Image: Thinkstock
Intel is reportedly preparing to fabricate “Loihi,” a self-learning “brain chip” that mimics how the human intellect functions, as a foundation for further developments in artificial intelligence.
Named after an active undersea volcano south of the island of Hawaii, Intel said in a statement Monday that Loihi includes a total of 130,000 silicon “neurons” connected with 130 million “synapses,” the junctions that in humans connect the neurons within the brain. The Loihi chip, which Wired reported will be manufactured next month on Intel’s 14-nm process technology, will be shared with leading universities and research institutions next year in a bid to advance AI development, Intel said.
Coincidentally, Microsoft said Monday that it, too, is working on ways to develop new avenues for alternative computing, including manufacturing actual chips and systems, as well as developing software to power quantum computers. Intel’s Loihi doesn’t employ quantum techniques, but it has a similar goal: Instead of trying to “brute force” its way into a solution, as traditional chips do, Loihi tries to mimic the parallel structure of the human brain to arrive at those same answers much more efficiently.
Intel said it believes the Loihi chip could be used autonomously, A Loihi-powered medical device, for example, could determine what a “normal” heart rate was and therefore be able to figure out when an abnormal heart condition presented itself.
Intel emphasized that its chip is intended to “self-learn,” teaching itself the answers to problems, using its digital array of synthetic neurons and synapses. Each “neuromorphic core” includes a learning engine that can be programmed to adapt. “As AI workloads grow more diverse and complex, they will test the limits of today’s dominant compute architectures and precipitate new disruptive approaches,” Michael Mayberry, corporate vice president and managing director of Intel Labs, wrote in a blog post.
Mayberry said Intel believes its Loihi implementation could be a million times faster than other typical spiking neural nets. “Looking to the future, Intel believes that neuromorphic computing offers a way to provide exascale performance in a construct inspired by how the brain works,” he wrote.
Synthetic brains developed in silicon aren’t totally novel. IBM’s DeepNorth brain chip uses 1 million neurons and 256 million synapses, the company said in 2016. And in 2012, Google and Stanford researchers developed a brain-like neural network for facial recognition.
What this means to you: Loihi is one of many new technologies vying to be the future of computing, as Moore’s Law loses its momentum. We’ve seen more and more chip generations—including Intel’s recent 8th-gen Core chips—manufactured on the same process technology as their predecessors. That why chipmakers have sought alternatives: more cores, quantum computing, and the like. There’s no guarantee that digital brains will be developed fast enough to keep up with the breakneck pace of Intel’s more traditional PC processor design teams. But, just as Intel’s chips are processing more in parallel, so are Intel’s researchers. Who knows which will win?
