Breakthrough technology: Artificial intelligence can read human thoughts

A group of American researchers has just announced a breakthrough technology that can convert what people see or imagine into text descriptions with high accuracy.

The “mind captioning” technology is surprising the scientific world by turning brain signals into accurate descriptions of images or scenes that people are thinking about. This is a new technology announced by a group of American scientists - a remarkable step forward in the field of artificial intelligence and neuroscience: “mind-captioning” technology.

The system can decode brain activity and convert what a person sees or imagines in their head into a written description with astonishing accuracy.

The work, published in the journal Science Advances, opens up new perspectives in understanding how the human brain interprets the world around it. At the same time, this new technology also suggests potential applications in supporting people with language disorders, such as after a stroke.

Mind-captioning technology can decode brain activity and convert what people see or imagine in their heads into text descriptions with high accuracy. (Photo: Nature.com)

According to a research team at the University of California, Berkeley (USA), this technique uses non-invasive brain imaging technology to record neural activity. Scientists then use advanced AI language models to predict and accurately describe what the participant is seeing or imagining.

“The system can predict at a very high level of detail what a person is seeing, which is unprecedented,” says computational neuroscientist Alex Huth.

For more than a decade, researchers have been trying to predict images or sounds based on brain signals. However, “translating” complex content like videos or abstract shapes into language remains a major challenge.

Previous methods only recognized single keywords instead of the entire context such as subject, action and surrounding environment.

To overcome this limitation, the research team of Tomoyasu Horikawa, a neuroscientist at NTT Communications Science Laboratories (Japan), developed a deep language AI model. First, they used a deep language model to analyze the subtitles of more than 2,000 videos, converting each segment into a “meaning signature” – a digital graph representing the video content.

The team then trained another AI system to recognize the connection between these “meaning signatures” and the brain activity of six participants as they watched the videos.

The results showed that AI could accurately “translate” what participants were seeing or imagining into complete descriptive sentences, not just listing keywords.

Experts say that “mind-captioning” technology is not only a milestone in neuroscience research but also opens up the possibility of wide practical applications. In the future, this technique can support people who have lost the ability to speak, helping them communicate through brain signals.

Although still in its early stages, “mind-captioning” marks a major step forward in the journey to decode the language of human thought, bringing the prospect of “mind reading” that used to only exist in science fiction movies closer to reality.

Tran Vy

Source: https://vtcnews.vn/cong-nghe-dot-pha-tri-tue-nhan-tao-doc-duoc-suy-nghi-con-nguoi-ar985947.html