Can artificial intelligence really replace the human brain?

This new study found that when simulating brain networks, by adjusting the electrical characteristics of individual cells, the learning speed of these networks is faster than that of simulating the same cells

They also found that the network requires fewer fine-tuned cells to achieve the same results, and the method consumes less energy than models with the same cells

The authors say that their findings can tell us why our brains are so good at learning, and may also help us build better artificial intelligence systems, such as digital assistants that can recognize voices and faces, or self-driving car technology

The first author Nicolas Perez, a doctoral student in the Department of Electrical and Electronic Engineering at Imperial College London, said: “The brain needs efficient energy while still being able to solve complex tasks outstandingly

The research was published in the journal Nature Communications

Why are neurons like snowflakes?

The brain is made up of billions of cells called neurons, which are connected by huge "neural networks" that enable us to understand the world

In contrast, each cell in an artificial neural network (the technology on which artificial intelligence is based) is the same, but their connectivity is different

They began to study whether simulating the brain by changing the properties of neural network cells can promote artificial intelligence learning

The lead author of the study, Dr.

To conduct this research, the researchers focused on adjusting the "time constant" - that is, each cell determines how fast it wants to do based on the actions of the cells connected to it

After changing the cell's time constant, they let the network perform some benchmark machine learning tasks: classify images of clothes and handwritten numbers; recognize human gestures; and recognize voice numbers and instructions

The results show that by allowing the network to combine slow and fast information, it can better solve tasks in more complex real-world environments

When they changed the amount of variability of the simulated network, they found that the best performing network matched the amount of variability seen in the brain, indicating that the brain may have evolved to have the amount of variability for optimal learning

Nicolas added: "We have demonstrated that artificial intelligence can be closer to how our brain works by simulating certain characteristics of the brain

"Next, we will study how to reduce the energy consumption of these networks and make artificial intelligence networks closer to the efficiency of the brain

“ Neural heterogeneity promotes robust learning ”by Nicolas Perez-Nieves, Vincent CH Leung, Pier Luigi Dragotti, and Dan FM Goodman, published 4 October 2021 inNature Communications .