IBM Making Plans to Commercialize Its Brain-Inspired Chip

In August last year, IBM unveiled a chip designed to operate something like the neurons and synapses of the brain (see “IBM Chip Process Data Similar to the Way Your Brain Does”). Now the company has begun work on a next generation aimed at making mobile devices better at tasks that are easy for brains but tough for computers, such as speech recognition and interpreting images.

“We’re working on a next generation of the chip, but what’s most important now is commercial partners,” says John Kelly, a senior vice president at IBM who oversees IBM Research and several business units, including two dedicated to the company’s Watson suite of machine intelligence software. “Companies could incorporate this in all sorts of mobile devices, machinery, automotive, you name it.”

Adding brain-inspired chips to products such as phones could make them capable of recognizing anything their owners say and tracking what’s going on around them, says Kelly. The closest today’s devices come to that is listening out for certain keywords. Apple’s latest iPhone can be roused by saying “Hey Siri,” and some phones using Google’s software can be woken with the phrase “OK Google.”

IBM’s TrueNorth chip architecture, as it is called, was developed through a DARPA-funded program intended to make it possible for mobile computers to run advanced machine intelligence software such as image or speech recognition without having to tap into cloud computing infrastructure, and using very little power (see “Thinking In Silicon”).

Kelly says that IBM is in discussions with leading computer system manufacturers about how TrueNorth designs could help them, but declines to name any. “We’re talking with the who’s who in the mobile space and the IoT [Internet of things] space,” he says. A TrueNorth chip would be added to device designs as a “co-processor” that works alongside the conventional processor and never powers down, says Kelly.

The TrueNorth chip unveiled last August is roughly the size of a postage stamp and has one million silicon “neurons” with 256 million connections between them that are analogous to the synapses that link real neurons. The chip consumes over 1,000 times less power than a conventional processor of a similar size. IBM has demonstrated how its network of neurons can be programmed to perform tasks such as recognizing different vehicles in video footage in real time.

However, because the TrueNorth chip architecture is very different from those in existing computers it requires new approaches to writing software. And its fake neurons work differently than the software-based artificial neural networks that companies such as Google, Facebook, and Microsoft have recently used to make breakthroughs in speech and image processing using a method known as deep learning (see “10 Breakthrough Technologies 2013: Deep Learning”).

Neurons in IBM’s TrueNorth architecture encode data using electrical on-off “spikes,” attempting to mimic the spiking signals of biological neurons. The simulated neurons used in deep learning do not use spikes.

Artificial neural networks that use spiking neurons—IBM’s included—have not been shown to match the performance achieved using deep learning on tasks such as speech recognition or image processing. Yann LeCun, who leads Facebook’s AI research lab and helped pioneer deep learning, has expressed skepticism that it will be practical to do.

Dharmendra Modha, who leads development of IBM’s brain-inspired chips, counters that spiking is critical if neural networks are to be run in a chip with high power efficiency. His team has begun to create tools that will make it possible to transfer trained-up deep learning neural networks onto a TrueNorth chip, he says.

“This chip was envisioned as a substrate onto which a large variety of neural networks can be mapped for real-time, ultra-low energy, ultra-low volume applications,” he says.

Terrence Sejnowski, leader of the computational neurobiology lab at the Salk Institute for Biological Studies, agrees that spiking neurons are important if compact computers are to become capable of doing intelligent things without guzzling power or tapping the cloud. They appeared in nature for a reason, he says.

New research from another pioneer of deep learning, Yoshua Bengio of the University of Montreal, suggests that the technique’s accuracy could be easier to transfer to spiking hardware neurons than was previously thought, says Sejnowski. Bengio, who collaborates with IBM on language software, posted a preliminary paperonline last week showing that tweaking the simulated neurons used in deep learning in a way that makes them more like spiking neurons didn’t harm accuracy on image processing.

Even if IBM’s brain chip architecture is reconciled with the techniques of deep learning, it will have competition. Google is already working on ways to crunch down artificial neural networks to run on existing mobile devices (see “Google App Puts Neural Networks on Your Phone”). Several companies, including leading mobile processor designer Qualcomm, are working on chip designs that would run existing deep learning software on mobile computers such as phones or in cars (see “Silicon Chips That See Are Going to Make Your Smartphone Brilliant”).

David DiVincenzo on his Tenure at IBM and the Future of Quantum Computing

Photo: Alexander Hellemans

Theoretical physicist David DiVincenzo is widely viewed as one of the pioneers of quantum computing. He authored a 1996 paper (PDF) outlining five criteria he predicted would make quantum computing a reality; it has become a de facto roadmap for most of the research in quantum computing since then. In 1998, with Daniel Loss, he proposed using electron spins for storing data as qubits in quantum dots, which might prove to be the best choice for creating a working quantum computer.

In 2010, DiVincenzo was invited by Germany's Alexander von Humboldt Foundation to become Director of the Institute of Theoretical Nanoelectronics at the Peter Grünberg Institute in Jülich, and a professor at the Institute for Quantum Information of RWTH Aachen University. Previously he was a research director at the IBM T.J. Watson Research Center in Yorktown Heights, N.Y.

We met DiVincenzo in his Spartan office at the Physikzentrum of RWTH Aachen University, which is located “ten minutes by bicycle” from The Netherlands, where DiVincenzo has made his home.

IEEE Spectrum: You turned to investigating quantum computing while working as a theoretical physicist at IBM. What caught your interest?

DiVincenzo: I became interested in around 1993. It was not very much of a field at that time, but it was a field. There were two very eminent IBM scientists who were already involved for much longer: Rolf Landauer andCharles Bennett. Landauer is remembered for his contributions to thefundamental understanding of computing. Questions like what is the minimum amount of energy required to perform computational processes.

Landauer was quite important in the original discussions of quantum computing because he provided a skeptical point of view.  He thought that the sensitivity to imperfections in matter would be devastating for quantum computing. But he was interested in the concept of error correction that arose at that time and that could be applied to quantum computing.  And this really turned the story around. Bennett was famous for introducing the ideas of quantum physics into information science and cryptography. In 1993, he worked on what is now known as quantum teleportation.

I was fascinated by these developments, and at that time, IBM was flexible enough so that I could just jump in. I started contributing various ideas, and the following year, the Shor Factoring Algorithm was discovered. This made it clear that quantum computing could be done.

Spectrum: So, the research culture at IBM was definitely an important factor in your research career.

DiVincenzo: I would say that the research culture at IBM was always distinct. There was a whole evolution over the decades. For years it was thinking of itself in relation to Bell Labs: Are we as famous as Bell Labs?  That’s the history of the 1970s. I joined the lab in the 1980s; I had many friends that were there from the beginning, and I think I had a feeling for what the culture was like. IBM tried to really build up its research in the 1960s. In that period, they were definitely looking at themselves hoping to be another Bell Labs, which was in its heyday. By the 1980s, I felt that at that point they really did not have to worry whether they [were turning out] science of a comparable quality as Bell Labs. The cultures were similar; they would take rather young scientists and immediately give them all the resources of an institute, basically, without any of the responsibilities. This is a fantastic model which has proven to be not so sustainable—at least not in the corporate world. [And beginning in the early 1990s, it wasn’t really sustainable within IBM.]

Spectrum: How did this change affect your work at IBM?

DiVinzenzo: IBM had a heavy financial crisis in 1993, its most severe one.  It had a moment when it was really questioning its whole business model and whether it should be broken up into smaller companies. IBM undertook a whole sequence of different steps, such as getting out of personal computers,  and each one made it appear that physics had become less relevant to IBM.  The physics department got much smaller that year, and I remained in that smaller department. But we had a fantastic time after that in quantum computing.

Spectrum: So at least the research culture at IBM survived.

DiVincenzo: Here I would say something about the culture of IBM versus Bell. Bell evolved into a very competitive internal culture. People were really knocking against each other. Internal seminars were quite an ordeal because you were subjected to really heavy scrutiny. Internal dealings among scientists at IBM were much more congenial. 

The rest of physics at IBM was suffering. IBM still has a physics department, but at this point almost every physicist is somehow linked to a product plan or customer plan. At IBM, right from the beginning, there was always hope that these physicists who were dreaming up interesting things could actually contribute. Research became more directed as time went on.

Spectrum: It is now five years since you joined two German universities. How would you compare the life of a researcher in Germany—a country traditionally known for its emphasis on academic freedom ever since the 19th century? Do researchers in Germany have the same freedom that researchers had at IBM during the 1960s?

DiVincenzo: No. But I think they are freer and have more flexibility than what you find in the U.S. academic culture. Of course, in the U.S., there is heavy attention given to third-party funding. In Germany, this is not completely true. If you have a chair, you actually do have fixed resources that go with the chair, which is not the case in the U.S. But it is typically not enough to do any major project, and it shrinks over time, so you should connect yourself with some third-party funding. However, there is here a pretty strong long-term consensus that we don't tinker with science funding too much.

Spectrum:  Basically, future quantum computers might be based on qubits of two types: atoms or ions suspended by laser beams, and ions or electrons trapped in matter, such as in defects or in quantum dots. Which will prevail?

DiVincenzo: We're close enough to the quantum computer that we kind of can foresee its complexity in a classical sense—that is, how much instrumentation is required for this to work as a quantum computer.  I think that systems that involve lasers add a really big  jump in complexity because they would require a laser system for every qubit. Say we need a million qubits; we will have a system with a complexity well beyond anything that has ever been done in optical science. 

Now, for example, with quantum dots there are no lasers in sight.  Everything is done with electronics and at gigahertz frequencies.  You will need controller devices containing transistors that work at 4 Kelvin. That is an interesting challenge. It turns out that some conventional transistors, right out of the foundry, do work at 4 Kelvin. This becomes the beginning of some real scientific research: How do you make a transistor that really functions in a good way at 4 Kelvin without dissipating too much energy? Energy dissipated in the instrumentation close to the quantum system at that temperature could be one of a whole set of challenges that will not be solved easily. My own personal view is that we're a decade or so away from some really new machines that at least will partially fulfill what we've been thinking about since 1993.  

IBM vs. Intel in Supercomputer Bout

The U.S. wants to one-up China in supercomputers and it's looking for a few good semiconductor architectures to help out.

The fastest supercomputer in the world is currently the Chinese Tianhe-2 running at a peak of 55 petaflops on Intel Xeon and Xeon Phi processors. The Collaboration of Oak Ridge, Argonne and Lawrence Livermore (CORAL) project financed by the U.S. Department of Energy (DOE) aims to one-up the Chinese with up to 200 petaflops systems by 2018. The three systems, named Summit, Aurora and Sierra, respectively, have also pitted IBM/Nvidia and their graphics processing units (GPUs) against Intel/Cray's massively parallel x86 (Xeon Phi) architecture.

"Over 100 experts were involved in picking two different architectures as mandated by the CORAL request for proposals," the director of science for the National Center for Computational Sciences at Oak Ridge National Labs, Jack Wells, told EE Times. "Two locations — Oak Ridge and Livermore — were chosen to go with IBM, and Argonne was chosen to use Intel processors."

DOE's purpose if forcing the National Labs to choose two different architectures is mysterious, but was stated to be "in order to meet DOE's mission needs." Other than the general rule of not single-sourcing anything important, the purpose was perhaps just what they said — to meet the very different needs of the three labs — ranging from designing new semiconductor materials to simulating the explosive power of U.S. atomic bombs.

Oak Ridge National Laboratory (ORNL), for instance, today announced its intensions for Summit with 13 Center for Accelerated Application Readiness (CAAR) projects chosen to run on its current Titan AMD/Nvidia-based supercomputer as a warm up for Summit's IBM/Nvidia architecture, which will run 5-to10 times faster than Titan, about from 135-to-270 petaflops for Summit.

"We got 100s of proposals for how to make best use of the Summit's CPU/GPU architecture," Wells told us. "But we narrowed it down to 13 which we hope to have ready to run on Summit when its installation is complete in 2018."

Of the 13 projects, only one directly involves searching for new semiconductor materials, specifically superconductors, by Research Scientist Paul Kent at Oak Ridge National Laboratory. The others involve climate simulation by Research Scientist David Bader at Lawrence Livermore National Laboratory, relativistic chemistry by Professor Lucas Visscher at the Free University of Amsterdam, astrophysics by Research Scientist Bronson Messer at Oak Ridge National Laboratory, plasma physics by Professor Zhihong Lin at the University of California-Irvine, cosmology by Research Scientist Salman Habib at Argonne National Laboratory, electron-structure by Professor Poul Jørgenson at Aarhus University, biophysics by Professor Klaus Schulten at the University of Illinois at Urbana-Champaign, nuclear physics applications by Research Scientist Gaute Hagen at Oak Ridge National Laboratory, computational chemistry by Research Scientist Karol Kowalski at the Pacific Northwest National Laboratory, combustion engineering by Research Scientist Joseph Oefelein at Sandia National Laboratories, seismology by Professor Jeroen Tromp at Princeton University and plasma physics by Professor Choong-Seock Chang, at Princeton Plasma Physics Laboratory. All of which will be prepared on AMD/Nvidia's Titan for IBM/Nvidia's Summit, which will be based on the "data centric" principles of the OpenPOWER Foundation's approach to handling "big data."

"With the world generating more than 2.5 billion gigabytes of data every day, a holistic approach to performance, across things like memory, bandwidth, and data movement, is necessary to ensure today’s supercomputers are not missing vital pieces of information infiltrating through the big data pipeline. Through this holistic perspective, IBM with members of the OpenPOWER Foundation are creating data centric systems that will solve these data challenges and minimize data movement within the supercomputer to radically reduce data-movement induced latency," Herb Schultz, IBM manager for Technical Computing and Big Data & Analytics told EE Times.

In fact, IBM just finished installing perhaps the most beautiful supercomputer installation in the world to handle Spain's big data in a former Barcelona church (see photo). In contrast, Summit may not be as beautiful (see photo) but it will pack Multiple IBM POWER9 processors and multiple Nvidia Volta GPUs connected with high speed NVLink, a large coherent memory of more than 512 GBytes, an additional 800 GBytes of NVRAM — configurable as either a burst buffer or as extended memory. Dual-rail Mellanox optical interconnects will be configured as a full, non-blocking fat-tree with a file system transferring 1TByte per second to a 120 petaByte disk farm.

Due to security concerns about Livermore Labs stewardship of the U.S. nuclear arsenal, its a little more secretive about the exact configuration of Sierra, but it does admit that the IBM Power9 based supercomputer with Nvidia GPUs will top 200 petaflops. Sierra will also support the DoE's Advanced Scientific Computing Research (ASCR) program and will target what it calls "non-recurring engineering" (NRE) research and development (code reuse).

Intel on the other hand, is open and talking about its 50,000 node Xeon Phi based Aurora supercomputer for Argonne National Laboratory which will run at 180 petaflops. (The Aurora contract was actually awarded to Intel Federal LLC, a wholly-owned subsidiary of Intel Corp. and also involves a second smaller supercomputer called Theta, that will serve as an early production system for the Aurora by providing 8.5 petaflops and requiring only 1.7 megawatts of power.) Intel's Xeon Phi based Aurora will use Cray's next-generation Shasta chassis and will be used to design more powerful, efficient and durable batteries and solar panels, according to Intel, as well as improved biofuels, more effective disease control, improved transportation systems and more efficient and quieter engines and wind turbines.

Communications will be handled by Intel's Omni-Path Fabric optical interconnect technology, non-volatile memories and Intel's Lustre storage system and software. The Theta system, on the other hand will be based on Cray's XC supercomputer chassis, with a similar, scaled down, memory and interconnection fabric.

IBM hire advisers to deal with restless investors - sources

Some top shareholders of IBM, disappointed by 11 straight quarters of falling revenues, are seeking help from activist investors to shake up the company, but have been turned down by both Bill Ackman's Pershing Square and Jeffrey Ubben's ValueAct, according to people with knowledge of the matter.

International Business Machines Corp (IBM.N) is concerned about a possible attack by prominent activist hedge funds, and is working with two investment banks to formulate a defense plan, according to the people, who declined to be identified.

When asked for comment, IBM said: "IBM is continuing to execute on our strategy - making investments in growth areas such as analytics and cloud, reinventing our core franchises, and returning capital to shareholders. We are managing the company for the long term."

The storied American technology giant, worth $157 billion today, has struggled to transform itself from a low-margin hardware maker into a cloud-based software and services company.

When Virginia Rometty took over as chief executive at the start of 2012, Wall Street was hopeful that she would be able to kickstart growth. Analysts praised the former systems engineer for her strategic thinking in guiding IBM's acquisition of PricewaterhouseCoopers Coopers Consulting in 2002.

As revenues continued to decline year on year, however, some IBM investors began to lose confidence management, according to people familiar with the matter. Last year, IBM withdrew its long-term operating earnings target for 2015, and shares of Big Blue are now down about 25 percent from a March 2013 high.

Some IBM shareholders are trying to persuade prominent activists to build positions in the company and come up with ways to boost value, people familiar with the matter said.

Pershing Square and ValueAct Capital both looked at IBM in recent months, but passed on making a move, the people said. A spokesman for Pershing Square declined to comment. ValueAct did not immediately respond to a request for comment.

Part of the activist funds' concern was that IBM, whose stock is trading at around $159, is too expensive and the company's structural problems could not be fixed easily, according to several sources.

Another reason, the sources said, is that some investors feel Rometty is doing a good job coping with a tough situation, so she does not fit the role of an underperforming CEO that many activists look for when they make a move.

Once best-known for mainframe computers, IBM has been pivoting to security software and cloud services, but growth in those areas has not fully offset weakness elsewhere. Revenue in 2014 fell to $93 billion, from $107 billion in 2011.

The company, which earns more than two-thirds of its revenue outside of the United States, has been hit hard by the strong dollar and it divested some $7 billion last year. Adjusted for these two factors, revenues were down roughly 1 percent last year, the company noted. Also the company posted $21 billion in pre-tax profit from continuing operations in 2014.

IBM in said in February it is targeting $40 billion annual revenue from the cloud, big data, security and other areas by 2018. It has divested about $7 billion in commoditized IT assets, such as call centers and chip manufacturing, and announced multibillion dollar investments in cloud data centers and its Watson supercomputer system.

NO DISCUSSIONS AT PRESENT

It is not unusual for investors and consultants to try and shop companies to activists. At a time when stock picking is becoming tougher, pressuring companies to buy back shares, spin off units, or replace a CEO can be appealing.

There are currently no discussions between IBM and any activist investor, two of the sources said. Still, the company's advisers are educating IBM's board on how to handle an activist and to conduct a strategic alternative analysis, they said.

Warren Buffett's Berkshire Hathaway Inc (BRKa.N) is a top investor in IBM. With Buffett often siding with management, his presence may make it less appealing for activists to jump in.

"These days no company is safe from an activist looking at it," said Damien Park, who works with large corporations as head of consulting group Hedge Fund Solutions.

"Everyone is on the list and either you already have been approached by an activist or you will be approached by an activist."

The $20 billion Pershing Square helped breathe new life into Canada's No.2 railway, Canadian Pacific (CP.TO), and the $15 billion ValueAct helped engineer a leadership change at Microsoft Corp (MSFT.O).

ValueAct sits on the board of Microsoft, which competes with IBM in some markets. ValueAct may have faced a conflict of interest if it were to take action on IBM.

Other large technology companies that have come under attack from activists in recent years include Apple Inc (AAPL.O) and eBay Inc (EBAY.O), both targeted by Carl Icahn; Relational Investor LLC took aim at Hewlett-Packard Co (HPQ.N); and Elliott Management Corp is pressing EMC Corp (EMC.N) to spin off VMware Inc (VMW.N).

IBM starts testing AI software that mimics the human brain

We haven't talked about Numenta since an HP exec left to join the company in 2011, because, well, it's been keeping a pretty low-profile existence. Now, a big name tech corp is reigniting interest in the company and its artificial intelligence software. According to MIT's Technology Review, IBM has recently started testing Numenta's algorithms for practical tasks, such as analyzing satellite imagery of crops and spotting early signs of malfunctioning field machinery. Numenta's technology caught IBM's eye, because it works more similarly to the human brain than other AI software. The 100-person IBM team that's testing the algorithms is led by veteran researcher Winfried Wilcke, who had great things to say about the technology during a conference talk back in February.

Tech Review says he praised Numenta for "being closer to biological reality than other machine learning software" -- in other words, it's more brain-like compared to its rivals. For instance, it can make sense of data more quickly than competitors, which have to be fed tons of examples, before they can see patterns and handle their jobs. As such, Numenta's algorithms can potentially give rise to more intelligent software.

The company has its share of critics, however. Gary Marcus, a New York University psychology professor and a co-founder of another AI startup, told Tech Review that while Numenta's creation is pretty brain-like, it's oversimplified. So far, he's yet to see it "try to handle natural language understanding or even produce state-of-the-art results in image recognition." It would be interesting to see IBM use the technology to develop, for example, speech-to-text software head and shoulders above the rest or a voice assistant that can understand any accent, as part of its tests. At the moment, though, Numenta's employees are focusing on teaching the software to control physical equipment to be used in future robots.

IBM rumored to lay off 110000 employees

Last week, there is a rumor that IBM is going to lay off 110,000 employees, which is about 25% of their global 430,000 workforce. IBM confirmed that they do have plan for the layoff but it could be perhaps 10,000, not 110,000 that was spread in internet like wild fire.

I believe that rumors always started somewhere. Perhaps the CEO did present his 110,000 layoff plan, maybe exiting some unprofitable business, to board of directors for approval but got shoot down and only 11,000 was approved and that's could be where the rumor started.

They sold their laptop business to Lenova, and sold their server business to Lenova as well last year. They give cash to Global foundry to beg them to agree to take their foundry business away from them. So it won't surprise me if the CEO has presented a business spin off plan to board of director that will affect 110,000 employees since IBM has been busy selling off their lower profit margin business units.