Big Blue at 100: IBM Fellow Bernie Meyerson on Perseverance and Big Bets
There aren’t many technology companies around who can claim to be 100 years old. You’ll probably hear that a lot today as various media outlets report that today is the anniversary of the day in 1911 that three companies came together to form the Computing Tabulating and Recording Corporation in Endicott, New York. It retained that name until 1924, when it became International Business Machines, or IBM.
The sheer number of things we take for granted in modern life that hail in one way or another from IBM is astonishing. The credit card you probably used to buy coffee this morning bears a magnetic strip that was developed by IBM. So was the bar code that’s stamped on practically every tangible product you buy. The memory and hard drive in your computer evolved from technology developed at IBM, as did digital computing itself. For a quick rundown that will blow your mind, watch the video that Kara Swisher posted here in February.
Obviously, an anniversary like this presents an opportunity for many things. One is the basic PR impulse of a corporation like IBM to assert its importance and impact and get people talking about it. So be it. But it’s also a chance to reflect honestly on the impact that technology has played in pushing human society ahead, and to understand how, in an industry that’s traditionally thought of as being populated by young companies run by young people, IBM has so effectively stood the test of time.
I talked recently with Bernie Meyerson, IBM Fellow and vice president of innovation, whose research years ago into a material called Silicon Germanium has a lot to do with why you can find a Wi-Fi network nearly everywhere today. We talked at length about the trajectory that IBM’s research has followed and about the unique mix of research and business acumen that has helped make Big Blue what it is today.
AllThingsD: Bernie, there certainly aren’t many IT companies that can say they’re 100 years old. The company has had its hands in so many things that just permeate modern life. How do you explain it all?
Meyerson: Most people don’t even know that many of the things they take for granted evolved out of work that IBM had done. We’re notoriously bad at explaining the level of progress that has been achieved in human terms. But to understand the distance you’ve traveled is mind-numbing. Here’s a good example. If you’ve ever seen one of those big cruise ships, think about this for a minute. We invented the first disk drives around 1955 or 1956. And the disk drive basically at that time was this monstrous thing using paint from a shipyard to make the memory part of the disk drive. It’s kind of a strange story, but it worked. In any case, if all we did was invent something and then walked away from it, then the amount of storage you have in a laptop today would weigh 250,000 tons. What’s typical data storage today would be the equivalent of two of those big cruise ships. When you put it in terms of gigabytes it loses all meaning. When you think of cruise ships it makes sense.
Well, thanks for not walking away from that one. Storage is one measure of technical progress but so is processing, which is essentially our ability to get things done and how fast. Do you have an equally colorful metaphor to illustrate progress there?
We moved to digital computing in the mid 1950s. It’s hard to compare what we were shipping then to the best machine we’re shipping today, which does about a petaflop, about 10 to the fifteenth power floating point operations. I’m about six feet tall. How tall would I be if I simply grew in height at the same rate that computers have grown in capability? If I had, I would today be looking down and seeing the moon whack me in the ankles, roughly. If you put numbers on it, it gets lost. It’s very easy to forget how far we’ve come, and very dangerous, too.
If you forget how far you’ve come, you don’t realize the importance of the tremendous investments required to go further. To continue the metaphor, if I’ve already grown so tall that the moon is whacking my ankles, don’t underestimate how hard it’s going to be to grow even further because I’ve already grown an unimaginable amount. The trouble is there are a lot of other folks at a lot of other companies who walk away. They don’t comprehend somehow the investment required to stay at it, or worse, they don’t value it. And it’s disconcerting to me as a scientist and technologist when there isn’t a focus on the drive necessary to keep things going. To some extent there’s a part of our culture at IBM that says you just keep driving, you keep innovating. And furthermore, you understand intuitively that the further you go, the harder it gets. The first step? Piece of cake. Further is not trivial. By the 47th step you’re walking straight up a cliff and it gets very difficult. If you don’t keep in mind the lessons of the past you’re doomed to repeat the same errors.
So how does IBM deal with that? Other companies have shorter horizons in their research. IBM does a lot more basic research that may or may not lead down a productive path. How does IBM manage that consciousness of what has come before versus the need for a shorter-term payoff?
It’s an excellent question and it basically comes down to culture. You don’t just create an organization that comprehends this overnight. This is basically the result of 67 years of learning. So when you look at IBM and particularly our research division, which does astounding things, we have a culture of making what we call big bets. These are things that are incredibly aggressive in terms of the targeted outcome. But they’re bets. Key word: Bet. You win or you lose. The bet aspect pertains precisely to your question, which is how do you focus on getting it done, and getting to closure, even if it’s a longer-term effort. The danger tends to be that if you do things that are really disconnected, and which don’t have a logical outcome, they can linger for a decade or two without any real progress. If you take a long view — a view that is nonetheless grounded in reality — the question comes down to how do we do it? It comes from this long, long-standing culture. Some things come to fruition quickly, others take forever. The question is having the right mix. That’s the element we’ve mastered. You have to have a diversity of projects under way. You can’t make all your bets on something five or 10 years out, but similarly you can’t place them all on something this quarter. Over the years, we’ve basically developed structures that ensure we have that diversity, and it has served us very well.
What’s a recent big bet that’s paid off?
Data and analytics. We felt that the use of IT to deal with big data — vast data sets that no human has a chance of dealing with — that sort of needle-in-a-haystack item is really something you can address with IT because of the computing horsepower that’s now available. In 10 years we’ve invested $60 billion in R&D and $14 billion acquiring 25 companies tied to analytics. And we’ve done some astonishing things in terms of accessing big data. One good example was Watson. That capability of dealing with big data has had some profound impacts near-term and it will have tremendous societal impacts over the long term.
We hear a lot about big data and analytics these days and I think it’s one of those abstract buzzwords that makes people roll their eyes. Can you give me a concrete example of analytics in action?
We worked with a hospital in Toronto treating premature infants. They have underdeveloped immune systems so when they become ill, they don’t really respond. They don’t have all the usual warning signs the way healthy infants do, so it’s hard to tell. By the time you know they’re sick, they’re so infected they usually die. The problem is this: How does a nurse working in the intensive care unit spot these infections early enough to treat the child so they don’t die? We asked: What if you took all the data coming off a child in the ICU — their heart rate, their blood gas measurements, brain activity, breathing activity and so on. This data flows out at a rate of tens of thousands of readings a second. If you compound all these into a database and then watch that child, every once in a while one of these children will get one of these infections and die. What you do then is you work back through all the cases like that to see if there is an early warning sign that you’re missing. All the other parameters for that child are normal right up until the moment they’re terribly sick. It turned out there was a warning sign, and it turned out to be that very early, a certain normal fluctuation in the heartbeat went away. Their heartbeats were more stable, but even so it was within the sweet spot so it wasn’t generating any alarms. That turned out to be one of the warning signs. Now they’re able to see 24 hours out or more when one of these children is getting ill before the best trained ICU nurse could ever spot it. It makes the difference between these children living and dying.
That’s a winning bet. What about the losing ones? What happens when the big bet turns out to be wrong? How do you manage that?
One of the things we do at IBM is that when we make a big bet, we practice the art of parallelism. You have two or three things that work in parallel and you measure against an absolute standard. The reason you do this is that you have to at some point choose a frontup and a backup plan. You never get it right every time so you have to have a Plan B. Which we’ve done with a lot of success. We had a case where we were developing a new material to use as an insulator on chips. The preliminary data showed that one we were using, lets call it Material A, was promising. But we also continued to work on another one we’ll call Material B. After nine months, A just wasn’t cutting it and B turned out to be the winner. And of course we had lost some of the investment, and we punted on A. This friend of mine with one of the trade journals came in to see me. He’s a good guy, but he came in and wanted to have a little fun at my expense about this. He asked how we were going to explain that we worked on A for nine months and it didn’t work. I thought about it for a moment and I looked at him and said, “Show me someone who has never failed and I will show you someone who will never lead.” You have to be willing to hang by your fingertips with the understanding that every once in a while you’re going to lose your grip. But you similarly have to make sure that somewhere you have a net so that the fall doesn’t wipe you out. It’s a very tough balance. It’s not that we avoid failure. Quite the opposite. We do hard stuff. I could list many many programs we’ve had to shut down because they weren’t cutting it. Most companies that have cataclysmic failures simply don’t know when to punt.