Dreamer Science

Artificial intelligence reads human thoughts

For the first time, a group of researchers at the University of Texas at Austin, USA, has succeeded in building a translator that can convert human thoughts into words without inserting devices into the brain (non-invasive). This unique achievement has been achieved with the help of artificial intelligence.

Earlier recognition of human thoughts was done through computer systems, but it was all done through the insertion of sensory devices into the human brain. In this research, the data obtained from the fMRI (Functional Magnetic Resonance Imaging) scan technology, which can monitor brain function through changes in blood flow in the brain, was used. The researchers had analyzed them by using the Large Language Model called GPT-1, which was introduced before the introduction of ChatGPT by the Open AI company.

Here, the researchers arranged for three volunteers to listen to podcasts for 16 hours each, and their brain activity was monitored through fMRI. The researchers then trained the GPT-1 language model to match those brain activations to the content of the podcast. The research participants were then allowed to listen to other new podcasts while the GPT-1 language model was instructed to predict the content of the new podcast by examining the brain activity data obtained in the process. In about half of the cases, it has managed to turn the thought into words almost correctly and in some cases extremely correctly.

Scholars point out that this research is a huge leap in this field. This can be introduced as an important step forward in the effort to build an interface that connects the human brain and computers. Many things, such as dream recording machines, which were once only limited to science fiction, may become real realities through this.

The research report related to this research was published on May 01 in Nature Neuroscience Journal.