DREAM MACHINE
Scientists have found a way to recreate images of faces using only data from a brain scan. One day, they say, the same technology could allow them to reconstruct people’s dreams. Jonathan Leake reports.
Scientists have created a mindreading machine so powerful it can extract images from people’s brains and then display them on a screen for others to see.
The system has been used to accurately reconstruct human faces based only on data from the brain scanner.
The researchers say the same approach could one day also allow them to reconstruct images from people’s dreams, memories and imagination.
Another future application could be to generate images of criminals from the minds of witnesses.
“Our methods yield strikingly accurate neural reconstructions of faces,” said Alan Cowen, a neuroscientist at the University of California, Berkeley.
“This represents a novel and promising approach for investigating face perception, but also suggests avenues for reconstructing ‘off-line’ visual experiences — including dreams, memories and imagination.”
In the research, the scientists showed six volunteers 300 faces while they were lying in an MRI scanner. This process showed how their brains responded to dozens of features ranging from blond hair and blue eyes to dark skin and beards.
Once they had built up the database of responses, across several areas of the brain, they showed the volunteers a set of new faces. Then they measured how each volunteer’s brain responded to the new image and, by comparing those responses to the database, reconstructed the image they were looking at.
At the heart of such research lies a simple but highly controversial idea: that all human thought processes have a “neural correlate” — a corresponding pattern of electrical and chemical activities in the brain.
For some, this idea is provocative because it implies that all our thoughts and feelings are little more than a complex set of chemical reactions — an idea that leaves little room for the concept of a soul.
For neuroscientists, however, this is exciting because it implies that they might be able to “read” all those processes, provided they can build instruments sensitive enough.
Cowen and his co-researchers, Brice Kuhl of New York University and Marvin Chun of Yale, believe that extracting facial images from people’s brains is just the first step in a process that will one day produce machines able to read minds in a far more detailed way.
“There is an ongoing philosophical debate over whether all mental phenomena are physical or at least some mental phenomena are nonphysical,” said Cowen.
“The prevailing answer among neuroscientists is that all mental phenomena seem to be manifested in brain activity, so the next ques- tion is whether we can access the neural correlate of a given mental phenomenon using current imaging methods.”
One idea, also controversial, is that such techniques could be used to probe the minds of people with
Something that looks like a high-definition movie of your dreams is not going to happen in the immediate future.
racial prejudices — to see if they perceive the faces of groups they are biased against differently from their own ethnic group.
Another might be to understand more of what happens in people with conditions such as autism who have problems with facial perception.
Kuhl said: “I study memory, and it’s hard not to be excited by the prospect of being able to reconstruct the images that we bring to mind when we remember something. We are certainly heading in the direction of reconstructing dreams too.
“Something that looks like a high-definition movie of your dreams is not going to happen in the immediate future, but we have already seen improvements in the sensitivity of these methods.”
Geraint Rees, director of the Institute of Cognitive Neuroscience at University College London, said the research was an “interesting technical advance.”
“This could be used to figure out what faces people are perceiving while they are dreaming, provided they were dreaming in a brain scanner.”
Gina Rippon, professor of cognitive neuroimaging at Aston University in Birmingham, praised the study, but said: “What would be really interesting would be to see if this technique could differentiate the same face but showing different emotional expressions. This would be a really insightful tool for the study of emotional perception problems in conditions such as autism and depression.”