Scientists unite with AI to record dreams

Yahoo

19-06-2025

Dreaming is a fascinating state where creativity runs wild, crafting vivid, cinematic scenes that can feel as real as everyday life—until you wake up and they vanish into memory or disappear from your recollection altogether.
But what if there were a way to capture, record, and playback your dreams in the real world?
At ATR Computational Neuroscience Laboratories in Kyoto, Japan, a group of scientists unveiled a complex system that uses functional magnetic resonance imaging (fMRI) and AI to 'record' people's dreams and play them back in a rough, estimated form.
The experiment focused on closely observing the brain's electrical and blood flow activity in a select group of volunteers as they slept. To achieve this, researchers combined an electroencephalogram (EEG) with an fMRI machine, allowing participants to spend several nights sleeping while their brain activity was recorded.
During the REM stage of sleep—when brain activity surges and dreams unfold—participants were gently awakened and asked to recount their dreams. This process was repeated hundreds of times, creating a rich database that linked distinct brain-scan patterns to specific dream imagery.
Using this carefully constructed database, scientists employed deep learning algorithms to attempt to reconstruct visual content while the volunteers slept.
During the tests, the system achieved approximately 60% accuracy, rising to over 70% on more specific categories like people or objects. Professor Yukiyasu Kamitani stated, 'We were able to reveal dream content from brain activity during sleep, which was consistent with the subjects' verbal reports.'
The idea of watching your dreams play out like a vintage film is undeniably captivating, but the implications for neuroscience and mental health are equally remarkable. In the future, this innovative approach could be harnessed to explore emotional and cognitive patterns in various mental health conditions, providing a non-verbal and impartial glimpse into the subconscious. As the technology advances, it holds the potential to deepen our understanding of both human consciousness and the subconscious mind.
However, while the technology is still in its early stages, the results it produces are often blurry and indistinct. Key elements such as colour, motion, narrative structure, and emotion remain largely elusive for now. Currently, dreams can only be captured in retrospect, after the subject has awakened, rather than in real-time.
Dr. Mark Stokes points out that "all of this would have to be done within individual subjects," meaning that a universal classifier capable of interpreting anyone's dreams is not feasible. In essence, a dream-reading machine is not a one-size-fits-all solution.
Nonetheless, this experiment represents a significant leap forward in the field of "oneirography," which involves the recording and documentation of dreams. As advancements in fMRI, EEG, and AI continue, the boundary between our internal mental experiences and external observation is starting to blur.
"Scientists unite with AI to record dreams" was originally created and published by Verdict, a GlobalData owned brand.
The information on this site has been included in good faith for general informational purposes only. It is not intended to amount to advice on which you should rely, and we give no representation, warranty or guarantee, whether express or implied as to its accuracy or completeness. You must obtain professional or specialist advice before taking, or refraining from, any action on the basis of the content on our site.