Artificial intelligence, deep learning... everyone wants a piece of the pie. The potential is enormous, as is the number of applications, but any superlative advance in software requires a similar leap at the hardware level. That brings us to Intel, which has just announced Loihi, a chip capable of imitating the human brain and learning from the feedback it receives from its environment, without classic training.
Anyone can imagine that the future will be full of things like autonomous cars and visual recognition... but that already exists today. Data multiplies, requires increasingly complex analysis, and traditional systems find it harder to respond to such demand. That brings us to the territory of quantum computers, artificial intelligence, deep learning, and neural networks. Advanced solutions for advanced problems place enormous pressure on the industry, with no respite. The human brain is the main source of inspiration, and there are many who seek to reproduce (up to a point) its functioning. In the past we have talked about several projects at IBM Research, but today it is Intel’s turn with its new chip Loihi.
Loihi has been in the ovens of Santa Clara for more than six years, but at the same time it represents several decades of study in neuromorphic computing. According to the official description, the chip mimics brain function by learning through various modes of feedback in its environment. Loihi uses the data it receives to make inferences and improve without the need for conventional training. Its design is based on the current 14-nanometer process, with 130,000 neurons and 130 million synapses. Manufacturing of Loihi will begin next November, and it will be as small as a pinky fingernail. For now there are only two prototypes, and even with favorable winds, we will have to wait two years for it to debut on the market.
Loihi is clear proof that Intel wants to become a giant in artificial intelligence. Last July we discussed the Movidius modules, but despite the convenience of that format for researchers, the company needs to aim much higher and offer very robust solutions.
