top of page

Techno-magic mind reading

Updated: Nov 11, 2023

By Liliana Chai and Gilina Voon Nov. 9, 2023


Jude Tantawy Art

“If you could have any superpower, what would it be?” One of the most common answers among children is the ability to read people’s minds. With the recent new GPT-1 technology, that dream may soon turn into a reality.


Researchers at the University of Texas at Austin (UT Austin) have discovered a method for translating brain activity into speech using AI and fMRI. Previously, language decoding systems, like Neuroprosthetic Speech Devices, often required surgical implants and were limited to decoding only a few words at a time. However, UT Austin’s new non-invasive technique can map out large amounts of speech with high accuracy, allowing for real-time mapping and increasing cost efficiency.


The research was conducted by a team of scientists led by Dr. Alexander Hunt. First, Hunt observed three participants who, over several days, listened to narrative podcasts like “The Moth.” During this time, the fMRI scanner recorded their blood oxygenation levels (which changed in response to brain activity) and language models like OpenAI’s GPT-4 were used to match patterns in brain activity with the words and phrases the participants were hearing. However, this method of translating brain activity made real-time tracking impossible due to lag between the signals fMRI measures and neural impulses in the human brain.


The GPT-1 model addresses the fMRI delay by generating numerous possible word sequences for a recorded brain response, ultimately choosing the most appropriate candidate. To test this technology, the same participants were scanned while listening to or imagining telling a story. While the generator could not reproduce the text exactly, it retained the intended meaning. For example, according to The Guardian, a participant listened to the words “I don’t have my driver’s license yet” to which the decoder translated her brain activity as “She has not even started to learn to drive yet.”

“GPT-1 can help people who cannot speak find a simpler and more efficient way to communicate. Previously, communication methods such as eye tracking devices were more time-consuming,” Senior Amir Salman said.

UT Austin researchers are optimistic about the potential of GPT-1 in medical care, such as enabling disabled people with stroke or motor neuron disease to communicate. Professor Time Behrens, a computational neuroscientist from the University of Oxford, argues that the technology can also be used to decode dreams or analyze how the brain comes up with ideas.


There are certain limitations associated with GPT-1 that will require further research. fMRI is expensive to use. Furthermore, training the model is a time-consuming process and requires a large number of subjects to account for diverse brain activity patterns. Participants can also impair the decoder’s accuracy by thinking about unrelated matters during the experiment. Moreover, while GPT-1 is capable of linking words to specific meanings, it struggles with several aspects of the English language such as using the appropriate pronouns and the correct point of view. There are also ethical concerns associated with AI mind-reading.


“The initial purpose of new technology was to make life easier, but it can often be abused. Mind-reading is an invasion of people’s privacy, as individuals are not voluntarily expressing themselves verbally or in writing,” Sophomore Ellisa Kim said.

With the advancement of technology, mind reading could be applied to grant individuals with speech impairments the ability to communicate or assist in criminal interrogations. However, when used improperly, tools like GPT-1 could infringe upon individuals' privacy, prompting researchers to be mindful of ethical concerns.

 

About the contributors

Gilina Voon

staff writer, photographer


Gilina Voon is a senior at Leland High School and is a writer and photographer for The Charger Account. During her free time, she enjoys traveling, hanging out with friends, and running.



Liliana Chai

artist


Liliana Chai is a freshman attending Leland High School and is an artist for the 2023-24 Charger Account. In her free time, she enjoys listening to music, playing piano, sleeping, arts and crafts, and writing poetry. She is looking forward to Journalism and hopes to explore new ideas while collaborating with other people.


Jude Tantawy

artist


Jude Tantawy is a senior at Leland High School and is an artist for The Charger Account. During her free time, she enjoys drawing & painting, cooking, baking, playing video games, and is always listening to music.

19 views0 comments

Recent Posts

See All

Drone deliveries explode onto the scene

By Vira Patil, James Yu Apr. 3, 2024 Despite their bland, unassuming appearances, cardboard packages often excite shoppers whenever one is delivered to their door. However, the waiting time fo

Open AI sora revolutionizes videography

By Joshua Yan Apr. 3, 2024 From disco-dancing kangaroos to petri dish pandas, OpenAI’s Sora has recently emerged as a massive pioneering force pushing AI development, being capable of generati

コメント


Screen Shot 2024-02-24 at 7.54.40 PM.png
Screen Shot 2024-02-24 at 7.55.49 PM.png

Facebook

Have any questions? Want to make any suggestions? Contact us at 

We'll reply as soon as we can!

  • Facebook
  • Instagram

Notice any mistakes?

Contact us here!

Recent Articles

Screen Shot 2024-02-24 at 7.55.11 PM.png
bottom of page