A team of researchers has developed an algorithm that allows an AI -powered “brain decoder”, trained in one person, translating thoughts from another with a minimum training
A revolutionary “brain decoder” of artificial intelligence (IA) can read a person’s thoughts, only with a rapid examination to the brain and almost without formation.
A study recently in Current Biology announced that a group of scientists at the University of Texas in Austin (UT Austin) has developed improvements in a “Brain decoder” that uses I was going to convert thoughts into text.
The new conversion algorithm is able to train in record time an existing decoder in someone else’s brain.
As it details, this brain decoder uses automatic learning to translate a person’s thoughts into text, based on the answers of his brain to the stories he heard.
Less effective was either anterior decoderwhich required participants to hear stories within a magnetic resonance machine for many hours. In addition, it worked only for the individuals where they had been trained.
In the new study, UT Austin’s computational neuroscientist Alexander Huth and the co -author Jerry tanga graduate student investigated the way to overcome this limitation.
The researchers trained the brain decoder of some reference participants, collecting functional magnetic resonance imaging while listening 10 hours of radio stories. Then they trained two conversion algorithms in the reference participants and in another set of participants: some spent 70 minutes listening to radio stories; Others spent 70 minutes to see short films not related to radio stories.
Using a technique called functional alignmentthe team mapped the way the brains of reference and objective participants responded to the same audio stories or movies.
Live Science quotes a test story included someone talking about a job they didn’t like, saying, “I’m a table maid in a freezing. I don’t know where I want to be, but I know it’s not here. ”
The decoder, using the converter algorithm trained in movie data, predicted: “I was in a job that I found boring. I had to receive orders and didn’t like them, so I worked on them every day. ”
Although not an exact correspondence (since the decoder does not read the exact sounds that people have heard), The ideas are related.
“What is really amazing and cool is that we can do this without using linguistic data,” Alexander Huth told Live Science.
“So we can have data that we collect only while someone is watching silent videos and then we can use them to build this language decoder for your brain,” he said.
