The technology giant IBM is currently developing a prototype of a chip similar to a human brain and states that within the framework of this project, the issue of improving the energy efficiency of artificial intelligence is being solved.
Concerns have been raised in public space that warehouses filled with computers powering machine intelligence systems could be a source of emissions. IBM claims that the prototype being developed can become the basis for creating AI chips for smartphones, which are characterized by a higher level of efficiency and do not drain the battery so quickly.
The technology giant stated that the quality of its development depends on components whose functioning, from the point of view of the algorithm, is similar to the connections in the human brain.
Scientist Thanos Vasilopoulos from the IBM research laboratory in Zurich says that the human brain, unlike traditional computers, can demonstrate a high level of performance while consuming a small amount of energy. He stated that increasing energy efficiency means that large-scale and complex workloads will be carried out with low energy consumption or with a limited number of batteries, for example, in cameras, smartphones, and vehicles.
The scientist also said that cloud providers will be able to use chips developed by the technology giant to reduce energy costs and reduce carbon footprint.
Most of the chips are digital. This means a method of storing information in the form of 0 or 1. The new chip uses memory resistors, often called memristors, which are analog and can store a range of numbers. The algorithm of memristors is similar to the principle of functioning synapses in the brain.
Professor Ferrante Neri from the University of Surrey says that memory resistors are what can be described as natural calculations that mimic the work of the brain. Potentially, a memristor can store the history of interaction with information in a similar way to the synapse mechanism in a biological system. He also noted that interconnected memory resistors can create a network similar to the human brain. The professor admitted the possibility that IBM’s efforts are bringing the appearance of chips, which are in many ways similar to the brain.
At the same time, Ferrante Neri stated that creating a computer based on memristors is a difficult task, and widespread implementation in this case means the need to solve several problems, including significant costs for the purchase of materials and certain production difficulties.
Memory resistors make the new chip more energy efficient, but it also contains digital elements. This facilitates the implementation of new developments in existing artificial intelligence systems.
Currently, many smartphones are equipped with AI chips. These components are used, for example, for photo processing. IBM hopes that in the future, chips in smartphones and cars will become more efficient, and provide a longer battery life.
The new chips, similar to the tech giant’s prototype, could save a lot of energy in computers that provide applications with powerful artificial intelligence systems. They could also reduce water consumption for cooling energy-intensive data centers.
James Davenport, a professor of information technology at the University of Bath, said that the IBM prototype is a potentially interesting development, but noted that this is only the first step towards solving a complex problem.
As we have reported earlier, IBM Intros Slew of New AI Services.
