- IBM announced that its researchers have made an important breakthrough in brain-like ("neuromorphic") computing.
- The new IBM artificial neurons promise high-density, low-power supercomputing inspired by biological neurons in the brain.
- Possible applications include ultra-efficient Big Data processing for the IoT, financial analysis, and social media.
Technology giants including Alphabet Inc-A (NSDQ:GOOGL) and Facebook (NSDQ:FB) are betting on Artificial Intelligence (AI). Facebook, which is becoming one of the most advanced technology research centers in the world, is developing new applications of AI that, according to the company, will soon begin shaping the ways users interact with computers. Alphabet, after impressing the world with its spectacular AlphaGo demonstration, is beginning to apply AI technology to its commercial cloud computing service.
Typical applications of AI, also indicated as cognitive computing or machine learning, include automatic recognition and analysis of context, natural language, and media content such as text, voice, images, and video. These applications are expected to power both enterprise systems and social networks in the next decade. It's important to note that, besides software, AI computing needs powerful hardware, servers, and data centers. In fact, AI applications challenge even today's supercomputers and drive the demand for High Performance Computing (HPC). Since the sixties, hardware performance has followed an exponential curve with a doubling approximately every two years (a 1000-fold increase over ten years) - the so-called Moore's Law. But it appears that Moore's Law is slowing down, hence the need of new approaches to HPC hardware.
IBM (NYSE:IBM) is another company in the small group of technology giants that are betting on AI technology. The company is beginning to commercially exploit its own AI platform, Watson, for initial applications including health care. But, in parallel, IBM is taking the lead in the development of specialized AI hardware devices - "neuromorphic computers" - that work like the human brain, and expects neuromorphic computing to power next-generation AI.
It's important to realize that the challenging frontier applications of AI supercomputing mentioned above - automatic recognition and analysis of context, natural language, and media content such as text, voice, images, and video - are the very applications for which the neural network of biological neurons and synapses in the human brain is specialized. While face recognition remains a challenge for automation, babies can effortlessly recognize the face of their parents. Moreover, contrary to power-hungry supercomputing data centers, the human brain performs these impressive computational feats at very low power - three meals a day. This explains why neuromprphic computing is considered as an important goal of HPC and a potential breakthrough.
IBM announced that its researchers have made an important neuromorphic computing breakthrough, using special phase-change materials - such as germanium antimony telluride, used in e-writable Blu-ray discs - to store and process data in artificial neurons. The new phase-change artificial neurons are 90 nanometers in size, but the researchers said they have the potential to shrink the process to as small as 14nm in size, Computerworld reports. A nanometer is one billionth of a meter.
"We have been researching phase-change materials for memory applications for over a decade, and our progress in the past 24 months has been remarkable," said IBM Fellow Evangelos Eleftheriou. "In this period, we have discovered and published new memory techniques, including projected memory, stored 3 bits per cell in phase-change memory for the first time, and now are demonstrating the powerful capabilities of phase-change-based artificial neurons, which can perform various computational primitives such as data-correlation detection and unsupervised learning at high speeds using very little energy."
According to the company, this demonstration marks a significant step forward in the development of energy-efficient, ultra-dense integrated neuromorphic technologies for applications in cognitive computing. The research result, inspired by the way the biological brain functions, could open the door to imitating the versatile computational capabilities of large populations of neurons, at densities and power comparable to the brain.
The new artificial neurons can be used to detect patterns and discover correlations in real-time streams of event-based data, for example in Internet of Things (IoT), financial analysis, and social media applications. Large populations of these high-speed, low-energy nano-scale neurons could also be used in next-generation neuromorphic chips. “Populations of stochastic phase-change neurons, combined with other nanoscale computational elements such as artificial synapses, could be a key enabler for the creation of a new generation of extremely dense neuromorphic computing systems,” said IBM researcher Tomas Tuma.
IBM's "TrueNorth" chip, a neuromorphic chip with 1,000,000 neurons and 256 million programmable synapses unveiled in August 2014, developed by IBM with funding from the Defense Advanced Research Projects Agency (DARPA), is one of the first neuromorphic chips on the market. It seems plausible that the new artificial neurons could power future TrueNorth chips.
"The artificial neuron is built to mimic what a biological neuron does," said IBM researcher Manuel Le Gallo. "An artificial neuron won’t have the exact same functionality but is still close enough that you can achieve the computation performed by the brain using these neurons." The researcher added that neuromorphic computing is expected to be more efficient than traditional computing, especially for processing large amounts of data.
Though commercial exploitation of the new research results is likely to be several years into the future, there are immediate advantages for IBM: the company is in a good position to receive large government contracts for both civilian and military projects using neuromorphic computing. The company could also become the leading supplier of neuromorphic hardware, and sell even to its competitors in the overall AI sector. Investors should feel confident that IBM is leveraging its world-class research capabilities to be a top player in the advanced computing applications of the next decade.