Singaporean researchers have developed a powerful AI system that can decipher brain wave patterns and generate pictures of what a person is observing. The team of experts conducted MRI scans of a group of 58 volunteers while showing them 1,200 – 5,000 images of food, buildings, animals, and human activities and gathered their data sets.
Each image is shown for a span of 9 seconds with breaks in between. The AI- system dubbed MinD-Vis correlates the images with the brain scans and generates a distinct AI figure for each volunteer.
These models equip computers with the ability to interpret thoughts and recreate the visuals that the individual is looking at.
MinD-Vis Interprets Brain Activity Like ChatGPT
Jiaxin Qing, a Ph.D. student and one of the main researchers at The Chinese University of Hong Kong (CUHK IE) elaborates on how MinD-Vis comprehends brain actions, similar to ChatGPT’s understanding of human languages.
Following this, the AI-system converts brain activities into a language which is comprehended by an open-source AI, termed Stable Diffusion to produce images from text.
Qing brought up that majority of the time, the decoded images matched what the participants had seen.
Seeing Beyond the Brain: Conditional Diffusion Model with Sparse Masked Modeling for Vision Decoding— AK (@_akhaliq) November 15, 2022
project page: https://t.co/6nTi6yb8xN
github: https://t.co/mnfKG8Nhj2 pic.twitter.com/DSxii6Exim
Fascinated by the concept of brain decoding, one of the volunteers, Li Rulin expressed his exhilaration in being part of the project that entailed utilising brain signals to generate natural modalities. “I’m also curious about what goes on inside my head, what it can produce, and what I’m thinking,” he added.
This research team is confident that this technology will eventually be able to help people.
According to Chen Zijiao from the School of Medicine at the National University of Singapore, the technology might help people without motor ability control their robots [artificial limbs…or] communicate with others by just using their thoughts instead of speech if that person couldn’t speak at that time.
The researchers credit recent advances in computer capacity, which enable effective data processing, and the improved accessibility of MRI datasets for the development of their mind-reading AI.
#Watch | Researchers from National University of #Singapore say they've developed MinD-VIS, a mind-reading #AI. The tech can 'read' a person's thoughts & re-create visuals s/he is examining. However, there are possible risks lurking & relative lack of #legislation in AI research pic.twitter.com/tbVCglTPkR— CNBC-TV18 (@CNBCTV18News) August 16, 2023
The researchers acknowledge the fact that it will take several years of development before MinD-Vis can accurately interpret the public’s ideas.