EA Interview Series, February 2016: John Salvatier

By Nicole_Ross @ 2016-02-28T18:13 (+8)

The EA Interview Series celebrates the amazing efforts of people in our community. We hope this reminds us of our wonderful achievements thus far and provides newcomers with a personal understanding of our community.

In this interview we want to highlight the work of John Salvatier, a researcher at AI Impacts who helped build Seattle EA and previously earned to give as a software developer at Amazon. Seattle EA was featured in a Seattle Times article about “the philanthropic generation gap” and its effects on arts funding as millennials take a more effectiveness-driven approach. John has presented at various EA meetups on “Why Effective Altruists should care about dressing well.”

John is a great example of how individuals can do good through multiple avenues depending on their skills and interests. If you know a lot about a particular topic, like fashion, you can help others in the community benefit from the key insights in that field. If you’re earning to give, but feel there’s an unusually promising opportunity to do direct good, like working for a nonprofit that focuses on ensuring humanity will benefit from artificial intelligence, then a career switch is possible.


It’s great to speak with you today, John. First, tell us how you got involved with effective altruism and the rationalist community, which I know led you into it.

I’ve been involved in LessWrong for awhile. In college I got really interested in economics blogs and through that got linked to the writings of Eliezer Yudkowsky while he was working on the sequences. I got hooked pretty fast. Eventually someone posted about the New York group; I was excited that there was a community aspect to it. I tried to get a friend to host a LessWrong meetup, which he eventually did. I took over hosting it and did that for several years, but that ended up feeling boring after awhile because LessWrong didn’t have a purpose. I had goals, but people didn’t agree on why we were meeting. So I mostly dropped it for two years, only hosting sporadic events.

Then one day I was driving back from Portland with Eric Rogstad and we had a discussion about effective altruism. He had been following people on Facebook, and I hadn’t really been doing that. He suggested hosting an EA meetup, but I didn’t think I would be that interested so I dropped the idea for awhile. Then, a few months later, I thought back to that conversation. I had been working at Amazon for about a year at that point and felt pretty set with my job and friends, but I felt like I needed a mission or a purpose.

I had long agreed with the arguments for spending a lot more of my resources on making the world a better place and that if I’m going to do that then I should try to do it as well as possible. I didn’t feel motivated at that point, but I thought that if I surrounded myself with people who also agreed with these arguments, then I might become more motivated to think about EA and do things. So Eric and I decided to host a meetup—and then a whole bunch of people showed up. It definitely increased my motivation: it felt like I could help convince other people in addition to figuring things out for myself. From there it went on to build a community I’m very proud of in Seattle. It’s still growing and getting better thanks to Elizabeth, Sydney and the other organizer’s efforts.

Congratulations on helping to build that community. What have you been doing professionally over that period?

I had done stats and software work for a tiny 3-person finance company for a couple years. Then I joined Amazon, and now I’m working for AI Impacts part time and trying to figure out what I want to do in the long run. I recently took a trip to FHI with AI Impacts and learned that technical AI safety work was more within my skill set than I had previously thought. There’s the MIRI style of work that is very logic and math heavy. I don’t have those skills and, while I’m smart, I don’t think I’m that smart, so that felt out of my reach. Some of the other work is more CS, game theory, and economics based, which are areas I’m more comfortable with and I think I could do something in that domain. Owain Evan’s research in value learning turned out to be quite accessible.

I had studied Paper Science and Engineering and Chemical Engineering at the University of Washington. I had been nervous about whether my dad would pay for college, so when a high school friend’s dad told me about the scholarships for the Paper Science and Engineering program, I went for it. I ended up doing a couple internships working in a factory making paper at a paper mill in eastern Washington. I thought it was pretty cool, but the pay wasn’t as good as CS work and it was located in the middle of nowhere.

What led you to leave earning-to-give initially and change your path?

While I was in Seattle I frequently thought about trying something more ambitious. I spent some time thinking about how to fix science and a few other schemes, but it didn’t seem like the plans I was entertaining were going to work. I felt nervous, like I couldn’t really succeed at that kind of thing. Eventually, though, Stephanie Zolayvar convinced me to move to the Bay Area and take AI risk more seriously. Before, I had agreed with the arguments, but didn’t feel motivated to do something about AI risk. I also think a lot of the problem as fear that I couldn’t do anything about AI risk and that it was obvious to others that I couldn’t do anything about it. Stephanie convinced me that it was really important, so I should get over this fear. A lot of this was by her example of taking AI risk seriously and through her conversations with me.

What are you doing with AI Impacts now?

One thing I’m working on now is the AI survey, which we’re hoping to run soon. I’ve been helping design it and have done a bunch of interviews with survey experts to figure out the pitfalls of writing and running surveys. I’ll have a post going up soon about that.

I’ve also been working on some blog posts and helping Katja Grace do some of her research on discontinuous progress in different technologies.


Why blogging?

While I don’t really enjoy writing and would like to get better at it, blogging seems like a useful way of exposing your ideas to a lot of people. I also find it valuable to be able to direct people to a blog post. For example with fashion, it’s helpful to be able to link someone to these posts and then discuss it if they have more questions. I think you’re able to influence a lot more people that way.

Is community building still important to you?

Making the current community bigger seems good but less important to me than trying to make it better. I stay in contact with various people in the Seattle EA group and talk with them about their efforts in making the community really excellent.

What does an EA movement that’s better look like to you?

On the local level, I think being better is having more and better discussions about what the good things to do are, being able to argue with each other better and touch on lots of different topics, even sensitive topics, and to be able to convince each other of things better. I also think being better includes helping each other along the skill pipeline so that people are able to do more direct work or more effective earning-to-give. I’d also like to see more of the good aspects from the rationalist community; there are real skills and techniques to learn from that sphere.

Do you have any role models?


I haven’t had many role models in my life, but I think Stephanie has been the major influence in my life that’s role model-like. I think of her as a mentor and have been inspired to take things more seriously, to pay more attention to how I think and to try to think in better ways by her. Her level of commitment to not letting the world die is extremely impressive to me and makes me want to tap into that commitment within myself more.

What are your hobbies/what do you do outside of work?

I’m still contributing to PyMC3, which is an open source project for doing Bayesian statistical inference. I’ve been involved with it for a long time and, while I don’t think that it’s an especially effective use of my time, it is tied in with my identity and my pride in myself.

I’ve also started taking West Coast Swing with a couple other EAs and rationalists, which has been pretty fun. I’ve taken some dance classes before, which sort of worked, but it feels like I got this more. I also play games like Space Alert occasionally and enjoy hanging out and bullshitting with people.


Thanks for your time, John. We’re really impressed with your work and commitment to EA.




The EA Interview Series is currently produced by Jacy Reese, research associate at Animal Charity Evaluators, and Nicole Ross, operations associate at GiveWell. Disclaimer: the views expressed above are those of John, Jacy, and Nicole, not their employers. We hope to continue producing one interview each month, as a way to highlight the incredible work being done in our community. There’s a good chance the people running this project will change over time as people have shifting availabilities.