Earlier this month, Michael Mitzenmacher told us about the record number of students attending his Harvard class online-only. Yesterday, Dick Lipton predicted that online learning will replace campus learning : “I see no reason that On [Online Universities] could not do as good a job as Un [Campus Universities] with this basic goal [Educate Students].” In the comments, Lipton questions the importance of credentials and whether social interactions really need the campus.

I have already written much on the topic but let me reiterate my message:

• In this new online world, professors are not content providers. They provide structure and motivation. They are role models. And most importantly, by their reputation, professors can provide certification. If someone gets a reference letter from Michael Mitzenmacher or Dick Lipton, I trust they know something about Computer Science, because I trust Michael Mitzenmacher and Dick Lipton. I suspect it is not easy to get these fellows to write fake reference letters because they have a high degree of independence (job security, good money, and so on) and their greatest asset is their reputation.
• Students are trained to expect classrooms. Many students need structure and constant attention. That is not a good thing! We are effectively training students to be good employees working in large organizations with much structure. Yet, this world made of large and stable organizations has already fallen apart. We urgently need to teach students to learn on their own, using the Web.
• Yes, there will always be campus classes, the same way there will always be physical libraries with actual books, and newspapers printed on paper.

Further reading:

• College students prefer classes with online learning, University of Wisconsin-Madison
• Stanford offers 10 free online Computer Science courses
• From online courses to… automated teaching
• Some myths about online teaching
• Why don’t people use university libraries?

Many consider Frank Hebert’s Dune the most important work of science-fiction ever written. Consider that Star Wars is just a variation on Dune. Yet, it was rejected by more than twenty publishers, before being finally published. It is likely that publishers rejected Dune precisely because it was such a radical departure for the genre.

Of course, being rejected does not mean you are original. It could also mean that you are sloppy or uninteresting. However, there may be valid indications of your originality such as:

• You have no competitor. Nobody quite does what you do.
• You cannot be scooped. You read new issues of journals looking for fresh ideas, but without fear that someone made you irrelevant.

As MacLeod put it: Don’t try to stand out from the crowd; avoid crowds altogether.

Further reading: The secret behind radical innovation and A recipe for interesting Computer Science research papers.

Researchers—at least in Computer Science—spend most of their days at a desk typing. Picking the right software for writing is important.

Most of my writing time is spent on LaTeX documents. I have tried typical Word processors in the past, but they get in my way. Indeed, by mixing document content and document presentation, Microsoft Word makes it difficult to maintain consistency. Word is meant for short-lived (or throw-away) business documents. It is easy to get started and get 80% of the job done with Word. However, as the document gains complexity, as the number of revisions grow, as the number of collaborators expands, Microsoft Word becomes inadequate.

Oh! I still use OpenOffice or Google Docs to produce quick-and-dirty documents. But for anything that is meant to have lasting value, that is research, I refuse to fall into the Word processor trap. It causes some friction with colleagues, but it is a price I am willing to pay.

I believe every single graduate student should learn to write without a word processor. And serious science students should learn LateX. Even if you do not care for LaTeX, at least explore alternatives to Word such as Scrivener.

In any case, you are unlikely to need more than a text editor to write your prose:  Charles Stross, one of the best scifi writer alive, wrote many of his novels using a primitive text editor (Vim). If you have never written without Microsoft Word, how do you know that Word is not holding you back?

Right now, I write using a regular text editor (Smultron for MacOS) and the TeX Live 2009 distribution. I save all my documents to a subversion tree. Using a version control tool such as Subversion makes collaboration easy, and it allows me to go back in time years ago. It is a good setup.

Programming is also a form of writing. For my experimental work, I program in C++, Java or Python, often using Eclipse. I find it is slightly better for programming than my standard writing setup (using only a text editor). Eclipse has great qualities:

• It stays out of the way. In particular, you can collaborate with people who are not using Eclipse without any problem.  For example, it will happily let you use handcrafted makefiles to compile your C++ programs.
• It offers incremental compilation of Java programs. Basically, it compiles as you type.
• It suggests corrections for many common compilation errors.

Essentially, while Java is still an awful language, Java with Eclipse is almost fun. Eclipse proves that sophisticated software can be helpful to programmers and writers.

Writing is hard and it will always be hard, no matter the tool. But at least, ease your pain!

See also Physical tools to improve research productivity.

Physics works with fundamental properties such as mass, speed, acceleration, energy, and so on. Quantum mechanics has a well known trade-off between position and momentum: you can know where I am, or how fast I am going, but not both at the same time.

Algorithms (and their implementations) also have fundamental properties. Running time and memory usage are the obvious ones. In practice, there is often a trade-off between memory usage and the running time: you can a low memory usage, or a short running time, but not both. Michael Mitzenmacher reminded me this morning of another: correctness. On some difficult problems, you can get a low memory usage and a short running time if you accept an approximate solution.

I believe there are other fundamental properties like latency. Consider problems where the volume of the solution and of the input is large: statistics, image processing, finding some subgraph or sublist, text compression, and so on. In such instances, the solution comes out as a stream. You can measure the delay between the input and the output. For example, a program that compresses text by first scanning the whole text might have high latency, even if the overall running time is not large. Similarly, we can give the illusion that a Web browser is faster by beginning the Web page rendering faster, even if the overall running time of the rendering is the same. As another example, I once wrote a paper on computing the running maximum/minimum of an array where latency was an issue.

It would be interesting to come up with a listing of all the fundamental properties of computing.

In the late sixties and seventies, we wanted universities to become more accessible. We founded the Open University, the Université du Québec, and many other universities with accessibility as part of their mandate.

The stated goal was to make degrees more accessible. We succeeded.

Yet, we are now facing an intriguing paradox due to this success. Technology, by making access easier than ever to access educational content, is also shaking the very foundation of the University. As an example of this transformation,  Michael Nielsen was pointing out this morning that you can watch 120 hours of lectures on Physics by Lenny Susskind, for free on YouTube. You are in deep trouble if what you are selling in 2009 are mass-produced lectures. The market price just went through the floor.

Lance Fortnow pointed us to a short essay by Martin Rees about technology and universities. Rees’ point is that technology creates a more level playing field as far as location is concerned. A hundred years ago, airplanes made it possible for Indian Mathematicians to travel to Cambridge where they could be taken seriously. In some sense, airplanes made Indian Mathematicians more globally competitive, though only marginally so. The Web—with repositories such as arXiv—pushes this idea further, an order of magnitude further. After all, Gregori Perelman won a million dollar and the equivalent of a Nobel prize by posting a few papers on arXiv.

The revolution is all around us, not just in Science. Recently, an unknown writer, Sam Landstrom, posted his novel MetaGame on the Amazon Kindle. No publisher, no ad campain. Sales rank of his novel? 540. Considering that Amazon reported selling more ebooks than paper books over Christmas, I am sure many authors envy Landstrom success. Yet, Landstrom did not need an office New York City to either write or publish his book. For all I know, he lives in his parents’ basement.

Thankfully, bona fide Universities have some form of monopoly on University degrees. Yet, like Rees, I think that we are coming to the end of the road for mass universities:

Traditional universities will survive insofar as they offer mentoring and personal contact to their students. But it’s less clear that there will be a future for the ‘mass university’ where the students are offered little more than a passive role in lectures (generally of mediocre quality) with minimal feedback.

One thing is clear to me: The value of a lecture in front of 80 students—or the equivalent as a webcasted show—is exactly zero. (From an educational point of view.)

Disclaimer: Yes, I know that I am making ennemies. Feel free to disagree with me.