Artificial intelligence can use your brainwaves to see around corners. The technique, called 鈥済host imaging鈥�, can reconstruct the basic details of objects hidden from view by analysing how the brain processes barely visible reflections on a wall.
Ghost imaging has been used before to聽reveal objects hidden around corners and normally relies on using video recordings of faint reflections cast by an object onto a nearby wall. 听补苍诲听 at the University of Glasgow, UK, have now replaced the video component with electroencephalography (EEG) brain scans.
In their experiment, a single person wearing an EEG headset connected to a computer stands in front a white wall and next to a wall painted grey, which obscures the view of an object and a projector. This projector is controlled by the computer and casts a series of special patterns onto the object.
Advertisement
Some of this patterned light reflects off the object and hits the white wall or diffuses through the room. The person can鈥檛 see the object in the reflections. However, a ghost-imaging machine-learning algorithm can build a basic 16-by-16 pixel image of the object using the EEG data.
The algorithm is partly guided by a feedback loop from the brain. While the algorithm is assembling the image, it uses the EEG readings to adjust the projected pattern to stimulate the person鈥檚 eyes better. If a pattern doesn鈥檛 result in enough EEG signal, the computer switches to a different one in real time.
鈥淲e could have done this with a camera [instead of EEG],鈥� says Faccio. 鈥淏ut the point is that you can actually do this with the human brain, which was not obvious because the brain is very different from a camera.鈥�
The new technology is a proof-of-concept for how AI can help people carry out tasks they normally cannot. Faccio says the human brain simply doesn鈥檛 have the kind of algorithmic capability to perform the calculations the AI does in the experiment, even if all the information needed for those calculations is already stored in the brain.
at SUNY Downstate Health Sciences University in New York City says the human brain doesn鈥檛 use all visual information that enters the eye in order to produce images. In fact, the way we normally see is tailored for the brain to use very little energy instead of processing every single stimulus. Many of the brain processes that help us see are subconscious and inaccessible to our conscious thought, he says.
at Glendon College in Toronto, Canada, says that, in contrast, EEG doesn鈥檛 discriminate between conscious and unconscious activity in the regions of the brain that process visual images. In other words, these readings contain information that we aren鈥檛 aware of.
Faccio says that the next step for the research may be trying to image hidden objects faster by connecting the computer with multiple people wearing EEG headsets or to try to differentiate between activity in the visual cortex that we are conscious of and that which is subconscious. Ultimately, Faccio and Wang would like to build a wearable device that allows the wearers brain to collaborate with an AI.
Faccio will present this work at the in Vancouver, Canada, on 11 July.
