Artificial intelligence has learned to "read" human thoughts

Researchers from the University of Texas at Austin have developed a neural decoder that can translate thoughts into text. The new algorithm allows for the analysis of brain activity using functional MRI and interprets it into coherent sentences. The findings were published in the journal Current Biology (CurBio).

How Does the Brain Decoder Work?

Previously, such technologies required many hours of training on data from a specific individual, significantly limiting their practical application. However, the new algorithm has sped up the process considerably:

• Adaptation to another person takes only 70 minutes.
• The decoder trains on data collected while participants watch movies or listen to audio.
• The algorithm analyzes the brain responses of different people to the same stories, enabling the transfer of the trained model from one person to another.

How Does AI "Understand" Thoughts?

The decoder does not reproduce thoughts literally but interprets them by meaning. For example:

• A person thinks: "I am a waitress at an ice cream shop."
• The algorithm "reads" the thought as: "I was at work, which I considered boring."

The system also works with silent films, opening new possibilities for individuals with aphasia — a disorder that impairs the ability to speak.

The Future of the Technology

The next step will be testing on people with aphasia and creating an interface that allows them to express their thoughts without words. Researchers hope that in the future, this technology will become an important communication tool, aiding those with speech impairments to regain their ability to communicate.