New AI Technology in Japan Interprets Thoughts and Visual Images
Artificical Intelligence has come to a place of interpreting the observations or recall of a human from brain data. This development represents a potential game changer in human-computer interaction.
The Approach Conceive Language from Brain Signals
Participants inside an fMRI scanner viewed thousands of video clips. By means of a deep language neural network, the researchers drew abstract meanings from the video captions. The AI model then matched these meanings to the brain responses created by individual participants. The algorithm learned to compose grammatical sentences over time by choosing words that semantically matched the observed brain signals.
Predictive Accuracy for Visual and Imaginary Events
The testing demonstrated noteworthy levels of predictive capability for their application
- Visual Recognition According to their predictions of description for novel video clips, the AI recognized the specific clip among 100 choices about 50 percent of the time.
- Memory Recall The accuracy for systems in tests where the participants just thought about the video they had viewed was 40 percent.
Importance for Neurointerfaces
A promising finding was that the technique was able to work without any contributions from the brain's classical speech territories. While this technology cannot yet be applied to practical and everyday use, they appear to hold some promise for the future of advanced neurointerfaces.
