What is the potential of neuromorphic computing?


At the end of September, Intel unveiled the Loihi 2 neuromorphic chip. Made to resemble the human brain in its functioning, this type of processor is arousing the curiosity of researchers and companies, who already see potential applications in various fields.

As its name suggests (or not), a neuromorphic chip mimics the functioning of the human nervous system and in particular that of the brain. A traditional computer has a so-called “Von Neumann” architecture, that is to say that the memory part (the storage) and the processor part (the calculations) are separate. Information is processed sequentially and synchronously, based on a regular clock rate. It is an operation which differs from the human brain by its rigidity.

In the case of a neuromorphic processor, the different “neurons” (one million for Loihi 2) communicate with each other, allowing information processing in parallel, asynchronous and based on events. Clearly, neuromorphic processors having the ability to process information irregularly, they have a better ability to adapt to events. Beyond increased power, this chip has several advantages.

More powerful, faster, less energy intensive

The fact of being able to calculate a large amount of information in parallel at any time is an evolution in terms of the speed of execution of calculations and the complexity of these. Using more complex algorithms can be up to ten times faster. The computer will thus be able to learn and adapt with less latency, even in real time.

The Loihi 2 neuromorphic chip. © Walden Kirsch / Intel Corporation

This type of communication directly between artificial neurons is therefore more efficient than the communication of traditional computers, where information must constantly go back and forth between memory and the processor. This efficiency also translates into lower energy consumption.

Several projects already under study

Among the sectors interested in neuromorphic computing, there is first of all medicine. In 2020, Intel supported a project launched by a university and pediatric hospital in Israel: a robotic arm mounted on a wheelchair to help people with disabilities with certain tasks of daily living. The use of the Loihi chip should allow the creation of an intelligent robotic arm that adapts to its user. Another big advantage is the low power consumption of neuromorphic chips, since it would allow the price to be divided by 10 and the device’s battery to be recharged more quickly. Other medical uses are envisioned, most particularly for prostheses, or for adapting the dosage of drugs in real time as a function of factors such as the level of glucose or of insulin.

© ALYN Hospital

Always on the lookout for technological innovations, the field of security also sees several potential uses of neuromorphic processors. According to robotics researchers at the University of Zurich, they could make drones more autonomous and more precise, for example to explore a disaster area and spot victims.

In 2020, researchers at Intel Labs and Cornell University succeeded in providing the first version of Loihi with a “smell” to warn of the presence of dangerous chemicals. This capacity could be expanded to find drugs and explosives or to diagnose certain diseases.

Cybersecurity is no exception, since the ability to learn in real time would be of great help in detecting viruses and other malware while consuming less energy. A characteristic which is important for this type of software, which must be permanently active.

The automobile industry could also use this type of chip for security reasons, in particular for emergency braking systems or to detect if a part of the car is defective.

Today, these chips remain prototypes and are not marketed. Intel has nevertheless taken an important step in this direction with Loihi 2, by making free software available to researchers and developers.

The fact that a computer can function similarly to the human brain, especially if it begins to have senses such as smell and touch, can cause concern and raise social and of course ethical issues. Rest assured: there is no question here of computers provided with emotions, let alone consciousness. They remain computers, at least for now.

Leave A Reply

Your email address will not be published.