I’m pivoting from working on myself to trying to understand the world, and in particular some things related to AI risk. I’ll be blogging about it here.
Around September 2015, I decided to make self-improvement a priority. I identified my difficulty noticing or caring about my internal states, well-being, or level of motivation, as a key bottleneck in letting me do the kind of work I want to do, the kind where I have a high positive impact on the world.
A few weeks ago, I did a personal strategic review, and decided that I’d gotten far enough that this no longer felt like the bottleneck I needed to spend my scarce planning time on. I’ll still be working on this, but in a sense I’m letting it go on “autopilot”. This frees up time to start working on my new key bottleneck: understanding the world well enough to reliably pick projects that fulfill my values.
My plan is to treat this research as something like a full-time job. I’ll articulate what decisions I find hard to make or what actions I find hard to evaluate, and what corresponding uncertainty about how the world works - or how the future is likely to unfold - I would need to resolve in order to make a confident decision. Then I’ll research the biggest or most tractable drivers of my uncertainty until I feel that I have my own model of how things work.
At least in the beginning, much of this will be related to AI risk.
Setting my research agenda
The explicit goal of this project is to understand the world well enough to confidently pick a project that makes it better. If I don’t end up with anything near certainty, I at least hope to build a model good enough that, if I start on one project and observe some evidence that should change my mind about what course of action is best, I’ll be able to notice and adjust appropriately.
Learning to do the work
The secondary goal of this project is to learn how to do self-motivated high-engagement work for several hours per day. I expect this to be challenging, not so much for reasons of stamina, as because I seem to find it difficult to prioritize a single project in my life. I expect many of the initial days to reveal impediments I hadn’t considered before, and my plan is to do my honest best not to flinch or hide or pretend that I’m doing the amount of work I’d intended to, but to troubleshoot each problem as it comes up, as many times as it takes.
I’ll keep a process log, both to keep track of what intellectual processes I’m using during my research, and also things that affect my ability to do the work at all (e.g. timing, distractions, level of motivation).
I am creating a new blog specifically for the purpose of summarizing each day’s research at the end of the day. (These may be much less polished and engaging than the posts on my personal blog.) I hope to accomplish a few things by this:
- Metacognition - to summarize what I’ve done, I’ll spend time reflecting on the work I did, which should help me notice if I seem to have done something silly, wasted time or effort, or alternately, done something that worked especially well.
- Record of progress - I expect that in a project like this, one danger is to feel as though I’m not doing anything. A public record summarizing my progress will help me figure out whether this is the case; if I’m going in circles it will be obvious. In addition, if people I know read my research blog and mention things from it, my social reality will reinforce my own internal sense that I’m actually doing something.
- Collaboration - if other people read about what I’m doing, they might be able to save my time by pointing out possible shortcuts or prior work on the questions I’m trying to answer, or might simply engage with me on the research, improving the quality of my thought on it.
In addition, I’ll share my personal notes here, in case anyone wants to read those. I don’t expect much to come of that, but it’s an inexpensive way to increase transparency.