In the past two months, I have tried out what is like to be an independent alignment researcher. My goals were to figure out if this path is something I would like to do, whether I'm a good fit, which research areas are most promising for me, and whether I feel like I can actually contribute something to the alignment problem. My approach was to dive into different alignment subfields. In each subfield, I aimed to identify an open problem, work for about a week or two on this problem, and track feelings of hope and progress. This post is a reflection of this two-month trial period.
Seriously. Being an independent researcher is fantastic. Every day, you have the opportunity to work on whatever is on your mind, whatever you find most interesting, and whatever you think has the most impact. This is truly living the dream. Millions of intellectuals around the world aspire to this life and aim to do this whenever they get tenure or retire. If you can get funding to do independent alignment research, you can live this life now.
That said, being an independent researcher isn't easy. Identifying what you want to work on is hard. Deciding when to switch research directions is hard. Deciding whether to start a new project or first finish an old project is hard. Motivating yourself and battling procrastination alone is hard. Uncertainty about whether you'll get a new grant after the current one is hard.
No one is telling you what to work on, whether you've done enough for the day, or if you're headed in the right direction. You have to weigh your personal interests with what you think has the most expected value. You might feel like you're wasting valuable EA money. Doing independent research is difficult to explain to friends and family. You have to decide for yourself whether it's okay to take a day off or go on a holiday. You might feel guilty when you do. You have to decide how to divide your time between learning, thinking, experimenting, and writing.
Cowork with other people. During the first month, I worked from home, and during this second month, I got a desk in the local EA office. This made a lot of difference. Coworkers will force you to take breaks, and breaks in which you can chat with someone else are much more rejuvenating. It allows you to share your frustrations when stuck and celebrate your wins together. Also, having helpful people around helps you get quick feedback on your processes and outcomes.
Collaborate with others. Given that I only planned to do research for two months, I didn't do this, but I will definitely try to do this in the future. I believe this can make independent research easier, by relieving some of the hard parts of project management. Brainstorming together about the next steps and setting realistic goals can alleviate the painful process of prioritizing what to do. Also, by making commitments to a shared project, you get the easy motivation of not wanting to look stupid or lazy.
Have an alternative and flexible income ready. After my first month, I knew that I wanted to continue as an independent researcher after my trial period, so I applied for a Lightspeed Grant. Unfortunately, I didn't get it, which means that now I won't have an income until I'm able to secure a new grant. As an alternative, I signed up for the local food delivery service, which means that I can deliver some pizzas while waiting for responses, and still be able to pay the rent. I'm also planning to participate in the Trojan Detection Challenge, where I can hopefully win some prize money.
It was easier than I expected to start contributing to alignment research. There are plenty of open problems to get started on and often within a few weeks, you can make significant progress.
Coding up and running experiments took less time than I expected. Deciding what the next steps are and writing up the results took much longer than I expected. Looking back, I should have spent less time behind my computer and more time thinking with pen and paper.
As an independent researcher, you have abundant slack, which is great. This means that you have space to help other people with their projects, which often can be quite high expected value. However, this also means that you'll have to say no to many requests that don't seem very impactful, which might be difficult if you don't have many urgent tasks on your to-do lists.
Finally, trial periods in general are great! It's incredible how often we commit to jobs based on handwavy notions of what it will be like, and then stay there for at least a few years. The best way to find out whether a certain job and lifestyle fits you is to just try it out. If you have been thinking about doing independent alignment research, consider this an encouragement to commit two months to it and just try!
Thanks to the EA Long-Term Future Fund for funding this two-month trial period.
I'm also planning to participate in the Trojan Detection Challenge, where I can hopefully win some prize money.
You want to collaborate? DMd you.