Innovation through Collaboration: Building Research Communities

By Rich Uhlig

We find ourselves in a world awash with data growing at an unprecedented rate. And that data is changing in nature with respect to how and where it is collected as we blanket the planet with new sensing technologies. Cameras, microphones and location sensors in our smart phones, attached to our vehicles and embedded in the environments in which we live and work are shifting the flows of data between centralized data centers and the network edge.

At the same time, Moore's Law continues to expand what is possible to build in computing systems. The question is less about whether we can expect the economic benefits of Moore's Law to continue (i.e., further reductions of cost per device over time) and more about whether we can manage the design and programming complexity that it affords.

More:Intel's Pohoiki Beach, a 64-Chip Neuromorphic System, Delivers Breakthrough Results in Research Tests| Intel Labs (Press Kit)

As I discussed during my presentation at the DARPA Electronics Resurgence Initiative (ERI) Summit in Detroit this week, continuing to achieve the types of gains Moore's Law enables requires work that cannot be done in silos. Aligned to the ERI mission, Intel is pushing this forward through industry and academic collaborations. To move from lab research to real-world impact requires a truly integrated systems view and a commitment to collaboration across a spectrum of research areas.

That is our approach at Intel Labs. We start with a bold mission, explore solutions, create prototypes and nurture communities to reach the end goal. We work with academia, industry and government organizations such as DARPA at various steps of this process. Together, we explore the development of architectures that use specialized hardware - and software - to solve computing problems more quickly and efficiently. Doing so requires opening our innovations to the broad research community to push every breakthrough further.

• A close-up photo shows Loihi, Intel's neuromorphic research chip. Intel's latest neuromorphic system, Pohoiki Beach, will be comprised of 64 of these Loihi chips. Pohoiki Beach was introduced in July 2019. (Credit: Tim Herman/Intel Corporation)
• A close-up shot of an Intel Nahuku board, each of which contains 8 to 32 Intel Loihi neuromorphic chips. Intel's latest neuromorphic system, Pohoiki Beach, is made up of multiple Nahuku boards and contains 64 Loihi chips. Pohoiki Beach was introduced in July 2019. (Credit: Tim Herman/Intel Corporation)
• One of Intel's Nahuku boards, each of which contains 8 to 32 Intel Loihi neuromorphic chips, shown here interfaced to an Intel Arria 10 FPGA development kit. Intel's latest neuromorphic system, Poihoiki Beach, annuounced in July 2019, is made up of multiple Nahuku boards and contains 64 Loihi chips. Pohoiki Beach was introduced in July 2019. (Credit: Tim Herman/Intel Corporation)
• Rich Uhlig, managing director of Intel Labs, holds one of Intel's Nahuku boards, each of which contains 8 to 32 Intel Loihi neuromorphic chips. Intel's latest neuromorphic system, Pohoiki Beach, is made up of multiple Nahuku boards and contains 64 Loihi chips. Pohoiki Beach was introduced in July 2019. (Credit: Tim Herman/Intel Corporation)
• Researchers at the 2019 Telluride Neuromorphic Cognition Engineering Workshop are working to automate Western Sydney University's foosball table under Loihi control, operating on visual input from event-based cameras. Foosball offers an excellent test for rapid closed-loop sensing, planning and control algorithms, a sweet spot for neuromorphic technology. (Credit: Sumit Bam Shrestha)
• Rachel Gehlhar of Caltech's AMBER Lab and Terry Stewart of Canada's National Research Council are working to control the AMPRO3 prosthetic leg with Intel's Kapoho Bay Loihi device so that the leg can better adapt to unforeseen kinematic disturbances while walking. (Credit: Sumit Bam Shrestha)
• Rich Uhlig, managing director of Intel Labs, holds one of Intel's Nahuku boards, each of which contains eight to 32 Intel Loihi neuromorphic chips. Intel's latest neuromorphic system, Pohoiki Beach, is made up of multiple Nahuku boards and contains 64 Loihi chips. Pohoiki Beach was introduced in July 2019. (Credit: Tim Herman/Intel Corporation)
• Rich Uhlig, managing director of Intel Labs, holds one of Intel's Nahuku boards, each of which contains eight to 32 Intel Loihi neuromorphic chips. Intel's latest neuromorphic system, Pohoiki Beach, is made up of multiple Nahuku boards and contains 64 Loihi chips. Pohoiki Beach was introduced in July 2019. (Credit: Tim Herman/Intel Corporation)

» Download all images (ZIP, 43 MB)

Our work with neuromorphic computing provides a good example. We set a goal to achieve several orders of magnitude improvement in energy efficiency (think greater than 1,000 times) for selected workloads like sparse coding, constraint satisfaction, path planning and adaptive control - tasks that we believe will be important as data increasingly must be processed at the edge in realtime and in energy-constrained environments.

We publicly kicked off our neuromorphic work in 2017 when we introduced Loihi, a brain-inspired research chip, to an ecosystem of research partners.

In March 2018, we established the Intel Neuromorphic Research Community (INRC) to further the development of neuromorphic algorithms, software and applications. We have granted INRC members access to accelerate their research through Loihi cloud systems and USB form factors. Their work has spurred new investigations into real-world applications for neuromorphic technologies. We are already seeing a series of publications coming from researchers in the INRC about creative new applications for neuromorphic computing and impressive performance and efficiency results.

Today, we announced Pohoiki Beach, which accelerates this effort by providing greater computational scale and capacity to Intel's research partners. Pohoiki Beach brings together 64 Loihi chips, each with 128 cores, enabling the efficient modeling of a system of about 8 million neurons. We hope that providing ever-increasing scale of neuromorphic computing platforms will enable the research community to tackle entirely new classes of problems.

This work is a natural extension of the technology innovations we have pioneered for decades. But we can't do it alone in our silo. To keep pressing forward the logical - and exciting - evolution of Moore's Law requires collaboration across the entire research ecosystem.

Rich Uhlig is an Intel senior fellow and the director of Intel Labs.