Computer science research.

[pmb]

Computer science is a very young field. This is sort of trivially true when you compare it to, say, philosophy and mathematics, but also has consequences for my daily life. One thing it means is that seemingly-obvious questions have often not been asked. With a BS in CS you are qualified to begin answering many of them, if the question is explained well.

This is manifestly NOT true in math, and is less true in the more mathematical areas of CS. But for much of it, we don't even know what the right questions are. In theory we have P vs NP as an overriding question and also concerns about quantum computation (yet another area that is intensely mathematical), but in networks, we don't even know what the right questions are. In software engineering, we are feeling around in the dark, and in user interfaces, we just keep throwing things against the wall and hoping they stick. (note that these are broad generalizations, and practitioners in each of these could certainly find counterexamples, but I think the broad gist is true).

If you know some computer science and come up with a question that combines concerns in disparate subfields, the chances are good that your question has never been asked before, much less answered, that both the question and answer may be interesting to more people than just you, and that you may have all the tools you need to solve it just from your undergrad education. Are there any other fields where this is true?

Comments

[freyley.livejournal.com]

stem cell research,
proteinomics,
neuropsychology
many subfields of cognitive science
evolutionary psychology

Many of these you could ask intelligent questions with just a BA, but the equipment needed to answer them is more specialized and expensive.

[freyley.livejournal.com]

also, behavioral economics, which is basically psychology discovered through the use of statistics and large experiments.

[freyley.livejournal.com]

and while I'm on this roll, let me add physics. Because graduate string theory really doesn't teach you anything.

[freyley.livejournal.com]

er, theoretical unified everything physics...yeah, okay, that wasn't a typo, just a casualo.

[bookerz.livejournal.com]

Research, yes. Research that's accepted (or acceptable), less so? http://www.newyorker.com/reporting/2008/07/21/080721fa_fact_wallacewells is a fun yarn about the reception of surfer-scientist Garrett Lisi's string-theory replacement. It qualifies as research done outside the system, if not work that's possible to do with only undergraduate math/physics.

[lindseykuper]

Sounds about right. I might even add straight-up neuroscience. We know so little about how the brain does what it does.

[freyley.livejournal.com]

yeah. The dividing line between we know a lot and we know little is, entertainingly, the same dividing line as in computer science -- we know a fair bit about the individual neuron, but not much about networks of them. Which is, of course, a computation problem, though analog and involving delicate physical apparati.

[pmb.livejournal.com]

I disagree. Neuroscience needs some complicated and expensive equipment to perform its experiments. Unless you are positing some sort of theoretical neuroscience, which may as well be described as "guessing".

[freyley.livejournal.com]

We're all confusing two separate issues, and I don't know which is of importance to you.

* the ability to ask useful questions; to design useful experiments; the experience to do the above well
* the equipment to do those experiments

Computer science is special in that the equipment is cheap and readily available. The other part, lots of fields have. Most of those other fields either require expensive equipment or large numbers of victimssubjects.

[lindseykuper]

I'm not sure I believe that CS research is cheap, not when it comes to the specific subfields Peter mentioned. If one wants to explore new questions in networks and UI, I would think one would need to build networks and UIs that nobody has built before. Actually, to get there, one would probably need to build, throw away, build, throw away...it's starting to sound expensive.

I'm studying programming languages. The equipment is definitely cheap and readily available. But it's one of the older areas of CS. I don't think a BS is enough to start answering the questions in PL.

Then again, what do I know? I don't have a BS.

[freyley.livejournal.com]

Networks and UI research are still cheap when compared to anything else we've talked about on this post. He's making the assumption that your labor is free, which is a reasonable assumption in the case of a research interest. However, if it weren't, it would raise all the fields equally -- no new question is easy to answer, all require significant labor.

And setting up a dozen computers costs very little compared to building neuroimaging.

[lindseykuper]

And setting up a dozen computers costs very little

But we're talking about answering new questions. Aren't a lot of those questions going to be questions of scale?

[freyley.livejournal.com]

Are they?

I've no idea.

[lindseykuper]

Yeah, I think so. One of the reasons computer science keeps being so interesting is that we keep having to deal with problems that we haven't encountered before in terms of scale.

I saw Peter Norvig talk a few weeks ago when I was in Mountain View, and the gist of his talk was "More data, more better." But how are we going to deal with all that data? Every day we have more than we had yesterday! Better figure it out.

Indeed, one reason I'm so excited about programming that treats computation as the evaluation of mathematical functions is that it can be a natural way to approach solving huge-scale problems which want to be broken up into lots of pieces to be solved independently and in parallel.

[pmb.livejournal.com]

Although it should be pointed out that Peter Norvig is at the forefront of a group of AI researchers who think that intelligence is based on lots and lots of data with relatively simple algorithms.

I think in CS we have more problems than scale, although in networks there' a lot of scaling-type-stuff. But most people build and use simulators rather than building out any hardware, and simulators can run on just about any hardware.

"There are only two hard problems in Computer Science: cache invalidation and naming things. Phil Karlton"