AI must be given the ability to 'sleep and dream' just like humans to prevent 'catastrophes'
ARTIFICIAL intelligence can learn better if it dreams and sleeps like a human, researchers say.
Scientists hope to create AI that can replicate this human behavior, so it can learn to perform tasks better than models that don't need time to sleep.
A team of researchers, including Professor Concetto Spampinato from the University of Catania in Italy, outlined this theory in a new study.
“We propose Wake-Sleep Consolidated Learning (WSCL), a learning strategy that leverages complementary learning system theory and the wake-sleep stages of the human brain to improve the performance of deep neural networks for visual classification tasks in continuous learning environments” , they wrote. .
The team hoped to prevent a phenomenon called “catastrophic forgetting.”
That happens when an AI model that has learned new tasks forgets everything it has learned.
This can happen when an AI is trained to learn something additional to a task at which it is already proficient.
The researchers conducted an experiment to demonstrate the “importance of dreaming” and noticed “significant performance gains.”
Just as humans form long-term memories during their sleep, AI was found to remember tasks better after a rest phase.
The models were trained on datasets during an 'awake' phase before entering a 'sleep period'.
During the sleep period, the AI models were shown images of what they had learned to help them remember.
This was said to be similar to humans during a dream phase.
Despite the positive results, some experts don't think that mimicking the human brain is the way forward for AI development.
According to the New Scientist, expert Andrew Rogoyski said: “The human brain should not be considered the ultimate architecture for intelligence.
'It is the result of millions of years of evolution and an unimaginably wide range of stimuli.
“We can develop AIs that have completely different structures than their biological designers.”