligence models with dementia.