Mind-Reading AI Technology Translates A Person's Thoughts Into Text

Technology capable of decoding our thoughts is no longer a far-fetched dream. According to neuroscientists at the University of Texas, for the first time ever, data has been decoded from non-invasive techniques and then reconstructed into meaningful stories.

Sounds like mind-reading, right? It kind of is! According to the new study published in Nature Neuroscience, scientists were able to extract language (and even exact phrases) from fMRI (functional magnetic resonance imaging) brain recordings of three participants.

While most people will perceive this in the mind-reading superpower realm, it's important to note that such tech could be extremely useful to people who cannot speak due to various motor conditions.

Pexels

How language decoding works

Such language decoding models (also known as "speech decoders") essentially use recordings of a person's brain activity to fish out the words they hear, imagine, or say. Until now, such speech decoders have been used only with data from devices that were implanted surgically, which limits their utility.

Many other decoders that use non-invasive brain activity recordings have managed to decode single words or short phrases, but never continuous language. This new research used signals dependent on blood oxygen levels from fMRI scans that are able to detect changes in blood flow and oxygenation levels in different parts of the brain.

Also read: Samsung Bans ChatGPT And Other Generative AI Tools After Sensitive Code Leak

Pexels

For this purpose, they focused on patterns of activity in brain regions and networks that are able to process language. The researchers found that their decoder could be trained to reconstruct continuous language, including specific words and the general meaning of sentences.

To this end, the decoder took into account the brain responses of three participants while they listened to stories. Based on this, the decoder generated word sequences that were most likely to have produced those brain responses.

How the words were captured

In most cases, these word sequences did a great job of capturing the general idea of the stories while being able to include exact words and phrases in some cases. In some cases, researchers asked the participants to watch silent movies and imagine stories while being scanned. The decoder managed to get the gist of the stories in both cases.

Also read: Survey Says 83% Of Indian Voice Scam Victims Lost Money To AI-Generated Fraud

Pexels

For instance, one user thought "I don¡¯t have my driver¡¯s licence yet", and the decoder read it as "she has not even started to learn to drive yet." To achieve this, researchers made participants lie inside an fMRI scanner and listen to 16 hours of narrated stories while their brain responses were recorded that were then used to train an encoder.

The original GPT language model was used to help attach meaning to words and sentences. While this is a useful technology for many people, it could raise fears of concerns surrounding the right to privacy of thoughts.

What do you think about this fascinating technology? Let us know in the comments below. For more in the world of technology and science, keep reading Indiatimes.com.