Semantic Brain Decoder: Guessing Thoughts through Brain Activity
Scientists from the University of Texas at Austin have developed a “semantic brain decoder” that can predict what someone is thinking based on their brain activity. The decoder can capture the essence of what a person is thinking, rather than a direct translation. If a participant resists, the decoder produces gibberish.
The findings, published in the journal Nature Neuroscience in May, showed that the algorithm used by the decoder was quite impressive. However, it poses limited threats to privacy, according to Edmund Lalor, an associate professor of neuroscience at the University of Rochester.
The research team designed the decoder by listening to 16 hours of podcasts which the participants listened to while wearing an MRI brain scanner. After analyzing the MRI data, the researchers taught the decoder to recognize the language patterns that corresponded to various types of brain activity. Participants then listened to different podcasts or imagined themselves telling stories while the decoder made brief “guesses” of what they were thinking. After eliminating poor guesses, the decoder further explored the better ones with the earlier version of the AI chatbot ChatGPT, which predicts the next word.
The decoder then repeated this process until it produced consistent results. The scientists compared it to a transcript of the story the participants imagined telling or the podcast the participant was listening to.
Decoder’s performance
The decoder outperformed randomly generated translations, retaining the general meaning of participants’ thoughts. Although its predictions were not perfect, the results were quite impressive.
For example, “I don’t have my driver’s license yet,” was translated to “She has not even started to learn to drive,” while “That night I went upstairs to what had been our bedroom” was translated to “We got back to my dorm room.”
Not a lie detector
Huth and his team are aware of the risks associated with such technologies, fearing that future versions could be used to detect lies or to read people’s memories without their consent. In order to prevent this, the team ran several tests on the decoder to detect what it could and couldn’t do.
For instance, the scientists tried testing if participants could disrupt the decoder by silently naming as many animals as possible while listening to stories. This experiment showed that participants could indeed disrupt the decoder and that they could turn it off if they thought of something else. The tests proved that the decoder did not have intrusive capabilities against privacy.
According to Lalor, the AI algorithms are trained on data that may have human biases. For instance, some AI models connect doctors with men and nurses with women. In contrast, the participants of this research listened to podcasters of diverse genders, ethnicities, and sexualities. However, there were concerns since the researchers had less control over the data used to train the GPT models used in the research.
The Benefits of the Brain Decoder
The primary goal of creating the decoder is to help individuals who cannot talk after a stroke or other medical emergencies. However, it currently requires an expensive, immobile MRI machine to function and takes hours to analyze predictions. So the team is looking for alternative, portable methods to measure brain activity, including functional near-infrared spectroscopy (fNIRS), which measures similar signals as an MRI machine but from a helmet.
In conclusion, although the brain decoder may seem promising, it can only decipher thoughts, not memories. The researchers of this study agree that human cognition goes beyond predicting next words. The brain decoder has tapped into how humans interpret the world through language, but it falls short of accounting for the sensations that come with experiencing the world.
Editor Notes
Brain decoder technology has raised concerns about privacy invasion. However, the current decoder is still far from reading minds against someone’s will. The experiments done showed that the decoder could be turned off if the participants thought of something else. The primary use for the decoder is to help individuals who cannot talk after a stroke or other medical emergencies. The research team is also exploring alternative, portable methods to measure brain activity that do not require an immobile MRI machine.
The development of this technology is an exciting breakthrough for neuroscience. To read more about the latest news in science, head on to GPT News Room.
from GPT News Room https://ift.tt/zrPI6ey
No comments:
Post a Comment