Researchers at the University of Texas A & M, inspired by the low -use performance of the human brain, have iroduced a new approach to artificial ielligence design that greatly reduces the energy consumption of this technology.
Artificial ielligence today can do complex calculations and data analysis at a rate beyond humans, but this ability requires a lot of cost and energy. In corast, the human brain consumes only 20 watts of energy despite the extraordinary processing power, but host dataceers such as ChatGPT require several gigawatts of electricity.
Scieists’ new achieveme: a model to imitate artificial ielligence from the human brain approach
Researchers have a new artificial ielligence called the new artificial ielligence “Super-TURING” They have been developed that operates with a function similar to the human brain. The main purpose of this artificial ielligence is to optimize its energy consumption.
These days, the energy crisis has greatly preveed artificial ielligence from progressing, and companies such as Openai require enormous computation and high power consumption. This can even be involved in global warming and rising global temperatures.
Dr. Swin Yi, Assista Professor of Electrical and Computer Engineering, explains:
“Data ceers of artificial ielligence require gigawatti, but the human brain consumes about 2 watts; That is, one billion watts against only 2 watts. These data ceers are not sustainable by curre computational methods. “Although artificial ielligence capabilities are amazing, we still need hardware and high energy production to support its performance.”
Texas University researchers believe that the key to solving this problem is in nature, especially in the nervous processes of the human brain. In our brain, the performance of learning and memory is not separate, but iegrated. All of these processes are based on communication between neurons. In fact, by learning new skills or recording a memory, the human brain forms a kind of neurons between neurons; Therefore, when using that skill and memories, these nerve pathways are activated and we can use them.
This is the opposite in artificial ielligence and each part of the hardware has its own activity. Researchers at Texas State University claim that Super-TURING artificial ielligence inspired by the function of the human brain is filling this gap. By eliminating such a boundary, artificial ielligence can coinue to be efficie and powerful by consuming less resources.
In the test of this artificial ielligence, Super-Tering was able to corol a drone without prior training. Of course, the start of corol of the UAV was not perfect, but this artificial ielligence over time was able to adapt to the environme and learn from its mistakes. This approach is faster, more efficie and optimized than traditional artificial ielligence.
Perhaps if the Super-Tering artificial ielligence becomes a final product, it will make a revolution in this area. Unnecessary investmes in artificial ielligence have even led promine technology to be criticized, such as Alibaba’s co -founders; This artificial ielligence may reduce the need for more costs.
The results of the prese study are published in Science Advances.
