Deliberate Grad School

Among my friends in­ter­ested in ra­tio­nal­ity, effec­tive al­tru­ism, and ex­is­ten­tial risk re­duc­tion, I of­ten hear: “If you want to have a real pos­i­tive im­pact on the world, grad school is a waste of time. It’s bet­ter to use de­liber­ate prac­tice to learn what­ever you need in­stead of work­ing within the con­fines of an in­sti­tu­tion.”

While I’d agree that grad school will not make you do good for the world, if you’re a self-driven per­son who can spend time in a PhD pro­gram de­liber­ately ac­quiring skills and con­nec­tions for mak­ing a pos­i­tive differ­ence, I think you can make grad school a highly pro­duc­tive path, per­haps more so than many al­ter­na­tives. In this post, I want to share some ad­vice that I’ve been re­peat­ing a lot lately for how to do this:

  1. Find a flex­ible pro­gram. PhD pro­grams in math­e­mat­ics, statis­tics, philos­o­phy, and the­o­ret­i­cal com­puter sci­ence tend to give you a great deal of free time and flex­i­bil­ity, pro­vided you can pass the var­i­ous qual­ify­ing ex­ams with­out too much study­ing. By con­trast, sci­ences like biol­ogy and chem­istry can re­quire time-con­sum­ing lab­o­ra­tory work that you can’t always speed through by be­ing clever.

  2. Choose high-im­pact top­ics to learn about. AI safety and ex­is­ten­tial risk re­duc­tion are my fa­vorite ex­am­ples, but there are oth­ers, and I won’t spend more time here ar­gu­ing their case. If you can’t make your the­sis di­rectly about such a topic, choos­ing a re­lated more pop­u­lar topic can give you valuable per­sonal con­nec­tions, and you can still learn what­ever you want dur­ing the spare time a flex­ible pro­gram will af­ford you.

  3. Teach classes. Grad pro­grams that let you teach un­der­grad­u­ate tu­to­rial classes provide a rare op­por­tu­nity to prac­tice en­gag­ing a non-cap­tive au­di­ence. If you just want to work on gen­eral pre­sen­ta­tion skills, maybe you prac­tice on your friends… but your friends already like you. If you want to learn to win over a crowd that isn’t par­tic­u­larly in­ter­ested in you, try teach­ing calcu­lus! I’ve found this skill par­tic­u­larly use­ful when pre­sent­ing AI safety re­search that isn’t yet main­stream, which re­quires care­fully step­ping through ar­gu­ments that are un­fa­mil­iar to the au­di­ence.

  4. Use your free­dom to ac­com­plish things. I used my spare time dur­ing my PhD pro­gram to cofound CFAR, the Cen­ter for Ap­plied Ra­tion­al­ity. Alumni of our work­shops have gone on to do such awe­some things as cre­at­ing the Fu­ture of Life In­sti­tute and sourc­ing a $10MM dona­tion from Elon Musk to fund AI safety re­search. I never would have had the flex­i­bil­ity to vol­un­teer for weeks at a time if I’d been work­ing at a typ­i­cal 9-to-5 or a startup.

  5. Or­ga­nize a grad­u­ate sem­i­nar. Or­ga­niz­ing con­fer­ences is crit­i­cal to get­ting the word out on im­por­tant new re­search, and in fact, run­ning a con­fer­ence on AI safety in Puerto Rico is how FLI was able to bring so many re­searchers to­gether on its Open Let­ter on AI Safety. It’s also where Elon Musk made his dona­tion. Dur­ing grad school, you can get lots of prac­tice or­ga­niz­ing re­search events by run­ning sem­i­nars for your fel­low grad stu­dents. In fact, sev­eral of the or­ga­niz­ers of the FLI con­fer­ence were grad stu­dents.

  6. Get ex­po­sure to ex­perts. A top 10 US school will have pro­fes­sors around that are world-ex­perts on myr­iad top­ics, and you can at­tend de­part­men­tal col­lo­quia to ex­pose your­self to the cut­ting edge of re­search in fields you’re cu­ri­ous about. I reg­u­larly at­tended cog­ni­tive sci­ence and neu­ro­science col­lo­quia dur­ing my PhD in math­e­mat­ics, which gave me many per­spec­tives that I found use­ful work­ing at CFAR.

  7. Learn how pro­duc­tive re­searchers get their work done. Grad school sur­rounds you with re­searchers, and by get­ting ex­posed to how a va­ri­ety of re­searchers do their thing, you can pick and choose from their meth­ods and find what works best for you. For ex­am­ple, I learned from my ad­vi­sor Bernd Sturm­fels that, for me, quickly pass­ing a draft back and forth with a coau­thor can get a pa­per writ­ten much more quickly than ag­o­niz­ing about each re­vi­sion be­fore I share it.

  8. Re­mem­ber you don’t have to stay in academia. If you limit your­self to only do­ing re­search that will get you good post-doc offers, you might find you aren’t able to fo­cus on what seems high­est im­pact (be­cause of­ten what makes a topic high im­pact is that it’s im­por­tant and ne­glected, and if a topic is ne­glected, it might not be trendy enough land you good post-doc). But since grad school is run by pro­fes­sors, be­com­ing a pro­fes­sor is usu­ally the most salient path for­ward for most grad stu­dents, and you might end up pres­sur­ing your­self to fol­low that stan­dards of that path. When I grad­u­ated, I got my top choice of post-doc, but then I de­cided not to take it and to in­stead try earn­ing to give as an al­gorith­mic stock trader, and now I’m a re­search fel­low at MIRI. In ret­ro­spect, I might have done more valuable work dur­ing my PhD it­self if I’d de­cided in ad­vance not to do a typ­i­cal post-doc.

That’s all I have for now. The main sen­ti­ment be­hind most of this, I think, is that you have to be de­liber­ate to get the most out of a PhD pro­gram, rather than pas­sively ex­pect­ing it to make you into any­thing in par­tic­u­lar. Grad school still isn’t for ev­ery­one, and far from it. But if you were se­ri­ously con­sid­er­ing it at some point, and “do some­thing more use­ful” felt like a com­pel­ling rea­son not to go, be sure to first con­sider the most use­ful ver­sion of grad that you could re­li­ably make for your­self… and then de­cide whether or not to do it.

Please email me (last­name@this­do­main.com) if you have more ideas for get­ting the most out of grad school!