This is an edited transcript of a talk I gave last week at Commonwealth School, a high school in Boston that I attended from 2000 to 2004. I'm typing from memory, so in places it may be closer to what I intended to say than what I actually said.

It's been twenty years since I was a student here, but the place feels very similar. It's good to be back!

I want to start with a conversation, sometime around 2003. I was in the computer lab:

That's not actually the computer lab. Would you believe I looked through four yearbooks and didn't find a single picture of the computer lab?

Anyway, I was talking to a classmate, in front of some computers that looked a lot like that one, and we were having some sort of political or moral argument. They pointed out that what I was arguing people should do didn't match how I lived, or what I was planning to do with my life. This felt like a very dirty trick! They were cheating! Debate is an abstract combat of ideas, it's not supposed to connect to real life, is it?

I spent the next four years not thinking about this tension very much, and when I did I felt uneasy and moved on quickly. But then I met my now-wife, who had come to a pretty different conclusion. She pushed me pretty hard: how could I live a life that was so far from my ideals? How could I justify keeping so much for ourselves when some have so little?

I started thinking: what if I really took this seriously? I talked to others online, and the ideas bouncing around this nascent community coalesced into the effective altruism movement.

While in the movement there were and continue to be lots of different ideas about what good is and how to do more good, I think at the core that ties this group together is the idea of comparing different options and taking optimization seriously.

Over the years I've taken a lot of different approaches to putting effective altruism into practice. Each time I've decided what to do, I've had two main questions:

What needs doing?
What's a good fit for me?

Initially there was very little money, including for projects that I thought were extremely valuable. Perhaps I could earn money and fund these projects?

We called this "earning to give" which is not a new idea. Sometime around 1750, John Wesley, the founder of Methodism, advised his followers to:

Gain all you can
Save all you can
Give all you can

This is somewhat archaic language, but if you read his full sermon, he's exhorting people to earn as much as they can, avoid spending it on unnecessary things so they have as much left over as possible, and then spend that excess to help others.

I decided to start there. I would earn money, and donate half of what I earned. I did not actually succeed at this, at least at first: in 2009, working my first job out of college, I had not understood how taxes worked and failed to budget for them properly. Over the years, however, as my income rose and I got better at budgeting, 50% worked well as a target.

In deciding where to give, one of my main sources was GiveWell's recommendations. This is an organization that evaluates charities on their impact: how will the world be different if your money goes here vs there? They compared many options for making people better off: bednets, textbooks, deworming, clean water, etc. These are all good things to do, but some are much more valuable than others. Because we don't have the ability to give every approach all the money it needs, we need to triage and prioritize.

I spent the next few years earning to give, 2009-2017, and I liked my work a lot. I was learning how to be a software engineer, how the corporate world worked, how to get things done. But I was also increasingly wondering: were there better things I could be doing with my time?

In 2017 I was talking to a friend, Ben Kuhn ('08), who worked at Wave. This was a company founded by some people I knew in the EA movement, trying to improve the handling of remittances. The idea is people come from poor countries to rich countries, work, and then send money home. Unfortunately, they would lose quite a bit to fees. If this process could be automated, you could make a profit charging much lower fees, and so make a bunch of money while also increasing how much money was available for poor families. Wave would accept payments from American debit cards, and deposit the money into mobile money accounts and Kenya, Uganda, and Tanzania.

But what is mobile money? In 2007, Kenya got M-Pesa. This let people send each other money with their phones. Like Venmo, but older, and even with cheap "dumbphones". It was really valuable economically: Suri and Jack (2016) looked at the impact in Kenya, treating where agents happened to open as a natural experiment, and people were much better off.

Ben asked if I wanted to work with him to help build a mobile money system in Ethiopia. I wouldn't need to move there, I could program remotely. This seemed like a really valuable thing to do, and I decided to quit my job at Google and join them.

I ended up liking it a lot, though not the remote aspect. I got to write software for low-end phones and satellite connections, and experience a very different work environment. Unfortunately, after I'd only been there a few months our Ethiopian effort failed, and I was laid off. I do think it was worth trying, but it didn't pan out for me. (The company has since been successful in Senegal and elsewhere.)

I looked at a lot of different options for what I could do next. One that seemed pretty valuable was government. Governments control the flows of large amounts of money, and have even larger influence through their policies. It was election season in Somerville, and I decided to try running for city council. Over the course of the next 3 days I learned something important: this was not work I enjoyed. I spent a lot of time talking to potential constituents in the process of gathering signatures to get on the ballot, and I hadn't previously really understood how so many of my neighbors (that I like a lot!) had very different priorities than I did. They would be telling me about something that was a problem for them in the community, and I would just be thinking: "OK, but this is so much less important than the housing crisis. Can we talk about zoning?" But of course I couldn't say that; a good representative pays attention to what's important to everyone. Still: not for me.

Another thing I tried was an independent research project to evaluate risk from artificial intelligence. I knew some people through the effective altruism movement who are quite concerned that AI systems becoming more capable and powerful could lead to major problems. But then when I talked to AI researchers I knew through school, some of them working on state of the art models at places like Google Brain and DeepMind, they thought these concerns were science fiction speculation. Perhaps if I spent a while talking to both groups I could understand why they disagreed, make up my own mind, and help other people understand the conflict?

I spent 2 months interviewing a bunch of different people, including Dario Amodei, now the CEO of Anthropic, who apparently had more time back then. I wrote up conversation notes for many of these interviews, which I think were helpful to others, but I wasn't able to make much progress on the core disagreements.

I also learned something else: I really didn't like this work. While I liked the interviews themselves, and writing up the notes, overall, the process was just too lonely, with too much time by myself in a room reading things.

Another thing I looked into, though not very deeply, was biological risk. I had read arguments that we should be more concerned about pandemics, and it really is pretty worrying just how vulnerable people are. I remember how when if you put a CD into a computer it used to be that the computer would just start running whatever program the CD said to run. We pretty quickly realized this was a bad idea and stopped making computers do that, but this is essentially what happens when you breathe in a virus: your system starts following those external instructions. But at the time, in my brief looking, there didn't seem to be anything that was a good fit for my skills.

At this point I'd spent a few months thinking about a bunch of different things I could do, tried some of them, didn't really like them, and was getting a bit burnt out on the process. I wasn't finding anything that was a good fit: I wanted to be working in person, in Boston, on something technical. It seemed like earning to give was my only good option. I was somewhat sad about this outcome, since with more money available for the things I thought most needed funding, it seemed like my donations would not be doing as much good as they had been, but it's still seemed like I could be doing a lot of good this way and I knew it was a good fit for me.

In late 2017 I rejoined Google, and resumed earning to give. I was earning more money now, because I was more senior, and still giving 50%. I was also still learning a lot, but different things now. I learned how to make progress on efforts that cut across the company, leading without authority. I spun up in a new domain, ads this time, and got a good enough understanding of the technology that I could see ways to solve old problems. I became a manager, and benefited from some pretty good training, resources, and mentorship there. I was happy, productive, and excited about what my money was able to fund.

Still, I was pretty sure that if there was a way to apply my skills directly, on a project that was a good fit for me in terms of what I needed from a work environment, that would probably be a lot more valuable. In spring 2022 I was talking to a friend at 80,000 Hours, an EA career advice group, and they suggested I look into biosecurity.

I was quite concerned about biorisk, and unlike 5 years earlier they're now were projects where it seemed like my skills were pretty relevant. And it's a good field for in person work if you want to be in Boston!

But why is biorisk concerning? One way to think of it is that there's a big red button. If you press it, the world ends. Now I know none of you would press it, but if you stuck a button like that in the middle of a busy public place, probably someone would. Right now there are probably a few thousand people in the world who, if they really set their mind to, it, could apply their biotech skills to creating something globally catastrophic. As bio gets more accessible (yay!) more people are in a position to press this big red button (uh...) We need a range of new defenses, so we are less biologically vulnerable.

I talked to some people who were starting a new organization, the Nucleic Acid Observatory, that aimed to cover one particular vulnerability. There are two main kinds of pandemic that could be especially bad:

Wildfire: think a worse Ebola. Super infectious, messes you up right away. There are a lot of challenges, but at least you know you have a problem.
Stealth: think a worse HIV. Very infectious, but a long period from when you get infected until you come down with severe symptoms. How could we learn about one of these while there was still time to do something?

This is the core problem the NAO has been trying to solve: how can we detect a stealth pathogen? I decided to join. So what do I actually do now?

Technologically, we have a newly affordable tool that is potentially a really good fit for this detection problem. It's genetic sequencing, which lets you learn which nucleic acid sequences are in some biological sample. Nucleic acids are "A", "C", "T", "G", the genetic code that says how to make humans and all these other living things. Which includes viruses and bacteria. Every organism has its own genetic code, and you can read it with a sequencing machine.

You've probably seen the Biobot numbers for how much covid there is in the Boston wastewater? That's a great system, but the tool they are using, qPCR, depends on knowing exactly what to look for. What's really valuable with sequencing is that you don't need to decide in advance what to look for. You can pull it all in, and then do your work in the computer.

So just like with Biobot we can start with a sample of wastewater, but instead of qPCR we sequence it. This gives us many observations of nucleic acid sequences, representing the vast variety of different things present in the wastewater. One thing we do with this look for human viruses that seem different than you'd expect. Perhaps a natural mutation, perhaps bioengineered, but not a tidy match for the known genetic code of the virus. So this is what I'm doing now: trying to build out an early detection system for engineered stealth pandemics.

To wrap up, I was thinking it might be helpful to look back at a timeline of my different approaches. I'm going to use donations on the y-axis, not because donations are all that matters, but because they illustrate what I've been doing altruistically over this time. This is our family's donations on our combined income, though over most of this period a large majority of the money came from my earning to give because my wife has been doing lower-paying non-profit work:

You can see how as I became more senior as a programmer. I was able to earn quite a bit more, and donate quite a bit more. Then you can see a dip in 2017 when I took a pay cut to work at Wave, and then try to a few different things that didn't end up being a good fit for me before going back into earning to give. In 2022, and especially 2023, you can again see a large drop off as I joined a non-profit halfway through 2022. I'm still giving 50%, though it's now 50% of a much smaller income.

I'm not actually sure it makes sense for me to still be donating half, since the money that goes to pay my salary is donated and maybe I should just take an even lower salary, but I've stuck with it at least for now.

So over the years I've tried to turn my beliefs into practice a few different ways:

Earning to give, trying to bring in as much money as I could so I could send it to projects I thought were really valuable.
Social enterprise, working for a for-profit company that I thought could have a really positive impact in the world.
Trying things to see if they were a good fit, with political office and looking into AI disagreements
Non-profit work, trying to reduce risk from biotechnology at the NAO.

This is one specific path, where I've looked at my options, compared them with my skills and interests, and looked at what is a good fit for me. Other people have ended up in different places, and I think that's really good; I'm glad people in the EA movement are pursuing a wide range of approaches!

I'll stop here. Thank you very much! Questions?

[Note: some of these were in the main Q&A, others were when a group of students came up to me with more questions after, and I mostly don't remember which is which. I also don't recall the order of questions and may have forgotten some. And even more than above these responses are going to be influenced by what I'd say now since I remember the questions better than my answers.]

Q: How do you think about art and music? Are you saying people who want to make the world better shouldn't go into artistic careers?

Art and music clearly bring a lot of joy to a lot of people, and a world without them would be much worse. On the other hand, I think this is somewhere it's helpful to look on the margin: what is the benefit of an additional person going into art or music? What is the benefit of that person going into reducing global poverty, harm to animals, or global catastrophic risk? So many people are eager to get into art and music that we're far from a world in which we suffer from too few options here.

Q: How do you think about careers that cause harm? If you go into earning to give, how do you compare the harms of the work you are doing to the benefits you can have through donation?

First, if you are doing enough harm through your career that even considering the value of your donations, you are causing harm on balance, that's clearly not a good choice. Don't steal people's crypto savings to donate! Figuring out whether this is the case is really hard, and it's even harder because there's a thing that happens where we tend to overestimate the harms of careers, downstream from news prefering to cover the negative aspects of things. A clothing manufacturer polluting or paying people poorly makes the news, but stories about the pollution from what it replaced, what their employees would have been doing otherwise, or the diffuse benefit of somewhat cheaper clothing don't come up. And if it would be ok for someone to work some job and not donate then it's clearly better if they work it and do donate. But looking into the potential harms of your work is an important consideration in earning to give.

Q: What were the specific donation options you saw when you were initially earning to give that were less attractive in 2017?

When I started earning to give, as I said earlier, there was a lot less money for things I thought were really valuable. One of those was giving career advice to people who wanted to do good with their career in a really effective way. I talked earlier about 80,000 Hours, and in 2012 when I first donated to them this made a significant difference to the amount of money they had available. I still think they're doing great work and funding them is valuable, but by 2017 this was clear to a lot of people and they needed my money less.

Q: Why don't effective altruists prioritize giving locally? When I've talked to EAs in the past, they haven't been very receptive to this idea. How can I convince them how important it is?

When EAs work on reducing poverty, the general question is, where can you do the most to improve people's lives for a given amount of funding or effort? In rich countries we have much more extensive social services than in poorer countries, and while things are far from perfect the easy cheap options have already been put in place. A strategy of helping the people physically close to you means that the people who happen to live near rich people get much more help.

If you think EAs are going about this wrong, and you see opportunities to help people in rich countries more cost effectively than in poor countries, I'd encourage you to write up your thoughts and consider posting them on the EA Forum. The community relies heavily on written communication, and I at least find it much easier to sort out a detailed debate if it's in text.

Q: What do you think about asteroid mining?

Some people think this is a good way to acquire valuable resources, other people think it will be a waste of money. I'm happy for the people who think it will be profitable to fund it, and if they're right they'll make a lot of money.

Q: Do you know [specific contra dancer]

Yes!

Q: If you are interested in effective altruism, is there anything you would recommend doing differently in college? Or is this something you would suggest people think about after graduating?

My main advice for college is pretty general, and mostly doesn't depend on whether you're interested in EA: when you are deciding what to study, think about what careers this might lead to and look into whether you'd like them. This may sound obvious, but at least when I was in college and picking classes, this wasn't something I thought about or remember people discussing. Instead, I took linguistics and computer science because they seemed fun. With linguistics this didn't work out too well: while I like undergraduate linguistics I'm much less interested in the kinds of work people do professionally in the field. On the other hand, with computer science, I was very lucky: I graduated with extremely marketable experience at the beginning of a unprecedented demand for software engineers, and I like the work a lot. Of course AI and other changes make seeing into the future this way quite difficult, but I think it's better to try than not!

Comment via: facebook, lesswrong, the EA Forum, mastodon