I've been part of Google Research for about ten years now, and I'm also an eager reader of histories of classic research organizations like Bell Labs or Xerox PARC. So I thought I would bore you all today with my theory that research organizations are best explained as providing a combination of three things to the company that funds them: a portfolio, prestige, and a prize.
"Portfolio" is my shorthand for the conventional explanation of research orgs -- they consist of a collection of risky projects with individually low chances of success but hopefully high returns conditional on success. Similar to a mutual or venture capital fund, the research org lowers variance by holding at least a handful of such projects.
I won't deny that research organizations sometimes deliver net profits, but I don't think that is their main objective, because the sorts of projects they fund look quite different than the ones funded by purely profit motivated actors like VCs. However, the sort of projects they fund do often look like they were picked so as to maximize their prestige: they try to hire distinguished professors to work on them, they try to work on impressive sounding areas like "quantum computing", they measure their success by how often they publish in highly selective journals, etc.
And let me add that I think that gaining prestige is a very rational thing for many large companies to do! If you have a lot of money, turning some of that money into prestige will open doors that money alone will not. Prestige will help with virtually every action a company can take: hiring, retention, advertising, improving public relations, swaying legislation, influencing standards committees, the list goes on and on. Prestige means other actors (including employees, antitrust regulators, and zoning officials) will look up to you, try to ally with you, and try to emulate you.
[On the other hand, it probably isn't fully rational for all of us to give so much slack to prestigious individuals or companies. On the other other hand, prestige is very plausibly the glue that holds together human civilization, so maybe we shouldn't rebel against that particular instinct.]
Finally, there is prize. Simply put, a position in a research organization can be a valuable carrot to dangle. There are some engineers, say, who wouldn't work on ads optimization for ten years in exchange for money, but would in exchange for a sinecure that lets them research in perpetuity the application of Hyperkahler manifolds in understanding consciousness. (It probably says something that it took me a full minute to come up with an example that was both prima facie ridiculous and also not close enough to any current GR project that I would be accidentally offending.)
There are some obvious synergies between these three P's. Most immediately, prizes wouldn't even be regarded as prizes unless they came with prestige. Less obviously, making money can be prestigious as long as certain conditions are met; for example, if you had to solve a previously recognized hard problem to do so. Sphere packing and machine learning both became more prestigious after they became useful.
Here's one application of my theory: if you manage your research organization purely to maximize its chance of profit, you are going to have a bad time. You won't be able to hire or retain impressive researchers (who lend prestige to your institution purely by associating with it and not necessarily by doing anything!) and many of your most impressive projects won't be able to justify their costs because their benefits will be published and public. And the worst of all, the damage will be almost entirely invisible, as it will consist of people applying to work at your competitors because they do more cool stuff.
[Linkpost for https://thomaswc.com/blog/my_theory_of_research_organizations.html ]
I've been part of Google Research for about ten years now, and I'm also an eager reader of histories of classic research organizations like Bell Labs or Xerox PARC. So I thought I would bore you all today with my theory that research organizations are best explained as providing a combination of three things to the company that funds them: a portfolio, prestige, and a prize.
"Portfolio" is my shorthand for the conventional explanation of research orgs -- they consist of a collection of risky projects with individually low chances of success but hopefully high returns conditional on success. Similar to a mutual or venture capital fund, the research org lowers variance by holding at least a handful of such projects.
I won't deny that research organizations sometimes deliver net profits, but I don't think that is their main objective, because the sorts of projects they fund look quite different than the ones funded by purely profit motivated actors like VCs. However, the sort of projects they fund do often look like they were picked so as to maximize their prestige: they try to hire distinguished professors to work on them, they try to work on impressive sounding areas like "quantum computing", they measure their success by how often they publish in highly selective journals, etc.
And let me add that I think that gaining prestige is a very rational thing for many large companies to do! If you have a lot of money, turning some of that money into prestige will open doors that money alone will not. Prestige will help with virtually every action a company can take: hiring, retention, advertising, improving public relations, swaying legislation, influencing standards committees, the list goes on and on. Prestige means other actors (including employees, antitrust regulators, and zoning officials) will look up to you, try to ally with you, and try to emulate you.
[On the other hand, it probably isn't fully rational for all of us to give so much slack to prestigious individuals or companies. On the other other hand, prestige is very plausibly the glue that holds together human civilization, so maybe we shouldn't rebel against that particular instinct.]
Finally, there is prize. Simply put, a position in a research organization can be a valuable carrot to dangle. There are some engineers, say, who wouldn't work on ads optimization for ten years in exchange for money, but would in exchange for a sinecure that lets them research in perpetuity the application of Hyperkahler manifolds in understanding consciousness. (It probably says something that it took me a full minute to come up with an example that was both prima facie ridiculous and also not close enough to any current GR project that I would be accidentally offending.)
There are some obvious synergies between these three P's. Most immediately, prizes wouldn't even be regarded as prizes unless they came with prestige. Less obviously, making money can be prestigious as long as certain conditions are met; for example, if you had to solve a previously recognized hard problem to do so. Sphere packing and machine learning both became more prestigious after they became useful.
Here's one application of my theory: if you manage your research organization purely to maximize its chance of profit, you are going to have a bad time. You won't be able to hire or retain impressive researchers (who lend prestige to your institution purely by associating with it and not necessarily by doing anything!) and many of your most impressive projects won't be able to justify their costs because their benefits will be published and public. And the worst of all, the damage will be almost entirely invisible, as it will consist of people applying to work at your competitors because they do more cool stuff.