The unsuspected medical virtues of ChatGPT revealed by a study

A new study reveals an unexpected effect of ChatGPT-like language models.
Discovering that you know nothing or very little. This is the feeling of many observers with ChatGPT. For example, we would not have imagined that artificial intelligence would have a role to play in helping to detect Alzheimer's disease, which today affects 55 million people worldwide.

A very effective tool

However, researchers at Drexel University in Pennsylvania have just published an important study which shows that GPT-3, the program which is used by ChatGPT, is able to identify certain clues in a person's writings which make it possible to locate the early stages of the disease. The efficiency would reach 80%, a very honorable level.

Quoted by our colleagues from Unite.ai, Hualou Liang, co-author of this research explains: "The most commonly used tests for the early detection of Alzheimer's disease examine acoustic characteristics, such as pauses, articulation and voice quality, in addition to status tests. But we believe that improving natural language processing programs offers another avenue to support the early identification of Alzheimer's disease. ".

Discovering that you know nothing or very little. This is the feeling of many observers with ChatGPT. For example, we would not have imagined that artificial intelligence would have a role to play in helping to detect Alzheimer's disease, which today affects 55 million people worldwide.
A very effective tool

However, researchers at Drexel University in Pennsylvania have just published an important study which shows that GPT-3, the program which is used by ChatGPT, is able to identify certain clues in a person's writings which make it possible to locate the early stages of the disease. The efficiency would reach 80%, a very honorable level.

Quoted by our colleagues from Unite.ai, Hualou Liang, co-author of this research explains: "The most commonly used tests for the early detection of Alzheimer's disease examine acoustic characteristics, such as pauses, articulation and voice quality, in addition to status tests. But we believe that improving natural language processing programs offers another avenue to support the early identification of Alzheimer's disease. ".

So how does it work ? GPT-3 is well suited to spot certain language features that predict dementia, the authors explain. By training this language model on this specific point, it could identify these markers even better.

Suiting words to deeds, the scientists “fed” GPT-3 with patient voice recordings to test the relevance of AI in diagnosis. Some had Alzheimer's disease while others did not. As we have seen above, the results were very conclusive.

And the researchers conclude: "Our results demonstrate that text integration, generated by GPT-3, can be used reliably not only to detect people with Alzheimer's disease compared to healthy controls, but also to infer the subject's cognitive test score, in both cases based solely on the voice data."

Discovering that you know nothing or very little. This is the feeling of many observers with ChatGPT. For example, we would not have imagined that artificial intelligence would have a role to play in helping to detect Alzheimer's disease, which today affects 55 million people worldwide.
A very effective tool

However, researchers at Drexel University in Pennsylvania have just published an important study which shows that GPT-3, the program which is used by ChatGPT, is able to identify certain clues in a person's writings which make it possible to locate the early stages of the disease. The efficiency would reach 80%, a very honorable level.

Quoted by our colleagues from Unite.ai, Hualou Liang, co-author of this research explains: "The most commonly used tests for the early detection of Alzheimer's disease examine acoustic characteristics, such as pauses, articulation and voice quality, in addition to status tests. But we believe that improving natural language processing programs offers another avenue to support the early identification of Alzheimer's disease. ".