AI Decoder: Artificial Intelligence Translates Human Brain Activity Into Written Words

New artificial intelligence developed to translate human brain activity into written words
The technology is designed to be trained to decode complex language from a person's thoughts for extended periods
The study is noteworthy because it does not require participants to get implants or be restricted to only a few words

A new type of artificial intelligence was created that can translate a human person's brain activity into written words using non-invasive scanning methods.

Researchers from the University of Texas at Austin were responsible for the new AI technology that can be trained to decode complex language from a person's thoughts for extended periods.

AI Decoder Translate Brain Activity Into Words

In a statement, an assistant professor of neuroscience and computer science and a co-lead of the study, Alex Huth, said that the latest development of the non-invasive method marked a huge leap forward compared to what others have achieved in the past. These were usually efforts that resulted in single words or short sentences.

There are currently other systems being developed elsewhere, but the one in Texas is noteworthy because it does not require participants to undergo surgery to get implants. They are also not restricted to only a list of words they use, per IFL Science.

Similar technology has been used in OpenAI's ChatGPT and Google's Bard chatbots capable of generating text information from prompts. The new decoder was trained on hours of data obtained from an individual listening to podcasts while getting their brain scanned by functional magnetic resonance imaging (fMRI).

Afterward, with the participant's consent, they can have their thoughts decoded while listening to a new story or imagining telling a story. The artificial intelligence model then generated a stream of text, translating the brain activity into written words.

While the newly-developed decoder cannot translate a person's thoughts word for word, it is commonly able to get a general idea of what the participant is thinking. After extensive training, the artificial intelligence produced text that was good and sometimes an exact representation of the participant's thoughts roughly half the time.

Trained to Decode Complex Language

An author of the study, Alexander Huth, said that the AI could get the ideas behind the words, the semantics, and the meaning. Despite the results, the artificial intelligence decoder cannot read a person's mind and only works if the participant is cooperating with the scientists, according to NPR.

Despite this restriction, systems capable of decoding language could someday allow people who cannot talk due to brain injury or disease to communicate with others more efficiently. Furthermore, this technology helps scientists understand more about the human brain and how it processes words and thoughts.

Huth argued that the study's results showed that it is not simply language stimulus but the AI getting the meaning behind something of what was happening. He noted that the fact that this type of technology is even possible is "very exciting," said the New York Times.