Generating images of what someone is thinking, shown on screen. It sounds like science fiction, but it is now on the road to becoming science fact. Already, images have been generated albeit at low resolution bearing definite resemblance to that of the dream.
Achievements have been made both through three methods, two of which rely on raw computing power, and the other on clever experimentation.
Firstly, there was a need to establish whether the brain gives out the same signals for a particular thought or action whether it is awake or asleep. This was achieved by recruiting people known as ‘lucid dreamers’ who are able to both be asleep yet partially awake at the same time to be able to signal that they are dreaming. Measurements on these volunteers showed that indeed mental signals were the same for the brain whether it was ‘asleep’ or ‘awake’
The second experiment consisted of showing volunteers thousands of film clips of many varied types, and monitoring their mri (magnetic resonance imaging) responses produced by their brains. A computer looked for correlations between specific images or sequences and the mri responses obtained. They then fed their computer 5000 hours of You Tube sequences and asked the machine to predict, based on the correlations what the matching MRI response would be. Finally the volunteers were asked to watch a further two more hours while the computer picked its own sequences based on the new mri signals it received. . The sequences picked were melded together to produce an estimate of what was being watched looked like. The results often gave recognizable simulations on screen (see gallantlab.org )
The third study by Francisco Pereira and colleagues at Princeton University, used a similar method to the one above but successfully attempted to see what topics his subjects were pondering. His method was to examine data conducted in an experiment 4 years previously where volunteers were shown 60 labeled objects. They were then asked to imagine the same objects while they had their brains scanned Again, a portion of the results were used to program his computer while the other half were used to see what the computer generated itself on being given the scans. While the accuracy was found not to be at a level to totally identify the object, it was enough to determine what type of object it was i.e. it could not distinguish a carrot from a celery stick, but could determine it was a vegetable.
These are early days with studies performed in a fairly limited way. Both the last two studies however could be repeated at many times greater numbers, which would no doubt increase ‘focus’ considerably.
‘Mind reading’ by machine has thus started and will only become more sophisticated and accurate. While it may have worrying implications in the wrong hands, it also has considerable potential to assist mankind. It may, for example unlock those who have suffered stroke or injury and are locked in their own minds. It could help the disabled to lead more normal lives. As with many scientific breaks through it will be up to society to decide their use, controls and limits.