Why I Want to Work at MIRI

Death sucks.

Death sucks a lot.

I talked about this somewhat in, “On Grief“, about how painful death is on those of us who are still here. Losing people you care about forever is agonizing. To never see them again, never make them smile again, all the unhad conversations and silent moments together, this is unacceptable, but we’re left with it anyway.

There’s also, of course, the loss to the guest of honor at the funeral. All your hopes and dreams, everything you wanted to do and say, the experiences you could have had, all gone. All value gone, turned into a game over screen with no continues and nobody to see it.

We should not leave out the loss to the world, either. A person is gone, taking with them their unique viewpoint and all of their experience. Everything they knew, every skill they had, every idea, gone, just like that, because of a biological hiccup.

It’s unacceptable. It is not to be borne.

The only real resolution that I see to this problem is a superhuman artificial intelligence, aligned to human values. Nothing else offers a complete solution the way more intelligence does. With what we humans have we’ve pushed death back, raised the length of the average lifespan bit by bit as our generations passed, but we still die in the end.

Even if SENS is fully successful, and aging becomes a process that is fully managed and no longer a threat, we still have risks – I’ve been told someone ran the actuarial calculations and found that in the absence of aging we’d average about eight hundred years and die to accidents. It’s certainly far better, but we’d still have to say goodbye.

Aside from that of course we’re still left with the problems of war, disease, and violence. While we’ve reduced these, they’re still factors in the ending of lives, and this doesn’t touch on x-risks – biotech, nanotech, unaligned artificial intelligence, climate change, all threats to us that SENS can’t touch, and solving any of these doesn’t resolve the others, except for AI. An optimizer that’s more intelligent than we are can actually touch these things.

So I’ve made the case now to work in alignment, but why MIRI?

MIRI’s working on a model called HRAD – Highly Reliable Agent Design. Machine learning systems that we have been building thus far are largely trained by rewarding the system for making the ‘right’ choice when given problems, which causes the system to update slightly in that direction. The problem with doing things this way is that we end up with a system that’s a black box – we don’t know what it is that makes it decide what it does in any sort of detail, and giving them novel inputs is a crapshoot – one example is of a system intended to scan photographs for tanks. The researchers fed it a bunch of pictures that had tanks, and a bunch of pictures with out, then tested it on the other photographs in the set, which it classified perfectly.

When they presented it to the military, the military tested it and found that it didn’t do better than chance. It turned out that in all of the training data, the pictures of tanks were taken on cloudy days, and the non-tanks were on sunny days. The system distinguished weather, not tanks.

We need AGI that values what we value, and I think MIRI is doing the best work in this direction, and I want to be a part of it.

Because I want to do the highest leverage thing that I can.

To stop the AI from eating us all.

To end death.

We’re looking at now, now

At the start of this year, we moved. The place where we’d been living had issues, chief amongst which was that there was some sort of compound that it leaked into the air. Several people had noted it, that there was an increase in depression and a reduction in executive function and motivation while they were there, and I’m fairly sure it hit me, too. Certainly I got depressed while I was there, and it shot my production level to hell. Admittedly, I had reason to be sad, having lost a longtime friend, but not like that.

So we moved. Found a place at another rationalist house, known as Tesla House, which is short for Doctor Tesla’s House of Electrical Fun. It acquired its name with an electrified towel rack to start. Not, to be clear, an electric towel rack, which heats to dry the towels. A towel rack that when hung had penetrated electrical cabling, and now provided a shock when touched. That was exciting.

There’s also the interesting way about half of the outlets were wired backwards; less exciting, but still annoying. The house has a few other ‘features’, but I don’t want to spend all of this post talking smack about a place that I’m actually fairly happy in.

As I had time away from Liminal, and with some antidepressants from my doctor, things started to clear up, and with some ADHD meds, I even started to be productive, but it wasn’t until I started a polyphasic sleep experiment that things really went FOOM. I tried an E2 schedule, with 5 hours and ten minutes of sleep daily, and those three hours per day really started to add up.

As well, I think my expectancy started really building during this time. Suddenly I was doing things, in a way I hadn’t before. Instead of being a drag to get things done, it was exciting! I had hope of doing meaningful things in the future, which is a truly novel experience to me – I’ve been used to thinking that I could be useful to others, but that my role was still largely support – to help others rise.

As I’ve mentioned, I record the work I do, and I went from managing 20 hours in a week, barely (and not that, in the worst of my depression) to 40, to 60, to 80. I’ve done 90 in a week, although that feels like an unsustainable amount right now. I average 60-80 now, regularly.

During the first polyphasic adaptation attempt, I was kind of a mess chemically. I was on antidepressants and ADHD meds and THC and nicotine and stimulants – I’ve described it as “running myself like a chemistry experiment,” because I was. This came back to bite me in the ass. 60 days into the attempt, I hit a nasty depressive patch. While snap depressions like that hit me sometimes, I’ve always been able to sleep them off, which wasn’t the case with this. My current belief is that due to all of the things I was on, my sleep quality was too screwed with to be able to manage proper compression, and the stimulants were papering over the sleep debt accumulation, so I looked adapted if you didn’t look too closely, but I clearly wasn’t.

It was during this experiment that I had my falling out with the Fleet crew. They had decided to double down on veganism, and my boots caused an argument that led to us parting ways and my talking to some of my partners about their philosophy. We discussed some of the errors they saw, and on the whole, I think despite having provided some of the elements of my powering up, I’m better off having gone my own way. I’m grateful for what I got from them, but I don’t think that their philosophy or action plans are going to lead to FAI; since that’s what I’m after, it doesn’t really make sense for me to be following along with them.

Instead, I’m working on my own projects. I’m studying math, music, programming, and of course all of the books I’ve been reviewing. I lead an AI safety reading group, I’m still doing experiments to gain more time in my days, I’ve been doing an insane amount of stuff compared with any other time in my life. I quit THC and nicotine, and while the residue stayed in my body long enough to mess up a second polyphasic adaptation attempt, I’m sleeping much better now.

The frustrating thing is that it feels like it’s still not enough. I’m not learning fast enough, not growing fast enough; I have so very far to go to ascend to the level I need to be at. All I can do is keep pushing and trying to grow stronger. At the least, I know I can keep pushing, and I know that that’s the secret to getting anywhere – keep going.

But I need to go faster. I need to get stronger.

One day at a time.