It is still not perfect, but artificial intelligence has managed to read the essence of thoughts.
A group of scientists managed to train artificial intelligence to read people’s minds. The technology behind artificial intelligence is not too different from of the popular ChatGPTwhich can have a meaningful conversation with the user based on their input, with the input in this case being the activity of the human brain.
The University of Texas team is on Monday published his research in the journal Nature Neuroscience. Their method uses magnetic resonance imaging to interpret what the examinee says “hears, speaks or imagines”. This is not the first time that scientists have developed technology that can read minds, but it is the first successful method that does not require electrodes connected to the subject’s brain.
AI model, called GPT-1, is currently the only technology that can coherently interpret brain activity. Other techniques can extract a word or a short phase, but GPT-1 can form complex descriptions that explain the essence of what the respondent is thinking.
For example, one participant in the study listened to a recording in which someone said: “I don’t have a driver’s license yet”. The language model interpreted the recording of brain activity as “she hasn’t even started learning to drive yet”. Therefore, although it does not read a person’s mind literally, it can recognize what is being thought about, all based on the blood flow in the brain. Accurate reading requires, at least for now, a medically invasive method that monitors other neurological activities.
The AI was trained on three volunteers who spent 16 hours listening to stories while in an MRI machine, as well as based on Reddit comments and autobiographies. This allowed the GPT-1 model to link brain activity for words and ideas.
After training it, subjects were played completely new stories, and GPT-1 was able to roughly determine what the story was about, or what the subjects were listening to. The study also included playing movies without sound, and similar results were obtained based on the volunteers’ visual interpretation.
However, the GPT-1 was most accurate when trying to figure out what subjects were listening to, rather than what they were thinking about independently, which is most likely due to the abstract nature of people’s thoughts rather than the concrete ideas formed while listening.
Reading human minds is undoubtedly the most controversial application of artificial intelligence to date, even surpassing the first layoffs to be made by AI. Although the team of scientists imagines that the technology will be applied to people who cannot communicate verbally, the scientists themselves have expressed fear that GPT-1 will not always be used on a voluntary basis.
See how artificial intelligence turned our imagination into images:
See description
Scientists invented artificial intelligence that reads minds
