IBM's R&D Projects Include RFID, New Search TechniquesIBM's R&D Projects Include RFID, New Search Techniques

IBM has always invested heavily in research and development, and that's not changing any time soon. To hear more about some of the projects and strategies being pursued by IBM's R&D labs, information associate editor Laurie Sullivan recently sat down with Paul Horn, senior VP of research at IBM at his Yorktown Heights, N.Y., office. The following are excerpts from the interview:

information: IBM sold many of its radio-frequency identification device patents to Intermec in 1997. IBM has admitted in the past that it missed some big opportunities, notably network routers. IBM manufactured the first commercial router before Cisco Systems ever existed, but execs couldn't convince higher-ups that IBM ought to care about routers. Will IBM look back at RFID devices and say the same thing?

Horn: We'll see. We made a decision that the value from RFID is in the application and not in the device. The real value is in solutions. So that's where we put our focus and sold the intellectual property on how to build unique devices.

Maybe we were wrong. We'll see over time. I'm comfortable with the cards. I wouldn't want to be in the RFID tag-manufacturing business now. There's more value in solutions, consulting services, and middleware applications as RFID gets ubiquitously deployed.

information: Does IBM still have RFID intellectual property?

Horn: At the solutions level.

information: Are there any stipulations in the Intermec contract that would prevent IBM from developing and selling device-level RFID?

Horn: I don't remember the specific contract, but I would be surprised [if there were], because we never sign anything that inhibits our future research. We sell or license our intellectual property and it's almost always nonexclusive and almost always has no restrictions on future development.

information: Has IBM ever gone back to rethink a strategy and start again?

Horn: All the time. A classic example is we built the world's best first relational database on a mainframe. For whatever reason, we decided not to take the technology and research commercial [and develop it for] the Unix space. After some time, we went back and realized we let Oracle get huge and maybe it's about time we took this technology and did something with it.

Same thing is true with RISC architecture. We invented RISC. Won the national medal of technology for it, and gave Sun a huge head start. Now we're ahead. But it took us too long. Hindsight is 20/20. We're better than most companies in finding disruptive technologies and embracing them, but there are lots of examples where we didn't.

Sometimes you may be too late. There are some markets that it doesn't pay to get into unless you can be a big player. Networking hardware is an example. That's one of the nice things about having a strong technical and research team.

If there was a huge high-margin semiconductor business in RFID technology, we still have a bunch of experts here.

information: OK, so IBM has no plans to get back into RFID devices or the chip business, but is it fair to say the company won't close the door?

Horn: That's correct. That's fair. Right now we have no plans to go into RFID device or chip manufacturing. IBM is focused on middleware, solutions, and applications.

information: What are some of the other innovative projects IBM's research and development division is working on today?

Horn: We have a project where the computer can crawl the Web similar to Google, but instead of [relying on] a keyword search, it reads the text, understands the text, and allows you to mine the context, not just keywords. The researchers bought back this idea because they looked at all the unstructured information on the Internet and tried to figure out how to get value from it.

information: How could a company use that technology?

Horn: You could use it to manage the brand. We did this for a record company and discovered the buzz in chat rooms was a terrific indicator of future record sales. The buzz in the chat room started and within two weeks the record sales [in a specific local region] increased. The software, Web Fountain, is a prototype piece of research software that requires many servers to scan the Web. With the software, companies can manage their brand, predict future sales in a specific geography, or determine what attributes will push sales higher or lower. But this is a combined research and consulting service, because once companies gain access to this information they [will want to] reengineer their business processes to have the correct products in regions at the right time.

information: Could this type of research turn into a business unit or division within IBM Global Services?

Horn: It could. We have 13 of these micro-practices that are embryonic, incubating service applications. The Web Fountain research application is part of Text Analytics, which is one of the 13. If any are successful, we can make them part of the consulting services business.

information: Is there a similar application for the retail industry?

Horn: We have an Emerging Business Opportunity [unit] in the company for retail, and research is working closely with them. This is focused on the in-store experience and they look at how handheld devices can change the [retail] experience and [whether it makes sense to establish] network platforms within the store to closely collaborate with supply chains and minimize inventory.

Another example is a screen attached to a shopping cart that you plug in your choices. The screen would tell you where to go [to find items in the store] and the specials available in the store. A big piece of what we do is in self checkout and standard checkout, and then connecting that information back to the inventory systems, so the store knows there were three bags of Fritos sold at [a specific] register and how many Fritos bags are left on the shelf. IBM provides the hardware for the checkout and the middleware within the store.

We've also done prototypes with companies interested in shopping patterns. They had video cameras, and we provided software [on the back-end] that captures a person [on the video camera] and track the person's walking pattern through the store. The software picks several people at once and follows them from camera to camera throughout the store. The application works on image recognition. This [application] allowed the store to optimize the floor plan and determine where they should put the merchandise. We've also been talking with security companies about the same software for airports.