Scientists warn AI must be given ability to ‘sleep and dream’ to avoid ‘catastrophic phenomenon’

Tech experts are expressing the importance of computer owners giving their devices power-off time, rather than conveniently putting them into sleep mode.

But now, it seems AI might need the same treatment.

Scientists have recently discovered that AI might work better if it's given time to catch some z's and even dream.

A team of researchers, including Professor Concetto Spampinato from the University of Catania in Italy, proposed a theory that suggested AI could benefit from 'sleep' periods, like humans, and perform better than AI models that don't need to sleep.

AI performs better after a periods of rest / Yuichiro Chino / Getty

The researchers wrote: 'We propose Wake-Sleep Consolidated Learning (WSCL), a learning strategy leveraging Complementary Learning System theory and the wake-sleep phases of the human brain to improve the performance of deep neural networks for visual classification tasks in continual learning settings.'

The team hopes to combat a phenomenon called 'catastrophic forgetting' which means an AI model can forget previously learned abilities when it is taught new ones.

More specifically, New Scientist explained: 'A model trained to identify animals could learn to spot different fish species, but then it might inadvertently lose its proficiency at recognising birds.'

To test their theory, the researchers trained the AI models on datasets during an 'awake' phase before going into periods of 'sleeping' - in a way similar to dreaming in humans.

During the 'sleep' phase, AI was shown images to see if it could recall what it previously learned.

Interestingly, allowing AI models to 'dream' resulted in 'significant performance gains.'

Without rest, AI could experience 'catastrophic forgetting' / zf L / Getty

There's a lot of research out there about the importance of sleep, in that humans consolidate a collection of short-term memories throughout the day and consolidate them into long-term memories. Similarly so, AI seemed to remember tasks better after periods of rest.

However, despite the results, some experts think the findings should be taken with a pinch of salt, and don't see crossing the behaviours of humans and AI as the way forward for AI development.

According to expert Andrew Rogoyski: 'The human brain should not be regarded as the ultimate architecture for intelligence.

'It’s the result of millions of years of evolution and an unimaginably wide range of stimuli.

'We may develop AIs that have structures completely different from their biological designers.'

Furthermore, physicist Stephen L. Thaler, president and CEO of Imagination Engines, cautions against taking the term 'sleep' too literally when it applies to AI.

'Instead, it needs to cycle between chaos and calm,' he highlighted.