Researchers at the University of Texas in Austin unveiled a system which uses artificial intelligence to transform a person’s brain activity into written words.
The team of academics, led by computer science doctrinal student Jerry Tang and assistant professor Alex Huth, used a model similar to the one which serves as the basis of ChatGPT to turn brain activity released while “listening to a story or silently imagining telling a story” into a “continuous stream of text.” Unlike other solutions, the system, which the researchers call a “semantic decoder,” does not need surgical implants and is entirely noninvasive.
“This is a real leap forward compared to what’s been done before, which is typically single words or short sentences,” Huth said in a press release. “We’re getting the model to decode continuous language for extended periods of time with complicated ideas.”
The decoder learns how to translate an individual’s thoughts as he or she listens to podcasts and the system concurrently measures brain activity with an fMRI scanner. The decoder then produces text meant to mirror the intended meaning of the individual’s thoughts rather than a word-for-word transcript of the exact words which enter the individual’s brain. In one such example, the decoder said a participant had listened to a speaker say “she has not even started to learn to drive yet” when the participant actually heard “I don’t have my driver’s license yet.”
Researchers added that the system, which was described in a recent study published in Nature Neuroscience, could not decode words for participants who had not willingly trained the system or who intentionally sought to defy the system. “We take very seriously the concerns that it could be used for bad purposes and have worked to avoid that,” Tang commented. “We want to make sure people only use these types of technologies when they want to and that it helps them.”
The system has implications for stroke victims and other individuals who are conscious but unable to speak physically, although the current version relies upon an fMRI machine and is therefore impractical for use outside of a laboratory.
Leading corporations and investors have poured considerable funds into developing AI tools meant for consumer products and business solutions over the past several months. Microsoft, which invested billions of dollars into ChatGPT creator OpenAI, announced that the system would be incorporated into the search engine Bing and internet browser Edge, allowing users to more easily locate information. Google likewise revealed that Bard, an experimental conversational AI service, would soon be added to the company’s search engine.
The semantic decoder study was published as medical researchers in the United Kingdom unveiled an AI tool which can improve predictions of whether a nodule will develop cancer, thereby helping physicians to identify the disease far earlier than current methods.
Some analysts have raised concerns of possible widespread technological employment even as the novel AI applications create productivity increases for workers and businesses. One recent forecast from Goldman Sachs predicted that AI could eliminate 7% of positions in the United States, largely in sectors that rely upon office work such as administrative support and legal, while positions in sectors such as building and grounds maintenance, construction, logistics, and healthcare support are predicted to remain broadly intact.