# IBM Shows Both Focus and Breadth at Think 2018

IBM Think in Las Vegas has subsumed all the company’s other conferences. Gone are Edge, Interconnect, World of Watson, Insight and Connect. One ring to rule them all.

This change in *modus operandi* has a side effect, however. The experience is like a 12-ring circus. You can’t be everywhere at once, and many great things are going on simultaneously. I stumbled almost by accident upon the tail end of Neil DeGrasse Tyson’s speech. There was a crowd filling every possible nook, craning to listen, and like a former Soviet citizen, I got in line before I knew what we were about, figuring it must be good.

And I picked up the thread just as he was hitting his stride in his “flat-earther” refutation. His simple proposition was that all solar eclipses feature earth shadows that look like round disks. There has never been one that looks like a pencil or even an oval. And he had a funny picture of the sun with a pencil-shaped shadow across its face. Debunking anti-science with observable evidence, in the best tradition of science, DeGrasse Tyson demonstrated one of IBM’s key messages at Think: science matters.

The floor acreage devoted to four technology nodes with repeated themes — Cloud & Data, Security & Resiliency, Business & AI, and Infrastructure — gave the ample facilities at the Mandalay Bay a feeling of the streets of Osaka, where intersections of tremendous avenues and elevateds stretch across the city, one after the other, in dizzying repetition. The scale was grand. CEO Ginni Rometty, comfortable in her role, having navigated IBM’s transition for nearly seven years now, addressed a crowd of more than 10,000 nearly filling Mandalay’s arena space. Total attendees clocked in at 30,000.

Certainly, there was lots of nuts-and-bolts product activity. Both customers and partners gave clear indications of the energy they are investing in current products and solutions. Feeding this interest was a stream of IBM commercial announcements, including advances in products incorporating blockchain, Watson, artificial intelligence, and leading-edge encryption techniques, technologies that only recently were still in the lab.

For customers, there’s something reassuring about knowing that their technology provider is engaging in research expected to bear fruit five years out and more that will be vital for their businesses’ survival. Even though these technologies won’t be on the market for a while yet, they indicate that IBM is not only taking care of the present, but has an eye out for what customers will face later on.

An example of this future-oriented viewpoint is the current R&D poster child at IBM, quantum computing. Research is running apace, and the company already has a working system, IBM Q, with a public Web interface. Grad students and tire kickers can try out a small quantum computer with a drag-and-drop visual interface that looks like a music staff. The five lines of the staff represent five quantum bits (or qubits). IBM’s research partners can already access a 20 qubit version, and from that system commercial quantum applications are just beginning to emerge.

I had a chance to sit down with Dr. Robert S. Sutor, vice president, IBM Q Strategy and Ecosystem. Sutor is a mathematician by training and thus able to explain some of the more obscure aspects of quantum computing. “The math is much easier than the physics,” he asserted. He noted that the 50 qubit system on display on the show floor was “80% of a real machine,” describing it as “beautiful” and comparing it to “a nice chandelier.” Among missing pieces were the vats of liquid nitrogen and helium needed to keep the qubits near absolute zero (15 millikelvin or -459.643° Fahrenheit). How near? Absolute zero is a short skip away at -459.67° Fahrenheit. Colder than even interstellar space, this is the temperature the qubits have to be to remain stable long enough to measure. Also missing: electronics that Sutor indicated would easily give away trade secrets to a knowledgeable viewer.

Sutor is working on figuring out what sort of problems are suited to quantum computing. He talked about its oddities: “superposition,” by which the qubits, in the most simplistic terms, are both zero and one at the same time, and “entanglement,” where the state of one qubit affects the state of another. While the physics of superposition may take a Ph.D. to understand, the math is more accessible. Let’s say a qubit is pointing at a location on a city grid three blocks north and seven blocks east. This location could be called both “north” and “east,” but it is more east. So, in the quantum world, it would be reckoned east at the measurement phase.

So, how might all these abstract concepts help with human commercial activity? Sutor is looking for what he calls “quantum speedups,” ways to recast problems so that they can be solved exponentially faster. He cited Shor’s algorithm, proposed by MIT professor Peter Shor, which uses a number of steps, including number theory, to transform a large-number-factoring problem into a different type of problem, which can be solved by a quantum computer to get an exponential speedup. It’s as if rather than having to solve exponentially harder problems with each additional bit, you could just count across the exponents instead. A ten-thousand-item list is a hundred times longer than a hundred-item list, but it’s only double if you solve at the exponential level between 102 and 104. The 2 and 4 become the operating elements.

Identifying the sorts of problems that might have this speedup is an important part of Sutor’s work. Promising domains include those seeing exponential growth, like IoT data. But early applications include multinode structures like chemicals, in which atoms interact in complex ways simultaneously.

If a quantum computer can represent such a complex state, including all simultaneous interactions, then it’s a much simpler problem than trying to compute all the possible combinations of states one after another, in the linear fashion of a classical Von Neumann computer (like the one you use every day), which faces exponentially larger numbers of possibilities with each additional node. A quantum computer just adds a new node to its bubbling pool.

Sutor gave as an example a caffeine molecule, which has only 95 electrons, and yet to describe the state of those electrons at one instant in time requires 1048 bits. The earth, he noted, has only 1049 atoms. A quantum computer with 160 qubits, however, could represent caffeine, he said, taking a problem from impossible to doable “in the next few years.”

It’s early days yet. Although the company was showing a model of a real 50-qubit machine on the floor, even larger versions are going to be needed to approach the modeling of interesting-sized molecules, perhaps with thousands of qubits. Other problems of this nature might be material design, drug discovery, and air traffic management. But rerouting the entire gird of airplanes while they’re in the air in real time will take many more qubits than are available today. So, we’re talking a ways off here.

In a future hybrid of classical and quantum computing, the two architectures will cooperate on artificial intelligence, where a quantum machine does generalized linear algebra and a classical machine crunches large matrices.

So, back to that earlier factoring problem. If quantum computing makes finding very large prime factors a snap, then encryption based on such factors will no longer be useful, since it can be broken quickly. But ever on the lookout for future potential, IBM is already working on ways to *defend* against quantum computers wielding hacking algorithms that don’t yet exist.

During Arvind Krishna’s “5 in 5” science slam (five people, each on stage for five minutes, present an advanced research idea that will affect society in five years), Cecilia Boschini, an IBM researcher based in Zürich, Switzerland, described lattice-based cryptography, a multidimensional approach that can protect against quantum attacks. Work on lattices, essentially stacks of matrices filled with rational numbers, has been going on since the late 1990s. One of the endearing characteristics of lattices is that they can be used as the basis for algorithms that are easy to construct but hard to crack. There is no efficient algorithm — classical or quantum — that can solve lattice-based problems in better-than-exponential time, thus making them robust against even quantum attacks.

IBM spends a lot on R&D (almost $6 billion or 7% of its 2017 revenues), including on advanced semiconductor process node technology, an area where the company no longer manufactures. IBM likes to work on hard problems where it has expertise that few others possess. These areas, blue sky initially, become market opportunities sometimes in as few as five years. From the company’s point of view, science is the beacon of the future. And customers for existing IBM products find this focus on the future reassuring.

At the show, IBM rolled out a new logo and catchphrase: “Let’s put smart to work.” The logo consists of five dashes arranged in an arc. The dashes will be familiar from earlier company graphics, but have been extracted and isolated to highlight how advanced science ends up in IBM technology that does real work for real customers.