Scientist Develop AI System that convert Brain activity into Written Text




Researchers have made a significant advancement in mind-reading technology with the development of BrainLLM, an innovative system that combines brain scan data with large language models to translate neural activity into coherent text. Unlike previous methods that selected from predetermined word options, this approach generates natural language directly from brain signals.
The system works by using functional magnetic resonance imaging (fMRI) to capture brain activity while a person thinks about words or concepts. A specialized neural network then transforms these brain patterns into a format that can be processed by a large language model, which generates appropriate text. Testing showed that BrainLLM produced more contextually relevant and coherent text than traditional methods, performing particularly well when analyzing activity in language-processing brain regions like Broca's area.
While this technology represents an important step toward direct brain-to-text communication, researchers acknowledge several limitations. The accuracy varies significantly between individuals, and fMRI technology is currently too bulky and slow for practical real-time applications. Future research will explore using more portable technologies like electroencephalography (EEG) to develop more practical brain-computer interfaces that could potentially help people with speech impairments communicate more effectively.
#BrainDecoding #AIResearch #NeuralTranslation #BrainComputerInterface #CognitiveScience

Tech