Presented neuromorphic learning chip, able to compose music

Presented neuromorphic learning chip, able to compose music


At a recent forum IMEC ITF2017 the representatives of the Belgian research and education center demonstrated the first of its kind self-learning neuromorphic chip, the principles of which imitate the principles of operation of the human brain. The chip has the ability to self-education, which is enough for “the works” chip simple musical compositions.

At the core of the chip is the technology OxRAM. Specialists of the research center IMEC have created a “building blocks” of hardware and software systems of the simplest artificial intelligence. The combination of these funds allowed us to put on a small crystal chip that consumes low amounts of energy, function, machine learning and self-study. Using these functions, the chip itself carries out the analysis of the data, and the greater the number of associations is found in the data, the stronger will be the connection between the individual elements of the chip. Thus, the “inside” of the chip is formed what kind of neural circuitry.

Crystal chip is manufactured on a 65 nm technology. It contains elements of traditional CMOS logic and memory arrays of MRAM (Metal Oxide Resistive RAM). This memory is 100 times more efficient from an energy point of view than “normal” types of memory. Due to the large amount of memory, the chip can create and use 40 stable associative relations.

During the demonstration, the chip “learned” to compose music, receiving data of several musical works. The rules of composition, and genre chip “invented” himself. As one of the staff of the centre IMEC, Parvin Ragavan,

“Now we already have the necessary hardware, software package and programming environment that will move far ahead in the area of neuromorphic computing. And now we had the opportunity of creation on the basis of the new chips complex intelligent systems. The level of intelligence of the system will grow in quadratic or even exponential, depending on the number included in neuromorphic chips”.