Latombe: 'Innovative universities must attract top researchers' 

Jean-Claude Latombe is a professor at Stanford University's Artificial Intelligence Laboratory. He was also a member of group of external experts which evaluated research at  INRIA , the French national institute for research in computer science and control.

To read a shortened version of this interview, please click here.

There are several models of organising research infrastructure, including independent, competing private universities and a model dominated by state-run institutions. What do you see as the major advantages of each? 

First, I think that multiple models should co-exist. I am generally more in favour of private universities (after all, Stanford is one), but they can work well only in certain conditions. First, no single university can do everything well, like simultaneously doing cutting-edge innovative research and providing practical training to a huge number of students. Hence, private universities must choose the areas in which they expect to excel. Second, a philanthropic culture must be established to support private universities. 

Advantages of private universities usually include smaller bureaucracy and the fact that decisions are made locally and more quickly. Life at Stanford is mostly shaped by its faculty, while my experience in France has been that life at state-run universities is mostly shaped by politicians that often have little knowledge or understanding of the academic system or what research is. 

Does this mean private universities are more adaptable than state universities? 

Private universities tend to be more innovative and flexible. They are certainly a good model for research-oriented universities, though there are some notable counter-examples in the US, like UC Berkeley and UT Austin. 

Top research-oriented private universities like MIT and Stanford attract good students from various backgrounds. These universities provide much financial support in the form of scholarships (for undergrads) and teaching or research assistantships (for graduate students). 

Although I am less familiar with other private universities, I suspect that they seek mostly affluent students. So, state-run universities are still needed. 

It is also clear that private universities must actually receive much state support in order to fund research programmes. Stanford or MIT would not survive well without the grants we get from NIH, NSF, DoE, Darpa and others. So, private universities are still quite dependent on the state. 

Does the funding model make a difference to how successful a university is? 

Overall, I don't think that whether a university is private or state-run is the key issue. What really matters for a research university is to recruit the best professors and admit the best students, to encourage diversity, not only of people, but also of ideas, and to create a vibrant atmosphere where excellence, impact, and creativity are valued by all for the benefit of all. UC Berkeley shows that a state-run university can achieve this. 

The report by the visiting committee evaluating French institute INRIA highlights the importance of increasing pay scales. In general, how important do you think salaries are in attracting people into research? 

Salaries are only one aspect of the full picture. At INRIA and in research universities, quality of colleagues and of students is even more critical to attract new excellent people. In research universities, teaching load is also important. It's quasi-impossible to both do cutting-edge research and be an excellent teacher when the teaching load of a professor is close to 200 hours per year. The quality of the infrastructure is also very important. 

In the end, however, salaries still play a big role. They reflect the value that is given to a professor's work. Over time, low salaries are demoralising, because they send a signal to the professors (even if they love their work) that they are not highly valued. Higher salaries encourage people to challenge themselves, precisely to justify their salaries. 

Are European salaries a lot less than US salaries? 

Yes, salaries at top research universities in the US are much higher. 

Interdisciplinary research is also highlighted. Why is this so important, and what can be done to incentivise collaboration? 

Research in core domains is certainly important as well, but most new contributions are made at the intersection of established disciplines. Computer science itself was born at the intersection of Applied Math and Electrical Engineering. 

Doing interdisciplinary research is fun and may not need major incentives. But two issues must be addressed carefully: how to fund interdisciplinary research (while there exist no experts to assess the quality of such new projects) and how to evaluate and promote interdisciplinary researchers. 

There is no simple solution to these questions, but the solution that consists of doing nothing is worth than taking risks. 

Do you think it is most useful for countries/regions to focus investment on a small number of research areas? Or is there value in being moderately good at a wide variety of disciplines? 

I think it is somewhat counter-productive to focus investment too narrowly within disciplines. For example, in computer science and information technology, progress in one area (say, reasoning about uncertainty in robotics) may benefit a very different area (dealing with imperfect data in unstructured databases, like the web). 

On the other hand, a region can focus on broad disciplines like aerospace, or biotechnology, or nanotechnology. In my opinion, computer science and IT are likely to remain the angular stone for most disciplines; no one can overlook them. 

Are you an advocate of 'competitive clusters' (where research institutes and industry are closely linked, often in the same geographical area)? 

I am for what works. 'Competitive clusters' may be good in some cases, but I suspect that they often increase bureaucracy, time wasted in coordination, travel, and overhead tasks. I am not sure that academic research and industry should be closely linked. They do not work on the same time scale and with the same constraints. 

I am an advocate of having many links between professors and industry – such as professors consulting for companies, professors creating startups, companies providing advice on curricula and research projects – but I don't think that most research projects in university can be directly linked with industry. 

It is often said the more should be done to promote maths at school, but do you have any suggestion as to how this can be done? 

There is no easy solution. I believe that academic researchers should more often go and make exciting presentations in high-schools, but I am not doing it much myself. Also, if a country deeply thinks that it needs more scientists and engineers, then it should find a way to pay them better. Soon high-school students would get the signal. 

The report on INRIA suggests closer links between tertiary and secondary education. In practice, can you give an example of what this would involve? 

In addition to making presentations at schools, organising summer camps for high-school students would also help. 

Do you believe industry should have input into developing university curricula? 

To some extent. But I think it is more critical for professors to have an 'industrial culture', by doing consulting work or creating startups, so that they can better understand and anticipate the impact of current research on industry. Most of the time, a curriculum is delivered to a student that will be on the job market several years later. Few companies know their needs several years in advance. Researchers with industrial culture are in a better position to anticipate. 

Describe your current research? 

I currently have a small group of PhD students, since I plan to retire in a couple of years. My research is in two areas: robotics and computational biology. In robotics, I have been working for many years on 'motion planning', i.e. on developing techniques to allow robots to automatically choose their motions to achieve task goals. Currently, I have a student developing an autonomous rock-climbing robot. 

In computational biology, I work on modeling low-frequency motions of protein in their folded states. These motions (which occur at a time scale between 1 microsecond and 1 second) are critical to the functions of the protein (e.g., its ability to bind against other molecules). 

These two domains use similar mathematical and algorithmic tools, despite the fact that the domains look very different. 

What could the robots you work on be used for? 

The climbing robot is not targeted toward any application. However, the results can be used to build robots to explore steep terrain on the moon and on Mars. It can also be used to build robots to run on very rough and steep terrain (e.g. patrolling borders). 

In the recent past, I have contributed to creating a brain surgery robot (the Cyberknife, now developed by Accuray), which is now used in many hospitals around the world, not only to treat brain tumors, but also many other cancers (lung, pancreas, liver, spine, prostate). I believe that close to 100,000 patients have been treated with this system. 

My work in computational biology may help in the future to develop new drugs or new treatments (e.g. based on designing new proteins). But I don't work on these applications specifically. 

What at the major sources of funding for this kind of research? 

NSF, Darpa and other military agencies, NIH, industry and private donors. 

Since the economic crisis, are researchers in the US concerned about funding? 

Not really (at least less than I would have thought). One reason is that the impact is not immediate. Obama's stimulus package has also increased NSF and NIH research budgets. There is more concern about the impact of the loss in our university endowment. We have to significantly reduce the costs of our operations supported by the university. It will soon lead to making difficult cuts (e.g. less administrative support and fewer teaching assistants in our classes).