In which I collect tips from esteemed and eminent minds about how to go about pro-actively discovering stuff. More meta-tips than detailed agendas of discovery.
Miscellany of career tips
A current meta-question. One starting point: John Schulman’s Opinionated Guide to ML Research, which discusses stuff like this:
Idea-Driven vs Goal-Driven Research
Roughly speaking, there are two different ways that you might go about deciding what to work on next.
- Idea-driven. Follow some sector of the literature. As you read a paper showing how to do X, you have an idea of how to do X even better. Then you embark on a project to test your idea.
- Goal-driven. Develop a vision of some new AI capabilities you'd like to achieve, and solve problems that bring you closer to that goal. (Below, I give a couple case studies from my own research, including the goal of using reinforcement learning for 3D humanoid locomotion.) In your experimentation, you test a variety of existing methods from the literature, and then you develop your own methods that improve on them.
John links to some other articles on this theme. Richard Hamming You and your research based off his work at Bell labs has some interesting thoughts on social engineering and genius and effort.
On this matter of drive Edison says, “Genius is 99% perspiration and 1% inspiration.” He may have been exaggerating, but the idea is that solid work, steadily applied, gets you surprisingly far. The steady application of effort with a little bit more work, intelligently applied is what does it. That’s the trouble; drive, misapplied, doesn’t get you anywhere. I’ve often wondered why so many of my good friends at Bell Labs who worked as hard or harder than I did, didn’t have so much to show for it.
The notion of stereotype threat has gotten some press lately; I want to also bring people’s attention to the slightly less-discussed notion of solo-status.
Michael Nielson, Principles of Effective Research
People who concentrate mostly on self-development usually make early exits from their research careers. They may be brilliant and knowledgeable, but they fail to realize their responsibility to make a contribution to the wider community. The academic system usually ensures that this failure is recognized, and they consequently have great difficulty getting jobs. Although this is an important problem, in this essay I will focus mostly on the converse problem, the problem of focusing too much on creative research, to the exclusion of self-development.
How precious is my idea?
My understanding about which parts of research are hard and which things are valuable has changed.
Let us suppose that I was worried about being “scooped,” beaten to publication, which is a thing that people in science tend to worry about a little bit but not so much as I imagined before I joined in for myself. How valuable is an insight? It depends. An insight like “this might be worth trying who knows” is not valueless, but the opportunity cost of trying it is possibly months of specialist time or years of grad student time. Let us call such intuitions. Even if the certainty is high, the exercise cost is also high, so it is not worth keeping this kind of thing quiet; rather, you want to share this kind of thing as widely as possible so the work and risk can be shared between multiple researchers. Having good intuitions is useful but if anything you want to cultivate the
If my idea is “I can definitely solve this particular problem in a particular way,” the value of this intelligence is higher. Let us call these solutions. Often, simply knowing a solution exists in a particular domain can lead someone else rapidly to a solution because it constrains the search for solutions enough to allow for other people to find it quickly. This is because most of mathematics, at least for me, is trying bone-headed things that turn out to be silly, and merely narrowing the search space can make things drastically easier. So if you were in a competitive, idea stealing environment, existence of solutions tends to be valuable intelligence. It is not the whole way to a paper which will get you credit for being clever (and I personally value a paper which has more than one great idea in it) but a good solution is an important chunk.
On the other hand, academia can often operate like a cooperative endeavour, or even an competitively cooperative potlatch and giving away valuable things is a way to signal status and attract collaborators and funding.
Sidling up to the truth.
As a researcher motivated by big picture ideas (How can we survive on planet earth without being consumed in battles over dwindling resources and environmental crises?) as much as aesthetic ideas, I am sometimes considered damaged goods. A fine scientist, many claim, is safely myopic, the job of discovery being detailed piecework.
On one hand, the various research initiatives that pay my way are tied to various real world goals (“Predict financial crises!”, “Tell us the future of the climate!”). On the other, researchers involved tell me that it is useless to try and solve these large issues wholesale, but that one must identify small retail questions that one can hope to make progress on. On the other hand, they’ve just agreed to take a lot of money to solve big problems. In this school, then, the presumed logic is that one takes a large research grant to strike a light on lots of small problems that lie in the penumbra of the large issue, in the hope that one is flares up to illuminate the shade. Or burns the lot to the ground. The example given by the Oxonian scholar who most recently expounded this to me was Paul David and the path dependence of the QWERTY keyboard. Deep issue of the contingency of the world, seen through the tiny window opened by substandard keyboard design.
Truth, in these formulations, is a cat: don’t look at it directly or it will perversely slope off to rub against someone else’s leg. Your acceptance is all in the sidling-up, the feigning disinterest, and waiting for truth to come up and show you its belly. I’m not sure I’m am persuaded by this. It’s the kind of science that would be expounded in an education film directed by Alejandro Jodorowsky.
On the other hand, I’m not sure that I buy the grant-maker’s side of this story either, at least the story that grant-makers seem to expound in Australia, which is that they give out money to go out and find something out. There are productivity outcomes on the application form where you fill out the goals that your research will fulfill; This rules out much of the research done, by restricting you largely to marginally refining a known-good idea rather than trying something new. I romantically imagine that in much research, you would not know what you were discovering in advance.
The compromise is that we meet in the middle and swap platitudes. We will “improve our understanding of X”, we will “find strategies to better manage Y”. We certainly don’t mention that we might spend a while pondering keyboard layouts when the folks ask us to work out how to manage a complex non-linear economy.
Disruption by field outsiders
Is it just stirring the pot? How many, to choose an example, physicists, can get published by ignoring everyone else’s advances?
How do you know that your left field idea is a radically simple left-field idea that causes the entire field to advance? And how do you know that it is not the crazed ramblings of someone missing the advances of the last several decades, an asylum inmate wandering out of the walled disciplinary asylum in a dressing gown, railing against the Vietnam War?
Alon, Uri. 2009. “How to Choose a Good Scientific Problem.” Molecular Cell 35 (6): 726–28. https://doi.org/10.1016/j.molcel.2009.09.013.
Arbesman, Samuel, and Nicholas A Christakis. 2011. “Eurekometrics: Analyzing the Nature of Discovery.” PLoS Comput Biol 7 (6): –1002072. https://doi.org/10.1371/journal.pcbi.1002072.
Azoulay, Pierre, Christian Fons-Rosen, and Joshua S. Graff Zivin. 2015. “Does Science Advance One Funeral at a Time?” Working Paper 21788. National Bureau of Economic Research. https://doi.org/10.3386/w21788.
Duchin, Moon. 2004. “The Sexual Politics of Genius,” 34.
Dyba, Tore, Barbara A. Kitchenham, and Magne Jorgensen. 2005. “Evidence-Based Software Engineering for Practitioners.” IEEE Software, 2005.
Newman, M. E. J. 2009. “The First-Mover Advantage in Scientific Publication.” EPL (Europhysics Letters) 86 (6): 68001. https://doi.org/10.1209/0295-5075/86/68001.
Nissen, Silas B., Tali Magidson, Kevin Gross, and Carl T. Bergstrom. 2016. “Publication Bias and the Canonization of False Facts,” September. http://arxiv.org/abs/1609.00494.
Rekdal, Ole Bjørn. 2014. “Academic Urban Legends.” Social Studies of Science 44 (4): 638–54. https://doi.org/10.1177/0306312714535679.
Thagard, Paul. 1993. “Societies of Minds: Science as Distributed Computing.” Studies in History and Philosophy of Modern Physics 24: 49.
———. 1997. “Collaborative Knowledge.” Noûs 31 (2): 242–61.
———. 2005. “How to Be a Successful Scientist.” Scientific and Technological Thinking, 159–71.
Weng, L, A Flammini, A Vespignani, F Menczer, L Weng, A Flammini, A Vespignani, and F Menczer. 2012. “Competition Among Memes in a World with Limited Attention.” Scientific Reports 2. https://doi.org/10.1038/srep00335.