Artificial intelligence technologies, which have gained a great place in our lives despite being at an early level, sat on the agenda with a striking research. A cluster scientist has done a study to see if the intelligence of artificial intelligence is always at a level. Scientists, who benefited from the MOCA test used for Alzheimer’s and dementia situations, faced a result they didn’t expect much.

The research, which was published on December 20, 2024, now showed that artificial intelligence began to deteriorate after a while. Artificial intelligence models, which are actively used compared to the research, were partially lost their cognitive abilities over time. In other words, as in humans, artificial intelligence worsened as it was age.

New models gave more appropriate results than older models

Scientists, within the scope of the research OpenAI’s GPT, Google’s Gemini and Anthropic’s Sonnet models recognized artificial intelligence models in the 30 -point MOCA test. In the tests where the GPT-4O model received 26 points, Gemini was able to get only 16 points. So older models lost their capacity more. But there is a valuable issue here. Relatively old artificial intelligence language models were not enough in terms of their capacity. This is exactly the source of the problem. Although the researchers say that all models have symptoms of cognitive disorder, they are not mentioned in the first initial capacity of the models.

Let’s talk about the MOCA test used in this research while ending our news. The human beings inserted into the MOCA test, they are definitely trying to draw a time on the clock dial, start at 100 and always raise 7, and remember the highest number of words in a speech text. Beşer, who received 26 out of 30, is considered healthy. The test on artificial intelligence showed that only the GPT-4O was healthy.