Now, a machine to read your dreams

 Scientists have developed the world's first “dream reader” that can tell what a person was dreaming of, with reasonable accuracy.

The dream-reading machine developed by Japanese researchers is not yet advanced enough to describe the exact narrative or sequence of events seen in deep sleep. But it can successfully tell the objects in the dream.

Even though it sounds like a lift from Leonardo DiCaprio starrer Hollywood flick “Inception”, the dream-reader’s 60 per cent success rate negates to a great extent chances of error or coincidences.

Using a functional magnetic resonance imaging (FMRI) machine that can map brain activities corresponding to human thought and machine-learning algorithm, they predicted the content of visual imagery on the basis of neural activity recorded during sleep.

Three volunteers were chosen for the experiment. They were asked to sleep inside a FMRI with a head phone to eliminate sound emanating from the machine. As each one drifted off, the team monitored their brain activity with the FMRI and an electro encephalograph (EEG).

Once the volunteers became used to take a catnap within the MRI, the researchers woke them up from their slumber every 6-7 minutes and asked them to narrate their dreams, which were matched against the neural activity seen on the machine.

The scientists did not wait for an hour till the volunteers went to REM (rapid eye movement) sleep, which is the period of deep sleep. They allowed those three individuals to go to a light slumber, woke them up for recording their statement and asked them to go back to sleep.

The experiments were conducted for 10 days and each volunteer had to sleep three hours inside the MRI chamber daily. While in one-fourth of the cases, there was no visual content, the rest provided information on the objects seen in the reverie. The corresponding neural patterns were also recorded and matched against a particular object.

All images were then grouped into 20 odd categories and computers were taught about those patterns using machine-learning algorithms.

After machines were coached adequately, volunteers were brought back into the MRI chamber. This time in their reverie when they visualised something, the machine could tell what they were seeing.
Once again the trio were awakened to share their experience. When cross checked against the machine-prediction, scientists found computers were accurate in 60 per cent of cases. Bingo, the dream-reader has arrived!

Reporting their findings in the April 4 issue of Science, researchers at ATR Computational Neurosciences Laboratory in Kyoto said the study would lead to a better understanding of the functions of dreaming and spontaneous neural events.

Although researchers warn they are still far from having a machine that can fully read our dreams, Robert Stickgold, a neuroscientist at Harvard Medical School in Boston who studies dreams, describes the work as “stunning in its detail and success,” says a report published in the same issue of Science.

“This is probably the first real demonstration of the brain basis of dream content,” Stickgold said, describing the results as “incredibly robust.”

But Allan Hobson, a psychiatrist at Harvard Medical School who too studies dreams, contended the images studied by the Japanese team were “hypnagogic hallucinations” with a different underlying physiology from the classic dreams that occur in REM sleep.

DH News ServiceBut Allan Hobson, a psychiatrist at Harvard Medical School who too studies dreams, contended the images studied by the Japanese team were “hypnagogic hallucinations” with a different underlying physiology from the classic dreams that occur in REM sleep.