A team of researchers from the University of Technology Sydney, Australia, is developing an artificial intelligence (AI) system capable of translating thoughts into text.
Technology has the potential to pave the way for more intuitive and accessible interfaces. One of the possible uses is to help people with paralysis communicate.
The results of the study were made public during the conference NeurIPS, in the United States. At the event, researchers highlighted the system’s potential ability to improve human-machine communication, especially in interactions with robots.
How the study was done
During the tests, 29 participants wore caps equipped with electrodes to record brain activity while silently reading passages of text.
Then, brain activity was analyzed by the AI model, called DeWave, developed by the team. This non-invasive model efficiently translates brain activity into words and sentences, providing direct communication between thought and text.
For now, the process accuracy rate is 40%. Researchers expect this percentage to improve to a level comparable to traditional language translation, or speech recognition programs, which is closer to 90%.
In addition to its uses in helping people with paralysis communicate, the ability to decode thoughts could revolutionize the diagnosis and treatment of neurological conditions. Thus, providing significant advances in understanding and addressing brain disorders.
In this video (in English) scientists demonstrate the technology in practice:
Advantages of AI over Neuralink
Unlike more invasive approaches, such as Elon Musk’s proposed Neuralink implants, the technology developed by the Australian team demonstrates a non-invasive approach.
Therefore, in addition to the potential impact on communication for people with motor difficulties, the application of AI in healthcare is also promising. Furthermore, the researchers emphasize the versatility of the system, which can be used with or without additional technologies such as eye movement tracking.
2023-12-31 23:02:00
#Created
