I wrote these up for myself: it is advice for me, based on my own experience and personality. But you may find some of it helpful!
1) Err on the side of doing less
One of my strengths is that I am curious and enthusiastic about many different things. But this is a weakness when I spread myself too thin. Cal Newport writes about this problem a lot: while overachievers have a natural tendency to take more classes, write more papers, take on more activities, it’s often more effective – and more enjoyable – to do only one or two things and do them really well.
In 2017, I did my best work during a semester and a summer where I took no classes and had few formal responsibilities. That allowed me to dig deep into two papers and write them well (enough to pass). My natural tendency is: take a computer science class, take a math class, audit a philosophy of science class, take ballroom dancing, learn to draw, etc. etc. etc. But in those months that I was able to rein in this impulse, and just do one thing, I did better work and was more relaxed.
Implementation (for me): take one machine learning class extremely thoroughly – not two moderately well. And don’t use twitter or Facebook, which are basically the ultimate spreading-thin-machines.
2) Learn better by writing and explaining things to other people
I went to a lot of workshops and conferences this year. At one of them, a particularly complicated one on developmental psychology, my friend started asking me after each session: “What were the talks we heard? What were their main claims, and what were the primary pieces of evidence?” We found this astonishingly difficult to do from memory – a sign that we were not absorbing the talks as well as we thought we were (I suspect we were not alone in this). Active recall exercises like this are a great way to combat the illusion of understanding. If you can’t explain something in full sentences, then you don’t actually understand it yourself. The challenge of explanation can reveal a hidden lack of understanding.
Implementation: after a talk or a lecture, ask a friend if they can repeat from memory what the main claim of the talk was, and the evidence for this claim. Implementation (for me): write two paper summaries in January, and post them to this blog. I’ll summarize these two papers by the end of the month: Concrete Problems in AI Safety, and Building Machines That Learn and Think Like People.
3) When under pressure, resist the urge to isolate yourself
Last year, I had a couple of crisis points when I had huge deadlines looming. My reaction was to burrow away in my apartment like a hermit and go a little bit crazy. As the deadline got nearer, my mind started telling me things like, “now it’s time for an epic push to save everything!!!!”. I’ve learned that when I start thinking about EPIC WORK TIME, it’s actually a warning sign that I’m about to break down and stop working. I usually don’t have a problem with motivation–at these times, what I need to do is: calm down, think very small, and reach out to friends for help.
Implementation: get accountability from friends, and work dates set up, in the run-ups to big paper deadlines
4) If you’re not writing, it’s because you are not scheduling time to write
This is a more specific instance of: If you’re not phi-ing, it’s because you are not scheduling time to phi, which is true for many phis.
Implementation: In January, I am going to write at least two hours a day. I will write 30 minutes immediately upon waking up, and the rest during the day at a time scheduled the night before.
5) Don’t be complacent!
This summer, I flunked Tyler Cowen’s complacency quiz, which measures how routinized and boring your life has become. Without endorsing Cowen’s arguments about what complacency is and what causes it, I used the anti-complacency tips that come with the quiz to make myself less boring in small ways: asking real-live people on dates, wandering around the city without an agenda, listening to new music, and so forth. Cowen’s advice lead to many mini-adventures last year. I will keep following it in 2018.
Implementation: see for yourself by taking the quiz! http://tylercowen.com/complacent-class-quiz/