Is it a good idea for an academic researcher to work with industry R&D projects? Yes, in small doses:

We find that university-industry relations exercise a positive effect on university scientific productivity only when (…) these activities do not exceed 15% of the researcher’s total budget. (Source)

Thus, spending no more than a few days a month working with partners in industry is good for you. My own policy is that I will serve on advisory boards, but I will not lead a development team.

Some specific types of collaborations I have done as an academic researcher:

• I worked on an ISO committee. I could not stand it. Trying to parse hundreds of pages of technical documentation is boring. The noise-to-signal ratio is very high.
• With Sean McGrath and NRC, we licensed inDiscover.net to Bell Canada. This was a good experience because Sean is an excellent engineer. Research-wise it was helpful to build a collaborative filtering algorithm for actual users.
• I have served on several advisory boards. In a sense, it is ideal because it uses very little time and everyone tends to win.
• I got a few contracts as a consultant. This made no sense to me and so I stopped doing it last winter. My time is extremely valuable, so selling much of it is not interesting. Yet, clients need someone who will commit large chunks of time eventually. Moreover, clients need someone who will pick up the phone and fix problems when they need it, not when you are available.

Many of my Ph.D. students have admitted to being motivated by financial gain. Stanford is famous for their graduate-students-turned-entrepreneurs. Sergey Brin and Larry Page of Google fame come to mind. But I cannot yet point to a comparable success story. Still, I have seen several variations over the years:

• Some students believe that having Ph.D. on their business card will earn them higher salaries.
• Others believe that a Ph.D. is an ideal setting to come up with commercially-viable technology.

These approaches appear to violate Dijkstra’s second rule:

We all like our work to be socially relevant and scientifically sound. If we can find a topic satisfying both desires, we are lucky; if the two targets are in conflict with each other, let the requirement of scientific soundness prevail.

I have my own related experience. Shortly after getting my Ph.D., in the middle of a post-doctoral fellowship, I decided to run off and start a company with friends. My business plan was simple: I setup a web site where I said that I would do research for money. No ad, no networking, no product, no gimmick. Incredibly, people would call me with job offers. I did well in the sense that I never worried about money. (See this TV interview about some of my crazy industry work, in French.) I also learned much. My experience has helped me become a better researcher. Indeed, I have become very critical of published research. But I was never able to reconcile the pursuit of knowledge with commercial interests. I did a lot of crazy mathematics, programming, but I never had enough time to think deeply about any one issue.  And, as it turns out, I don’t love money nearly as much as I love knowledge.

Thus, I tell my students that they should be in it for the pursuit of knowledge. Some get upset. Others appear to ignore me. A few leave. I am still waiting for one such student to earn his Ph.D.

To sum up my opinion: if your goal is to become a highly-paid consultant or an engineer, the Ph.D. is an overkill.

What do you think?

Computer Science can be taken a natural science: the study of how the universe processes information. If it is a natural science, then does it build on Physics? Or does Physics build on Computer Science?

The answer is obvious (to me): Without algorithms there would be no Physics! Physics is built on the fundamental assumption that we can model the world using algorithms. Computer Science is the most fundamental natural science.

Does Computer Science make fundamental (and possibly falsifiable) predictions about the universe? Of course, it does! Take the strong Church-Turing thesis: the assumption that the universe is a Turing machine. It has deep implications:

1. There is no problem solvable by a human brain that cannot be solved by a machine. In particular, creativity and intuition are computable. Philosophically, we have no soul (not anymore than a PC).
2. We all live within a discrete computer simulation. Physics is digital. Continuous functions (such as f(x)=sin(x)) are approximations to the discrete functions governing nature, and not the reverse. We all live in the Matrix.

The predictions are falsifiable:

1. Come up with a task, such as writing poetry, that a brain can do, but not a digital computer.
2. Show that we cannot reproduce a physical process using a digital computer.

Yet, these predictions remain unfalsified to this day.

To put it bluntly: The most fundamental of all natural laws is that we live in a Turing machine.

Further reading: The World is Digital (blog post by Dick Lipton)

Consider the problem of hashing an infinite number of keys—such as the set of all strings of any length—to the set of numbers in {1,2,…,b}. Random hashing proceeds by first picking at random a hash function h in a family of H functions.

I will show that there is always an infinite number of keys that are certain to collide over all H functions, that is, they always have the same hash value for all hash functions.

Pick any hash function h: there is at least one hash value y such that the set A={s:h(s)=y} is infinite, that is, you have infinitely many strings colliding. Pick any other hash function h‘ and hash the keys in the set A. There is at least one hash value y‘ such that the set A‘={s is in A: h‘(s)=y‘} is infinite, that is, there are infinitely many strings colliding on two hash functions. You can repeat this argument any number of times. Repeat it H times. Then you have an infinite set of strings where all strings collide over all of your H hash functions. CQFD.

Further reading: Do hash tables work in constant time? (blog post) and Recursive n-gram hashing is pairwise independent, at best (research paper).

Bill Gasarch is proposing a manifesto on Open Scholarship. What a great idea! Imagine thousands of researchers openly agreeing on practices making research more effective! We could change the culture of scientific research without having to convince publishers or funding agencies.

To keep the ball rolling, here is my version:

1. Whenever you publish a paper in a conference or journal, post it on your website or on some appropriate archive (such as arXiv). In particular, as soon as you submit the final version to a conference it should go on online.
2. Post improvements and revisions to your work. Should you spot a mistake in one of your older research paper, revise it and post the result online!
3. If you give a talk, then post the slides online.
4. Make it easy for other researchers to get automatic updates when you post new content. (If you use arXiv, it comes for free if you claim an arXiv user ID.)

Further reading: Berlin Declaration on Open Access to Knowledge in the Sciences and Humanities and Toward author-centric science