The power and reach of Artificial Intelligence (IA) remains an unknown. Recently, Scientists have developed a new system that can transform thoughts into complex texts. The tool is an excellent novelty for the health sector, being able to help stroke patients (Stroke), paralysis or degenerative diseases.
The technology is being studied by researchers at the University of Texas, us United States, where the full study was published by the journal Nature Neuroscience. This machine with AI functionality allows the user's brain activities to be converted into conversations without any risk to patients undergoing.
How the technology works?
The system is called a “semantic decoder” and, according to CNB, uses a transformer model similar to ChatGPT and Google chatbots, for example, for generating texts. The operation of the technology is also coupled to an fMRI scanner.
This scanner is responsible for measuring patients' brain activity. But, so that the AI machine delivers the correct information, the semantic language decoder was trained for several hours by being exposed to podcast audio. Like this, the tool can retain relevant information.
The text does not provide an exact transcription. But, after all this training, the machine can generate a text flow for the patient from listening to conversations or when the patient imagines telling a story to other people.
Non-invasive method in testing phase
The technology is not considered invasive, i.e., would not pose risks to patients using the system. Even because no operational process is necessary for the messages to be reproduced. Offering autonomy to people unable to speak is a solution that has been studied by science for many years.
The difference, although, is that the advancement of Artificial Intelligence has enabled continuous language and the retention of more complex ideas. Although, despite the novelty, the decoder is not suitable for sale or use outside laboratories as it depends on the fMRI scanner.
Although, scientists hope that portable systems will soon be launched. “Towards a non-invasive method, this is a real leap compared to what has been done before, which are usually single words or short phrases.”, said one of the research leaders, Alexander Huth.
