Content note: discussions of death, i.p. x-risk and death from accident. Difficult medical diagnoses. Trying & mostly failing to keep it light.

A while ago, I have been working through AI scenarios to figure out what to do with my life. In this process, I conditioned on some future trajectory and asked: What would I do if…

The idea is then to assign a probability to all those trajectories, and make choices that are high in expected value (and low in downside).

So what about those doom scenarios?

Or put bluntly: What would I do if everyone was dead in 10 years?

Scary question, I’m not putting a lot of probability mass on this, yada yada, but good to have an answer to.

The most interesting thing about this hypothetical is that I can’t think of many things I would change. Once I opt out of my voluntary retirement plan contributions and take out a really long mortgage, what else is there to do?

We should already be spending time with people we love and doing work that feels meaningful to us. If the timeline was two years, liquidating all your assets and touring the world could be a shout (or does this just put you far away from most of your friends?), but with 10 years, that’s not a real option for most people (And again: if it is for you, why aren’t you doing that already if that’s what you want?).

In researching this question, I looked at forum threads of people who have received a medical prognosis of ~10 years. Those reflections mirror mine: focus on your relationships but don’t change much else.

Risk Budgets

Intellectually, this is a bit frustrating. Surely there must be something? So I started thinking about risk. A useful tool here is the Micromort, the 1:1,000,000 chance of dying.

Now, bear with me, dear reader, and suppose there are ways to trade risk for fun. Suppose also that I don’t want to die doing risky things, if not doing them would have been a good option. Under these assumptins, my policy should be, as enshrined inside every London tube carriage:

Do not take any risks.

Ok. But if I’m not being maximally stupid about my policy on risk, then I would instead say that I want my chance of dying from doing avoidable fun risky things to be small, say 1%. Then, assuming a remaining healthy lifespan of 75 years, I could aim to accruse a yearly risk of death from doing fun risky things of about 133 micromorts. If I expected to only live 10 more years, that number goes up to 1000 micromorts!1

Are there lifestyle choices2 that accrue risk slowly over time such that having them around for 10 years is acceptible but 40-60 years is too much?

Exercise and healthy eating are like this. However, they also have immediate benefits to well-being. Stopping them is a bad idea even if you don’t expect to reap the long-term rewards.

I stopped using sunscrean, in part for this reason3. But same as with unhealthy eating, it turns out that getting sunburnt sucks.

Recreational drug use kind of fits the bill. I’ve heard a wise greybeard in that community say that the only people who are still around stopped doing it fairly early in their lives… so if you don’t expect to be around anyways, the calculation here might change. But that’s not my cup of tea for unrelated reasons.

Then I thought about riding a motor cycle, thinking I was finally onto something. Much like extreme sports, going for a motorcycle ride is a lot of fun and probably won’t kill you, but if it’s part of your lifestyle for decades, it just might.

Risk budgets for new skills aren’t spent evenly

So I thought I had come up with a category I could mine for opportunities in the 10 year healthy lifespan case. But then I realized that for activities are the more dangerous the less experience you have. This pops up all over the place once you notice it.

I once worked at a German chemical plant, that manufactured prototypes of semiconductors. The workers would be experienced technicians, performing steps of the process manually, often using incredibly toxic substances. Our supervisor, an old Bavarian who’s been working there since 1945, said that everyone who joins is very careful for the first 6 months, at which point they have their first accident. If they survive, they are careful again for 6 months, etc. At some point the carefulness either sticks, they have a very serious accident, or they find a new job. That implies that the first year on the job is the most dangerous.

Same with some kinds of strength training, where you incur a lot of injury risk up front. After that, you are strong enough to protect yourself and consistent and experienced enough that you can afford to take a day off instead of pushing through pain.

The same is true of driving a car, as reflected by the decrease in insurance costs with driver experience.

And the same, unfortunately, almost necessarily is true for motorcycling, paragliding, and all the other ways we can spend micromorts on fun.

So when it comes to practical impacts of a shortened expected lifespan, it would seem I must return to the drawing board.

Risk always accrues unevenly

Other than frontloading risk when learning a new skill, there are other effects that make the distribution of risk on any occasion of doing an activity uneven.

I think about this when I cycle. Cycling in London is notoriously dangerous, but looking at accident statistics, most incidents occur on wet, foggy November days. Not cycling on days like these, or when I’m feeling tired and distracted, has an outsized impact on the risk I’m incurring. Or in other words, the cost in micromorts on those days is so high that it’s not worth it for the convenience and marginal health benefits I get for it.

Radical Lifespan Extension

Of course, if we allow for the possibility of stopping aging, mortality will just be from accident, homicide, or whatever diseases remain incurable.

At that point, expected lifespan is simply one over the yearly risk of death, which is a constant. If my lifestyle choices can reduce this risk by 10%, that extends my expected lifespan by 11%. So if we’re in any of those worlds, spending micromorts to have fun looks like a bad idea in any case, and we should heed the wisdom of TFL.

  1. Dying earlier is bad because it prevents more life later on, but risky things are probably more fun while you’re young. I’m lazy so I’ll just have the two balance out to a uniform allocation of risk. ↩

  2. I think the term lifestyle here captures the useful concept of a thing that I’m (a) choosing to do, that I could (b) easily choose not to do, and (c) impacts my life in a regular fashion. I think thinking about yourself as adopting a specific lifestyle is epistemically risky. While I’m not fully sold on identity zero, thikng of yourself as e.g. “the sort of person who rides a motorcycle” is probably not doing you any good. ↩

  3. … but mostly because saying “I don’t use sunscreen because of AI timelines” is a good way to troll people. ↩