Human brain still learns faster than AI, Oxford researchers claim
Despite swift progress of machine algorithms, computers cannot think as efficiently as people
THE human mind remains superior and more efficient than artificial intelligence, a study has found.
Research by the University of Oxford suggests that brains work in a quicker and fundamentally different way than machine learning algorithms. Animals, including humans, can learn something after seeing it once and do not need to spend any more time or energy solving the problem.
However, artificial intelligence (AI) relies on refining all the possible solutions to a problem to reduce the chances of an error before settling on the final answer.
This trial-and-error approach is known as “backpropagation”. Computers were built to do this and it was believed that humans thought in the same way.
However, the Oxford study found that humans employ a mental technique called “prospective configuration”, in which the brain lays out neurons in a way that enhances learning and makes it easier to remember something.
This ability speeds up the learning process, scientists claimed. It also allows for more information to be absorbed without corrupting previous lessons learned – something that machine learning algorithms struggle to achieve.
Dr Yuhang Song, the first author of the study, said that computers were slow at the human way of learning “because they operate in fundamentally different ways from the biological brain”.
However, the Oxford scientists believe that it is possible to apply the process to artificial intelligence, to create a computer brain that combines the power of AI with the efficiency of human minds.
“A new type of computer or dedicated brain-inspired hardware needs to be developed that will be able to implement prospective configuration rapidly and with little energy use,” Dr Song added. The idea around different approaches to soaking up new information was theorised after research showed that allowing neurons to settle into a prospective configuration reduces interference.
It was also said to better explains data from previous studies on neural activity and human behaviour.
The scientists say the different approaches can be thought of by imagining two bears approaching a river full of fish, one with an AI brain and one with a natural brain.
Both the animals can hear the water and smell the fish.
The AI bear processes the sound and the smell together to deduce that the river contains food.
On the other hand, the natural brain can separate these two senses. If these bears went deaf, the AI bear would think that because there is no sound, there is no food.
In contrast, the organic brain is able to realise that although the sound has gone, the smell is still there and the food is still present.
The research is in its early stages, and the team hopes that further work can unpick more of the details on how brains work and what underpins the nimble, flexible thinking the human brain can carry out.
Prof Rafal Bogacz, the study lead researcher, works in MRC Brain Network Dynamics Unit and the University of Oxford’s Nuffield Department of Clinical Neurosciences.
He said: “There is currently a big gap between abstract models performing prospective configuration and our detailed knowledge of anatomy of brain networks.”
Prof Bogacz added: “Future research by our group aims to bridge the gap between abstract models and real brains, and understand how the algorithm of prospective configuration is implemented in anatomically identified cortical networks.”
The results of the research were published in the journal Nature
Neuroscience.
‘Computers operate in different ways from the biological brain’