The theory group at IBM served multiple purposes. For the company, having a top research group serves as a kind of status symbol. Moreover, it can also be a great resource internally. For example, sometimes the right answer from one of the research groups will save a company on the order of millions of dollars. Of course, there is always the hope that some of the researchers will become interested in practical problems relating to products. Recently, Halpern claims that this has actually become more of a requirement. IBM, for example, still claims to support basic research, but when hiring will also ask questions about how you will fit in and interact with the more practically-oriented groups, and what impact your research may have on products. This shift seems to be occurring in many companies, with Microsoft being a counter-example. (They seem currently to be very interested in having a solid theory group, and have also invested in other research areas as well.) In general, the last four or five years has seen a change in how companies such as IBM view research. They seem to be asking for tighter connections between research and development. At IBM, many researchers themselves will do a lot of coding. Halpern suggested that this was not a good use of researchers' time. Zippel noted that it's easy for researchers to lose track of the system, and that therefore requiring that they write some of the code is not necessarily a bad thing. In short, if you work in industry, you will have to write code.
There are obviously tradeoffs between work in industry and work in academia. At a university, a lot of time will be spent on teaching, committee work, writing grant proposals and dealing with undergrads. On the other hand, in industry the focus may be more product rather than research oriented and a fair amount of travel may be required. When asked about the job market, it was noted that research labs are hiring, but that they are more and more focused on having direct connections to the company's products. It's not necessarily true that the research coming out of such labs is "worse" than academic research. Cornell regularly looks at research labs to try to lure people to academia.
Zippel's perspective was that rather than view research in industry as a better or worse version of research in academia, that it should be seen as simply another possible career path. Working in industry arguably can give you a better chance to do work that will have a direct impact on society. In fact, it's possible to work for a company, not in a research lab, and still do research, albeit research that is not traditionally done in a university setting. Lotus Notes is an example of something produced from such an environment. The possible downside is that it is not easy to get back into academia from a position like that.
One distinction between academia and industry is in how funding is acquired. More specifically, the audience in academia is usually other academics. In industry, instead of convincing one's colleagues that an idea is worth pursuing, it is necessary to convince bankers, venture capitalists and consumers. This will of necessity require a different sort of argument than when writing a proposal to NSF, for example. Consequently, one will end up evaluating ideas differently, usually with an eye toward whether an idea will contribute to something good for society and/or increase the GNP. On the other hand, the NSF, for example, is becoming more and more concerned with "returns to society", both intellectual and financial, and thus this difference between academia, industrial research labs and advanced development groups is becoming blurred. Another distinction is the time-frame allotted, that is, in academia tangible returns/results are not expected immediately, but in some areas of industry the window may be as short as three to six months.
On another note, in industry it is often necessary to understand what's "good enough." That is, if some module in the system is just written horribly, it may be necessary to step back and say it's ok to let that go, that there are other more pressing things to spend time on. The tempo in industry is "absolutely breakneck." The goal being to get good ideas out onto the market first and best. It's essential when working in industry to always be aware of what the current state of the art is and what the hot problems are. In some cases, a hot problem today may not be so important in a couple of years, and deciding where and what to spend time on is crucial.