A computer that stores and processes data on a single chip can distinguish between different people’s handwriting.
Professor Shahar Kvatinsky of Technion in Haifa says that the combination of RAM and CPUs can act like the human brain, opening up a whole new world of opportunities.
To demonstrate its capabilities, he and his team presented the chip with various examples of handwriting. The chip learned which was which in each example and achieved 97% recognition accuracy.
“We like to describe computers as ‘brains,’ but completely separate hardware for storing and using information is not an organic brain,” Kvatinsky said. says.
Chips that can both store and process information have many potential uses. For example, a chip could be built into a smartphone’s camera sensor to save energy every time someone takes a picture.
Smartphones typically have to convert visible objects into digital data every time they take a picture, which consumes energy. This chip allows the phone to process raw images and then store them in a compressed digital format.
Neurons in the human brain build stronger connections each time a person learns. Thus, the learning process strengthens the conductivity of new artificial neural networks.
A traditional computer has two pieces of hardware. Information is stored in memory, RAM, or hard drives, and processed by the CPU or processor.
Traditional computers can surpass the human brain only by relying on many processors and using a lot of energy. For example, a computer program to beat a chess master had to use 1,500 processors, and each game cost him $3,000 in electricity.
Human players expended much less energy in the same game. At the same time, I was also able to do multiple things at the same time.
“Commercial companies are in constant competition to improve their products,” says Professor Kvatinsky.
“They can’t afford to go back to the drawing board and rethink the product from scratch. You can release it when it’s done.”
His work was published in an academic journal nature electronics.