Will incredible artificial intelligence save us from Alzheimer’s?

For the past few weeks, the internet has only had eyes for ChatGPT. Powered by incredible machine learning, OpenAI’s chatbot has already shown that it can hold conversations, write fake texts, help students cheat…and one day it could be involved in the fight against Alzheimer’s disease.

It is a neurodegenerative disease with dire consequences for patients and their families. This usually results in a loss of memory and clarity that can take on catastrophic proportions over time. Patients then tend to express themselves in disjointed, incoherent, or difficult to understand ways.

The problem is that despite the promising potential, there is still no such treatment. Admittedly, there are many protocols that allow us to deal with the symptoms, but we are still not completely cured of this affliction.

Therefore, the challenge is to reach this diagnosis as soon as possible, which leads to serious consequences. Indeed, the effectiveness of treatment depends greatly on the time of treatment. If the disease is detected at an early stage, it is possible to significantly slow down its development.

But then we face another problem. Because even if you are especially careful, it is not always easy to detect Alzheimer’s disease in its early stages. Indeed, symptoms appear relatively late. ” Studies have shown that up to 90% of brain cells may already be dead by the time a diagnosis is made Valerie Daggett, professor of bioengineering at the University of Washington, explained that she developed a blood test to diagnose the disease.

Statistical tools to monitor the disease

A blood test can detect Alzheimer’s disease years before symptoms appear

The most obvious solution to prevent this problem would be to organize a systematic follow-up of the population, especially the patients at risk. You will usually need to undergo brain imaging sessions and extensive cognitive evaluations to give yourself any chance of reaching the correct diagnosis. Therefore, it would be an expensive, time-consuming monitoring protocol and therefore difficult to implement on a large scale.

A group of researchers from Drexel University in the United States preferred another approach to carry out this preventive work. Instead of searching physiological markers They plan to pass as the famous amyloid plaques of the disease statistical analysis of language.

The idea is to break down a person’s speech using computer tools to identify extremely discreet warning signs. Generally, this involves looking for a few serious acoustic cues; in other words, we are not interested in the deeper meaning of his remarks, but in the way the individual expresses himself. This approach works quite well. Approximately In 75% of casesthis allows for diagnosis before the onset of symptoms, which greatly improves management.

However, it turns out that artificial neural networks illuminate these discrete statistical deviations particularly well. And in parallel, artificial intelligence tools applied to language in recent years have been gaining momentum at an astonishing rate. This offers researchers new possibilities: instead of focusing only on the acoustic dimension, it is now possible to look directly at the meaning of words.

Generative AI to the rescue

So the Drexel researchers asked themselves a timely question: Could a program like ChatGPT detect early signs of Alzheimer’s long before human experts could?

“These language models like GPT3 are so powerful that they can see such subtle differences,” says Hualou Liang, lead author of the study. “If the subject has Alzheimer’s disease and it’s already felt in the tongue, we hope to be able to use machine learning to suggest an early diagnosis,” he said.

To test this hypothesis, they began collecting 237 audio recordings. They came from two groups of patients, healthy or suffering from Alzheimer’s disease. According to IEEE Spectrum, they relied on an unknown feature of ChatGPT to analyze them.

Indeed, the app can only offer a text response. It can also create a “placement”. In fact, it is a very rich mathematical representation of a piece of text. Normally, this allows the algorithm to compare the meaning of the two pieces. Here, the researchers channeled it to analyze the relevance of the comments.

And the results have been quite remarkable. In 80.3% of cases, based on positioning alone, the team was able to determine whether the target patient belonged to the Alzheimer’s group. Not yet revolutionary, but a significant enough improvement over what traditional acoustic analytical methods offer.

Towards new diagnostic tools based on artificial intelligence

Two conclusions can be drawn from this study. First, it confirms the potential of AI applied to language in the management of Alzheimer’s disease. This is great news, as researchers have thus paved the way for the development of new diagnostic tools that are not only very powerful but also less difficult than conventional protocols.

A second implication is that we must be very careful before developing such custom solutions. Because it should be remembered that ChatGPT is a commercial program developed by a private company. According to Frank Rudzicz, an expert also interviewed by IEEE Spectrum, this creates a big problem in terms of transparency and accessibility.

” These closed APIs are limited because you cannot inspect or make deep changes to the code. “, he explains. “ This means that we cannot necessarily perform a more sophisticated series of experiments that allow us to find the source of potential errors that need to be corrected or prevented. “he regrets.

Liang is also quite transparent about the boundaries of this work. For all these reasons, he insists, it is not yet a tool adapted to clinical issues; It’s the day before, not tomorrow, that experts in neurodegenerative diseases will hand over to ChatGPT.

But the most important thing and what should be learned from this work is the potential of this approach. As specific tools emerge and mature, humanity will have access to a very powerful tool to wage an unprecedented war against Alzheimer’s disease. Knowing the human and medical risks of this public health scourge, this progress should be unreservedly welcomed. And this is probably only the beginning of knowing the incredible potential of artificial intelligence in many areas of medicine.

The text of the study can be found here.