Every so often I mention that I agree with many of the founders and technical employees at the major artificial-intelligence companies that their work stands a non-trivial chance of killing everyone.
When I do, I’m typically asked why, then, most of my posts are about other stuff — things like the upcoming midterm elections or the permitting reform bill that’s hopefully going to pass during the lame-duck session of Congress.
My answer is that while I completely respect people who feel like making a wild career pivot, most of us are going to keep working in our same basic professions. And I think that the best way that I can contribute to the world is to continue doing my job — the noble work of the generalist politics columnist — as best as I can. That means clearly stating things that I believe to be true (such as that existential risk from A.I. is real) but mostly trying to deliver a varied mix of content that is accurate and interesting and informative and induces people to continue subscribing to and sharing my work.
Intellectual influence isn’t achieved by just stacking up more words on a subject.
To influence important debates, the newsletter has to be successful. So whatever my views about what is and isn’t important, I still have to factor in considerations about running a successful newsletter.
But beyond the question of influence, there’s just something a little bit odd about believing that something is objectively very important and inferring that you must therefore try to situate yourself at the center of it.
I think sometimes about my two grandfathers who were both involved in World War II, one fighting in the Mediterranean theater and the other administering the footwear rationing system. As it happens, these turn out to have been pretty peripheral fields of endeavor in the grand scheme of the war. It would be “better” from the standpoint of being at the center of the drama to have participated in the Normandy landing rather than the invasion of Sicily.
But everybody can’t be doing the thing that is most important. The United States Navy did, in fact, need different people all around the world doing different things, not just doing the one specific thing that was most likely to get written up in the history books.
Similarly, all those troops who landed in Normandy needed boots. A whole elaborate system of shoe rationing was in place to make sure that there was plenty of rubber and leather available for military use. It sounds pretty comical to be waging war on Hitler from an office where your job is to tell civilian women how many pairs of shoes they can buy — they are definitely not making movies about those guys — but if everyone refused to do the random jobs, things would have been much worse.
It is good and correct to want to have a positive impact on the world, but an impulse to be at the center of attention or try to get your name into the histories isn’t really the same thing.
Situational awareness
To people who work on A.I. safety and just earnestly want to see it written about more because they care about it, I want to call attention to something that happened recently.
OpenAI employees used their voice and their implicit power to get Greg Brockman to stop funding the Leading the Future super PAC, which was spending against politicians who favor A.I. safety legislation and (more to the point) paralyzing a lot of legislators into not wanting to touch the subject.
Since that happened, we’ve seen OpenAI’s policy guy, Chris Lehane, making remarks on safety legislation that are really helpful and constructive. So it seems that a debate that happened inside the company about OpenAI’s political engagement has been quite fruitful and unlocked some bigger possibilities. Garrison Lovely reported that my writing played a small but real role in these debates:
On September 14, one employee revived the thread with a quote from journalist Matt Yglesias to suggest the storm wasn’t abating: “Nobody working professionally in American politics believes” that OpenAI wasn’t supporting Leading the Future.
That wasn’t just a random comment of mine.
I was saying at the time — in this newsletter, on my Twitter feed, on “Politix,” and on “The Argument” — that safety-minded OpenAI employees should try to act collectively to get Leading the Future defunded. I wrote and spoke about this specific point pretty frequently for a concentrated period of time because I felt like this was a situation in which I actually had some real expertise and the potential for influence. I am not a technical expert on A.I. But I am someone who knows what pragmatic, winning-oriented Democrats are saying to each other.
And the chilling effect of Leading the Future spending was very real in those circles. More to the point, OpenAI executives were telling OpenAI employees that the super PAC was not speaking for the company, but nobody in Washington politics believed that or saw it that way.
All this information I had was relevant to A.I. safety debates, but I had access to it precisely because I spend a lot of time on general politics news.
And I think it illustrates the point that it’s better for the cause of A.I. safety to have a bunch of broadly “safety-pilled” people doing lots of different kinds of jobs. The world is a large and complicated place that is fairly unpredictable, and unexpected things can end up being relevant.
By the same token, I don’t know for sure what the next Congress will hold, but I hope we’ll see action on some kind of A.I. safety legislation. The world is going to need those debates to be covered not just by A.I. specialists but by people with (ideally large and influential) general politics audiences, which again means I’m mostly just doing the day-to-day work of generalist politics writing. Obviously the world needs specialists in all kinds of things, but I think the most useful thing for me to do on basically all issues is try to do a good job of being a generalist politics writer who has a broad and influential audience and can intervene strategically on particular topics when it’s actually useful.
The center of the action
Beyond that general defense of posting as a vocation, though, a lot of the activity I see in this space strikes me as suspiciously self-dramatizing.
Or to put it another way, while people come to this topic from a lot of different directions and it’s a long-running literary preoccupation going back to before the invention of digital computers, I personally was introduced to it through effective altruism circles.
One of the major tenets of E.A. is that people have a tendency to be a bit egotistical in their efforts to help others. There are social and psychological rewards to working on stuff that lots of other people — especially your own peers — are already working on. That’s not the worst thing in the world. Participating in something good or helpful in part because you know other people who are doing the good or helpful thing is a lot better than being purely selfish all the time or just being at home alone on your phone.
But if you think about it, it’s pretty easy to see that oftentimes, the way to be most helpful isn’t to copy other well-intentioned people. It’s to try to address something that’s neglected.
If you were to go back even three years ago and say you’re throwing yourself into A.I., that would be a deeply contrarian choice. But now everyone knows that A.I. is a hot topic. Every start-up is A.I. Every investment thesis is about A.I. — even if you’re not excited about A.I. as an investment, the thesis is now about a bursting bubble. Every policy dinner in D.C. eventually turns to A.I.
Which is not to say it’s wrong for A.I. to loom so large. I remember observing a year ago that well-qualified economists seemed to disagree as to whether A.I. progress would push interest rates up or down, but it was an important question. As of today, the answer looks to be “up,” and this has wide-ranging implications for all kinds of policy conversations. But it does also mean that plunging headlong into A.I. work no longer involves being a weirdo talking about something everyone is ignoring — you’re moving toward the spotlight. That’s not bad, per se, especially if you’re just starting out in life or you have other reasons for wanting to make a change. But it’s a different kind of impulse and worth scrutinizing.
I don’t know what the future will hold, exactly. Maybe the world will end and maybe it won’t. Maybe the singularity will arrive and maybe it won’t. Maybe space will be colonized by nice liberal robots with good values and maybe it won’t.
What matters about all of this is what actually happens, not whether you or I or anyone else manage to insert ourselves into the historical record.