IBM Research employs more than three thousand people worldwide. It is a corporate function and we have eight labs around the world. Four of them (Beijing, Tokyo, New Delhi and Haifa) specialize in software. In Austin we do research on processor technology and in Almaden, in Watson and Zurich we do research in basic science and technology, computer science and communications. The motto of IBM Research is to be vital for IBMs future success.
IBM has 320,000 employees and for a company like IBM to have a Research Division with more than three thousand researchers is quite a big number. It hits the bottom line and you better justify why you are doing it. Our reason for our existence, why we get our salary at the end of each month, is that we will have to show what we have done for the company recently or what are we promising to do for the company. There are nine areas that we focus on our research: (1) Communications; (2) what we call exploratory science (this is a catchall, which is good, because we put in quantum computing and all sorts of things that have right now no direct commercial viability); (3) personal systems (you might say that that cannot really be research, but of course it is in some sense, when you talk about usability factors and so on); (4) servers (they are our mainstay in terms of the hardware we are selling); (5) services (it is difficult to answer how you can you do research in services but I will address this later on); (6) software; (7) solutions (you can use research technologies to build solutions); (8) storage (another important area of IBMs business); (9) technology (we need it to build storage devices, processors, memories and communication links). Bioinformatics will increasingly become important. IBM Research has engaged in this field by creating the Computational Biology Center, and IBMs newly built Life Sciences business division is a considerable investment that has its roots in Research.
Many innovations that came out of IBM Research over the last decades impacted IBMs products but also the industry at large. Just a few examples: work on speech recognition which we did for many years and which now is getting to a point where it sort of works. But this is a more than thirty-year investment. More recently, we had breakthroughs in semiconductor technology like using copper instead of aluminum as the wiring. The Nobel Prizes that came about dont really fit into any of the nine categories of our research. I would claim they were not received because of management direction, they came about because we gave people enough freedom to run their own project and then it just happened. However, it turns out that the Nobel Prize for the scanning tunneling microscope has had significant ramifications for the continuing research we are doing now in our lab as well as for other areas in IBM and in the industry.
The disciplines in our research labs are computer science, electrical engineering, physics, mathematics, chemistry, and materials science. The values we create for IBM are technical leadership of the company, patents, luster, marketing support based on this luster, the vision we have for the company and directing the company plus the kind of people that we hire and train and who then sometimes move on into the business area of IBM.
How do we impact IBMs businesses? Essentially through all these various areas of technology that I mentioned and which address all parts of IBM. The challenge is that over the last couple of years services have grown such that in IBM global services there are roughly 150,000 technical employees. That is more people than are employed in the creation of software, servers and PCs, the traditional mainstay of IBM. Our research effort is not really directed to services. But I want to motivate research in services, because at the moment services is a people-intensive business. How do you grow the revenue in this field? You make services more efficient, get more out of the individual, deliver services electronically, develop better knowledge management, all sorts of things that we are only scratching the surface of. It is certainly an area where one can do research in.
The role of IBM research changed over the years. Research was growing because of economic pressures that were happening. If you have a company of say 200,000 people and you have only two hundred researchers you dont really care what they do. They dont cost much. If you have two thousand researchers it becomes rather important. That is why through the 1970s though IBM research was corporate funded, it had its own research agenda and occasionally it did some technology transfer. It was not done in a very coordinated manner because the money kept coming every year. In the 1970s IBM was the company in information technology. The profit margins were tremendous and so the IBM researchers could indulge in the kind of research and the kind of freedom they wanted. In the 1980s the company was already looking at how effective research was and tried to influence the direction in which the developments had to go. IBM started joint programs between research and the product divisions with a shared agenda that both parties, research and development, had to agree upon and IBM created collaborative teams to accelerate the transfer of research results. Clearly there was a focus on effectiveness.
In the 1990s the change truly came. The early nineties were really dire years for IBM, where a lot of my colleagues started asking: Are we going to survive as research or are we going to be blown apart and some of us be sent off to the product divisions and the other ones will be surplus?" In the Zurich lab, we were quite scared at that time because you could not send the best of us to some product lab which was maybe twenty to thirty miles away. To preserve research we tried to become more proactive here and started to work on customer problems. Part of the reason for that was because IBM in the pre-years running up to this catastrophic date had somewhat ignored customer requirements. The idea was to improve and really get out with the customer. It was unheard of that a researcher would go and directly engage with the customer. But the idea I call it research in the marketplace" worked.
In 2000 we see yet again a different role. IBM is no longer the dominant player. The company is still big but there are other strong companies. In information technology there have been a large number of start-ups and the competitive environment has changed a great deal. We still create a lot of technologies within Research , part of which we cannot transfer into the business. So now we are looking at how we can commercialize these technologies outside of the traditional framework. We have been doing this in an unorganized manner already in previous years but now we are looking at it in a more organized way.
As I already said, in the early nineties IBM was really on the verge of maybe no longer being around. What happened in research? In October 1992 we had the first all hands meeting worldwide in research. The idea was to define the strategic direction for research and to talk about IBM and the research financial challenges. In the 1980s in IBM it had always been the tradition to have two lab development sites to work on the same thing, just because one might fail and the other might succeed. In research we frequently had at least two projects on the same topic. But in the late 1980s and the 1990s there was a whole bunch of other companies out there that were already competing with us. There was no need any longer for internal competition. Therefore we had to change the way we did research and no longer compete with each other but complement each other. And we had to shift significant research resources to services, applications and solutions because that was deemed at that time to be the growth engine for IBM.
This happened then in the time period from 1992 to 1995. Our investments in services, applications and solutions had been five percent of our work and we increased that to twenty percent, largely at the expense of basic technology (minus twelve percent). This was difficult. A lot of people left and some retrained.
In 1996, we started something that we call industry solutions lab". That is a place where we can bring IBM customers, where we can show them interesting research prototypes and where they can interact with our researchers who typically give a talk about their topic of research. The customers talk about business problems that they may have. The idea is, first, to create a good impression with the customers and, second, to identify certain areas where we could collaborate and the customers could get some business benefits out of working with IBM research. Not for free necessarily, but paying. For example now we are doing work on high-performance computing with a large food company. That contact was established through a customer visit in the industry solutions lab.
Our view of technology life stages is such that we have two axes: the size of the investment that we do and the degree of consensus. For little investment in an interesting technology or science we will have low management control; empowered researchers define what they want to do. At this stage, it is not the management that decides that we shall work on this or on that. It is still the individual researcher who decides that this is an interesting topic where he/she wants to work on. It is not like in a product division where you tell people what they have to do. In the second stage, as the consensus and the size of investments grow, the senior management gets involved, essentially trying to move the technology further. If in the third stage the degree of consensus hits the senior executives, it means the level of investment goes up again. Then we try to move the technology to be really important for the business.
By doing a portfolio analysis of research projects, we observe from time to time that research projects are going to have an impact in the market. This means that a project produces something that can be transformed into a product and has some sizeable impact. In 1988 we were largely in a three to five year time frame. This has shrunk now in terms of the research time compression. In 1998 we were down to twelve to eighteen months. This shows that the speed of innovation, at least in information technology, has largely accelerated. People are not becoming any smarter than they were; it is because more people are working in the field. And there is more exchange of knowledge, that may also accelerate the process. In addition we see ourselves under pressure to work on shorter term projects.
We manage research in an intricate process that I will explain now: Essentially we have our ongoing research activities, the empowered researchers who define projects that they want to work on and the supporting business processes (the regular researcher is not really involved in this whole process). On an annual basis we do a global technology outlook where we try to predict where technology is going in the next ten years. Are there significant barriers that we will hit with a certain technology? The basic technology is on a geometric growth curve still (Moores law still holds to some extent) but we try to predict when it will stop. We know, for example, that in a few years we will get to the point where you cannot have the bits any smaller on a magnetic medium any more because otherwise they will lose their polarity and you will not be able to read it out. Then you know that you need something else in order to keep growing.
In the global technology outlook we explore the overall environment and create a vision. Subsequently we design a strategy to work towards that vision. The strategy determines what we want to do over the next year plus the years after that. It cannot change every year, otherwise it is just tactics. So we try and create themes that we will refine over time but that we more or less stick to for a couple of years. Out of this, in the course of the summer, projects get created or existing projects get morphed so that they fit the strategy. On that basis we generate a technical plan. At year end we look at the individual projects and set goals (at the beginning of the year these are the milestones) and check whether we have accomplished these goals. That is part of the measurement process. The measurements then directly influence the pay that an individual employee will receive because we have a variable pay bonus and an award system. It is a feedback loop in some sense. Then we go through this cycle again. The context in which we create this global technology outlook is by looking at global market trends, which is something that the corporation does and which is on a shorter time scale. But we also look on a longer time scale at how technology is evolving and what new opportunities this technological evolution creates. During that process, we have identified five main future trends.
The first trend: everything is getting faster and faster or cheaper and cheaper or smaller and smaller. Still for the next ten years we see no danger in terms of being on a growth curve. We know from past experience there was always some invention or some disruptive technology that changed things and allowed that curve to continue.
The second trend: everything is going online. Microprocessors are everywhere, everything is getting connected to the Internet. How does that affect things people are working on? What does that mean in terms of other research that then has to happen?
The third trend: the infrastructure to support everything being online and everything being connected has to become more intelligent. It has to have a higher bandwidth and it has to perform more functions than today. If everything gets connected there will be security and configuration implications. It also means that there is tremendous overhead in terms of human power that is required to keep this infrastructure running if we continue with the current course and speed. When you buy ten servers, you have to hire a person to maintain them. When you buy another ten, you have to hire another person. Growing the IT infrastructure means you will have to hire a lot more people than you may want to hire. That should not be the case. So what can we do in order to have the infrastructure run itself? This is a major effort which we call autonomic computing", with a strong involvement of IBMs research.
The fourth trend: the fact that everything gets faster and faster means that supercomputers become affordable, not only for the Department of Energy but also for normal" firms. Supercomputers are increasingly being used already in pharmaceutical companies, financial institutions, etc. Firms no longer use the computer just for pure infrastructure, squeezing out costs from the business, but in fact to grow their business by inventing new products, by running their business on the computer, by optimizing their supply chain, inventing new products, whatever. It requires entirely new algorithms, large integration of different pieces of software and additional research.
The fifth trend: now we have all these computers and put more and more things on them and more and more people are using computers. This means that the software also has to become more powerful. So software building blocks become the business. You no longer argue with somebody whether he/she uses Linux, Windows 2000 or another database, but rather you ask what the business level objects are that he/she uses and whether they are compatible with each other. That is essentially software building blocks at a higher level. Again, what kind of research do you need to do in order to get there?
For every research area we have so-called strategists who are our vice-presidents. They are all located in the United States. These strategists have to define certain sub-strategies for their strategy areas. The strategist I work with is responsible for services and software. Within this framework I as a sub-strategist am responsible for something which is called Internet Infrastructure, the intelligent infrastructure I mentioned before. Every spring I update the strategy documents and in the fall I write a plan about the kind of research we want to do, matching it with the strategy. There I look at the history, at the resources that go in and at the projects that contribute to this piece of sub-strategy.
First-line managers describe the individual projects that their group works on in a document. They say why they are doing this project, what it is about, what kind of milestones are included, what kind of background and resources they have and how they are linked with either the company or with other research labs or to the outside world. Do they have a chance of success with the work they are doing? It is always hard when you have a computer scientist saying he wants to work on a physics problem. That is not overly credible and so one would say that maybe it is not a good thing to work on. Or when you have a project that consists of two people and they try to accomplish something that would require ten collaborators. Then we give feedback and we try to morph the projects into something that have a little more chance of success.
Having said this, it sounds like we are planning research. We are doing this, to some extent. But at the same time, if you want Nobel Prizes you cannot plan for those. You have to leave enough freedom for people so that they can actually pursue areas that they are interested in and come up with some breakthrough discoveries. Why do we have this plan? We have it because the development organizations are all taxed to give money to research the corporate funding. Therefore they ask us what we are doing with this money and who is working for them. We never answer which persons are working for them, because that is then no longer research. Clearly we dont want to give them that much influence. But we pull out the plan and say that we have a certain number of people assigned to this strategic area of the company and a certain number to that area, and so forth. Then we can talk a bit more in detail as to why we have this done. So this plan is important in terms of documenting to the company how we are spending the money.
Our funding model implies that we have roughly sixty-five percent core funding. An additional twenty-five percent we match with joint programs funding that we get from the divisions. That means that the divisions get to influence our technical agenda (fifty percent of what we do). Then we have ten percent funding which comes from the outside. This could be government programs or commercialization of our technologies. We can keep the money that we make and invest it, which creates a bit of entrepreneurship. About one third of our core funding is devoted to so-called core matching joint programs, created in 1981. The idea here is that we have enough flexibility to work on the things that we want to work on as a research division and that we can leave people enough freedom to work on the things that they personally perceive as important. These programs are implemented now in virtually all IBM product, solutions and services organizations. They are managed by us or jointly (joint funding, shared agenda"). The target with these joint programs is to increase the speed to market. We do the complete job in the joint programs rather than simply handing it over. We used to do so, and then it often took another three years until something materialized. We manage this with so-called relationship managers. A relationship manager may be an individual or a team. He/she is responsible for the satisfaction of the research divisions customers or partners and provides high level access to the divisions resources. He/she develops a shared vision with the partner and is an active advocate for maximizing researchs value in the partners business.
Again, how can we document that we are doing good things for the company? We measure projects where we get external feedback from our partners or customers. They fill out an evaluation form. We get a partnership assessment from the people we work with, in particular if they give us money. Then we get a report card showing how much we influenced their strategy and business. The results for different business units are all quite good. But in some cases either research did not deliver or the business unit did not do well. Then you get lower grades. These assessments influence our variable pay, the bonus that we get at the end of the year.
These milestones, these measurement projects and these partnership assessments we call accomplishments". There we look at what kind of impact we had with individual projects, be it either in science and technology or be it impact on IBM, which means that it is a recognized impact, preferably in terms of dollars. We have categories of these, for example invited talks and papers at important conferences", publications in refereed journals", patent application rated file", and so forth. It is very important to document this and share this, both with the research population and the rest of the corporation.
What are the main forces of change in the twenty-first century? Communications are getting faster and more pervasive. There has been a wireless revolution and there is of course the Internet. That means that a lot of friction will get eliminated, channels will be dis-intermediated, people will be more directly connected. Computing becomes different. The paradigm changes. The Internet itself becomes the computer in a certain sense. We already have these grid computing initiatives. Deep computing, that is the IBM term for high-performance computing, becomes affordable. Computers become prevalent everywhere. This creates large amounts of data. To make use of this data in a business sense will require additional work. Customers move from task automation to using information technology for business advantages, to grow their revenue and consequently hopefully grow their profits. They use computing and communications to make money. This is why some people claim we see a bull stock market in the United States. Also the European recovery is largely attributed to the fact that finally information technology is bringing business benefits. Customers are moving this pervasively throughout their businesses. They put business practices, processes, law, what the consumer needs, into programs. Programs will in the future be used to trade service contracts and so forth. We see a closer integration of interdependent businesses. That is, companies become more virtual in the sense of that they will outsource a lot of things to partners electronically.
How will research evolve? As the computing will be less isolated, it is going to be integral to many components hidden from users. It is no longer viewed as computing". Teaming will become even more important. That means the pooling of deep expertise in various fields will be necessary to be able to do significant research. Teams will extend far beyond traditional computing and technology disciplines. Examples from projects of this kind we have been doing are weather forecasting, airline check-ins for customers or the large area of e-commerce. E-commerce touches not only upon all disciplines in computer science (security, networking, database, software engineering, cryptography, etc.) but also upon fields like economics, international relations, law, anthropology, sociology, etc. E-commerce is getting accepted as a discipline of computer science because it goes way beyond what we traditionally know.
Working jointly in the twenty-first century involves deep, multidisciplinary collaboration. One needs to leverage the disciplines and knowledge contained in many different institutions (hardware and software producers, specialists from industry, academia, think tanks). For example, in Zurich we are working with the Hochschule St. Gallen, which is known for its economics research, on joint projects. The purpose of this collaboration is to improve the way the world works, not simply to solve a narrow problem. It requires new structures for collaboration. IBM has founded the Institute for Advanced Commerce, which is intended as a catalyst to bring these things about. It is very much outward focused in terms of organizing conferences, publishing white papers, supporting theses in the field, carrying out selected projects with customers and creating a community of interest around this area. We are doing similar things with the Deep Computing Institute. There we have many areas, bioinformatics, decision support, simulation, visualization, topics that we are looking at and in some cases also working on.
With respect to teaming one of the things you need, even within the company, is to increasingly share knowledge between different groups to avoid overlap and become better. A large focus is on knowledge management. We are using the same tools that we sell to customers for exactly the same purposes internally and we have re-engineered many processes, in particular the patent process. There we have created specific awards for certain business areas to cause people to actually sit down and write down their good, patentable ideas. For the seventh consecutive year we have been leading in the number of U.S. patents issued to a single company.
Other changes are non-traditional paths to market. We leverage the worldwide labs and we have an organization by product target rather than discipline. However, we have countered this organizational form with professional interest communities that are organized by disciplines. The focus is on the values delivered to the corporation and this is tied directly to the bonus. For us with the corporation it is to discuss research in terms of the value it creates for the company, not the cost. IBM Research is part of the solution, not part of the problem. We have the strategic vision and we increasingly influence the way the company is going forward. We do integrate more than before with corporate-wide processes. While in the early years we were truly independent, we are now somewhat more integrated, but it is a fine line we have to walk to keep the amount of independence to truly allow independent research.