By Jeffrey Welser
One of the watershed moments in the history of computing took place on Dec. 9, 1968. Douglas Engelbart and his team at Stanford Research Institute presented a technology demonstration that included the first public showings of the computer mouse, hypertext, dynamic file linking and shared-screen collaboration over a network. Those advances turned out to be essential building blocks for personal computing and Internet, and the event came to be called “The Mother of All Demos.”
While only history will say for sure, I think we saw the glimmer of a similar new beginning last week at IBM Research – Almaden, in Silicon Valley. The IBM Cognitive Systems Colloquium signaled a shift from a singular focus on the von Neumann computing architecture, which has dominated computer science and the computer industry since the mid-1940s, to new architectures modeled on the human brain.
To be clear, I’m talking about systems that augment human capabilities–not machines that do our thinking for us. Also, let me make it clear that while von Neumann-style computing faces major challenges, I don’t expect it to disappear. There are plenty of things it’s really good at.
The event brought together leading thinkers on brain-inspired computing from industry, government, philanthropy, and academia–including authorities in computer science and neuroscience. If this were a wedding, guests representing those two sciences would be sitting on the opposite sides of the aisle.
The presenters and panelists were top notch. They included Karlheinz Meier, the co-director of the European Commission’s Human Brain Project; Dharmendra Modha, an IBM Fellow who heads up our SyNAPSE Project; Fei-Fei Li, director of Stanford University’s AI lab; Horst Simon, deputy director of the US Lawrence Berkeley National Laboratory; and Miyoung Chun, EVP of the Kavli Foundation–who played a catalytic role in the US BRAIN initiative.
The audience was just as impressive. Nearly 300 people from over 70 organizations attended, drawn, it seemed, by a sense that something important was happening that they didn’t want to miss out on. They included Turing Award winner Ivan Sutherland, the Neurosciences Institute’s Einer Gall, Jeff Krichmar of the Cognitive Robotics Laboratory at UC Irvine, and Jeff Hawkins, the mobile computing pioneer who now heads up AI startup Numenta. In addition to Dharmendra Modha, four other IBM Fellows were in the audience: Hamid Pirahesh, Stuart Parkin, Ron Fagin and Chandrasekaran Mohan.
While the names were eye catching, the content of the presentations more than lived up to expectations. Our presenters demonstrated nothing less than the state of the art in brain-inspired computing. In addition, I believe, the discussions during the breaks allowed people to make personal connections that will bear fruit for years to come. For a more detailed look at the program, visit a blog post that captured the event in real time.
We focused part of the program on IBM’s SyNAPSE Project, which was funded by the US Defense Advanced Research Project Agency. The IBM team has produced a chip modeled on the human brain, TrueNorth, with 1 million neurons and 256 million synapses that only consumes 70 milliwatts of power. The chip could run non-stop for a week on an iPhone battery. We’re in the process of building an ecosystem to help take the SyNAPSE technology from science project to products and services in the marketplace–and we’re looking for partners in industry, academia and the startup community.
The chip is designed to augment the human senses. Think robots, self-parking cars, sensor networks on gas pipelines and wind farms, and public safety monitoring applications. And think about your smartphone being used as an always-on mobile sensing device for contextual computing.
This chip and the software tools and libraries they created to support it represent a major advance in the field of non-Von Neumann computing. That means the technology is not based on the architecture laid out in 1945 by American mathematician John Von Neumann, which became the core architecture for most computer systems built since then.
In the Von Neumann architecture, data is routed back and forth repeatedly between the logic, memory and communications elements of computing systems. That linear approach requires a lot of data movement, which consumes a lot of time and energy. Our TrueNorth chip interweaves logic (neurons), memory (synapses), and communications (axons), eliminating what’s known as the Von Neumann Bottleneck. TrueNorth embodies a parallel, distributed, modular, scalable, fault-tolerant, and event-driven architecture.
At last week’s colloquium, Horst Simon, Deputy Director of Lawrence Berkeley National Laboratory, spoke about his hopes for the new brain-inspired computing architectures, saying, “We need technologies like TrueNorth to take us into the future.”
At IBM, we certainly hope that TrueNorth and the work of our SyNAPSE team provides a pathway to the future of computing. But we recognize–and welcome–the fact that there will be many experiments tried, and, potentially, multiple paths forward as we move beyond the single-minded focus on Von Neumann computing.
What I’m certain of is this: brain-inspired computing will become an ever-more-important factor in computer science, science, technology, industry, government, and society–and, as a result, we’ll be able to achieve things in collaboration with computers that we never would have thought possible before.
To learn more about the new era of computing, read Smart Machines: IBM’s Watson and the Era of Cognitive Computing.