The expert explained the difference between the human brain and AI
Modern artificial intelligence (AI) systems consume significantly more energy than the human brain because they use a fundamentally different approach to information processing. Alexander Kugaevskikh, PhD, Associate Professor at the Faculty of Software Engineering and Computer Technology, head of the Master's degree program in AI Systems Design and Development at ITMO University, told Izvestia about this.
According to the expert, the human brain consumes about 20 watts due to the fact that memory and computing are combined in it, whereas modern computers spend a significant part of their energy transferring data between memory and the processor.
"In the brain, memory and computing are combined. Synapses simultaneously store information and participate in signal transmission. This eliminates energy losses for data transportation, which in modern computers can account for up to 80% of the total energy consumption of the system," said Kugaevskikh.
The expert also stressed that the principle of operation of biological neurons differs significantly from artificial intelligence algorithms. Neurons are activated only when necessary, whereas graphics processors continuously perform billions of mathematical operations, even if some of the signals do not carry useful information. In addition, the brain simultaneously uses about 86 billion neurons and about 100 trillion synapses, providing a high degree of parallel data processing with minimal energy consumption.
Another reason for the high energy consumption of AI, the scientist called the learning process of modern models, which requires a large number of iterations using the error backpropagation algorithm.
"We can remember something by looking at an object once. Modern artificial intelligence models are trained using multiple forward and backward computing passes, which requires significantly higher energy costs. The brain uses other learning mechanisms, such as synaptic plasticity, so it does it much more efficiently," the expert explained.
Read more in the exclusive Izvestia article:
Between lines and neurons: Can AI replace the human brain
Переведено сервисом «Яндекс Переводчик»