Some weeks back, I had a conversation with someone who was in charge of a relatively large lab with extensive robotics infrastructure and support personnel. He mentioned his vision of large robotics centers playing a role not unlike that of the physicists’ particle accelerator or space telescope. Recently, as I was making plans for my own grant proposals, I started thinking about this issue and I found myself slightly uneasy with this vision.
Some common arguments in favor of these large experimental facilities (or even large integrated projects) include the following:
- Ability to generate splashy big results that then lead to positive public opinion (within the extended scientific community and with the lay public, who indirectly control the purse)
- Allows for specialization of labor – there can be highly specialized theorists, “computational” folks and “experimental” folks.
- Makes it easier to convince funding agencies that every dollar/euro/yen will go a long way, and benefit a large community.
However, my unease comes from the following causes:
- Although the splashy new results will do great things for the field, and so help everyone involved in numerous ways, going down a path dictated by them will force several researchers to structure the research agenda according to the politics of the situation rather than in a truly independent way.
- The link between theory and experiment in CS is rather tenuous, so forcing it upon anyone will be a disaster. As it stands, the state of robotics and AI seems closer to the days when there were two separate but coexisting communities of natural philosophers – some doing thought experiments (and maybe some small experiments) and others doing very detailed and meticulous laboratory work. Frankly, I see too few people seriously exploring the foundational issues in robotics proper (as opposed to allied areas for which robotics is a good testbed).
- Having massive facilities may only make it harder for someone with offbeat questions to make their case.
In the end, I suppose the community will increasingly find itself consolidating in response to a variety of driving forces. However, I feel that it is very important for there to be avenues for people to do foundational work and small (and medium) scale experiments in a relatively blue sky fashion, without being forced to immediately justify everything in terms of how things will scale up to the grand unified experimental setup. This last point applies in a somewhat general sense… spherical cows, while not sufficient, do play a necessary role along the path towards any truly ambitious and speculative scientific question.