To most people, predictions about machines soon becoming more intelligent than humans might seem like a far off possibility, but an announcement Thursday from IBM about its development of cognitive computing chips seems to bring us one very significant step forward in that direction.
IBM unveiled a working prototype of silicon computer chips Thursday that emulate the processes of the human mind rather than relying on the traditional architecture of computer chips, which have remained basically unchanged since the 1940s.
The research team’s design uses fewer transistors than traditional chips, and it features 256 digital processors that act as “neurons,” that do the computation, and synapses that learn and remember things. One of the chips has 262,144 programmable “synapses” and the other has 65,536 “learning synapses,” according to IBM.
Traditional integrated circuits feature many more transistors, and are programmed. The new IBM chips won’t be programmed in the same way, and they’ll process information differently.
One major breakthrough significance of the “neurosynaptic core” chip is that the technology could use much less power than computer chips do currently because they’re designed like a human brain (which is, believe it or not, extremely efficient at processing information.) They would also occupy less space than the supercomputers of today.
“This is a major initiative to move beyond the von Neumann paradigm that has been the ruling computer architecture for more than half a century,” said IBM’s Project Leader Dharmendra Modha, in a press statement. “Future applications of computing will increasingly demand functionality that is not efficiently delivered by the traditional architecture. The chips are another significant step in the evolution of computers from calculators to learning systems, signaling the beginning of a new generation of computers and their applications in business, science and government.”
John von Neumann was a polymath and mathematician who laid the foundations for modern computer chip architecture.
One of the first things the researchers apparently had the chips do to demonstrate their intelligence is to play Pong.
The researchers didn’t provide any details about this, but headlined a blog post announcement about the development of the chips with: “This cognitive computing chip taught itself how to play Pong.”
What the researchers were getting at is that the chips have an intelligence that can learn from their environment, and adapt.
“Cognitive computers are expected to learn through experiences, find correlations, create hypotheses, and remember — and learn from — the outcomes, mimicking the brains’ structural and synaptic plasticity,” says IBM.
In its announcement, IBM said that its research team “successfully demonstrated simple applications like navigation, machine vision, pattern recognition, associative memory and classification” with two prototype designs of the chips.
To come up with the design, the researchers said that they were inspired by neurobiology, and combined principles from nanoscience and supercomputing, and went from there.
They said that such chips could be deployed in applications like monitoring the world’s water supply for various metrics that could ultimately predict conditions that might spark off a tsunami, or conditions at road junctions that are likely to cause a higher rate of traffic accidents.
“Imagine traffic lights that can integrate sights, sounds and smells and flag unsafe intersections before disaster happens or imagine cognitive co-processors that turn servers, laptops, tablets, and phones into machines that can interact better with their environments,” Modha said.
The project is a collaboration between IBM and researchers from five U.S. universities: Columbia, Cornell, University of California, Merced and the University of Wisconsin, Madison.
Their progress has also landed them additional funding from the Defense Advanced Research Project Agency to the tune of an additional $21 million for further development. That brings the total cost of the project so far to $41 million, according to The New York Times.
The details of the announcement, and the goals of the project are quite mind-bendingly ambitious.
There is “nothing even close” to the level of sophistication in cognitive computing as this project, Richard Doherty, an analyst at the Envisioneering Group told VentureBeat’s Dean Takahashi.