Aella is a lot of things to a great many people. She’s a data scientist, a sex-research Substacker, and an orgy organizer. And now she is an AI safety advocate, with Plz Don’t Kill Us, an experimental residency program aiming to create viral content that communicates the potential threats of artificial intelligence. 

Her concern is that the AI safety movement has a messaging problem, with talk of existential risk and alignment failing to impress the nontechnical masses. Aella worries about how little the general public is paying attention, as the stakes are high.

Other groups are also trying to change the conversation. Last year, the Foresight Institute ran an existential meme competition, and January saw the launch of the Frame Fellowship, an eight-week AI safety creator program.

Plz Don’t Kill Us (opens in new tab) is betting on a lighter touch. Cofounded with Ronny Fernandez from Lightcone Infrastructure, a nonprofit focused on developing rationalist communities, the program will offer residencies to up to 100 creators, who will be asked to post daily short-form content about “AI doom.” Food and housing in Berkeley will be provided. Applications are open through June 15. 

Mentors include art-pop musician Grimes; Eliezer Yudkowsky, cofounder of the Machine Intelligence Research Institute; and Liv Boeree, the poker-playing host of the “Win-Win” podcast.

Expectations are high: Selected creators must post every day; if they don’t, they will be asked to leave. The bet is that enough volume might cut through the noise and spread skepticism about the real risks AI poses to society. 

We spoke with Aella about why she developed the program and how this might change the AI discourse. 

This interview has been edited for clarity.

“Plz Don’t Kill Us” is a pretty extreme name. Where did it come from?

When I first started making this project, I was just, let’s call it, “Please don’t kill us.” And everybody was like, “Sounds good.” There was no discussion. I liked that it’s very direct. I’m very pro directly and honestly stating what you believe.

It’s very straightforward. I would prefer not to be killed. It would be sweet if we could have the companies not develop a super intelligence that kills us all.

How worried are you about AI right now?

The world doesn’t understand the magnitude of the thing that is being built right now by a small handful of companies. It’s equivalent, IMO, to nukes. This stuff will very quickly have the potential to be world-altering in ways that none of us really are involved in. 

We didn’t vote for something like this to be so powerful and handled by so few. 

How is that worry affecting you?

I’m nine out of 10 worried. It’s impacting my life plans a lot. I’m doing more drugs and not really saving for retirement. I’m pretty worried, and I’m not super hopeful that we will be able to stop it.

I’ve done a lot of grieving in advance, going through the agonizing process of coming to terms with this as the potential reality. Now I’m like, Let’s just get the thing done that needs to be done and hope for the best.

Have you talked to people at the large AI labs about this?

I have. In general, they sort of think they can control it. [But] if we’re headed toward a world where a couple of super companies are controlling the most powerful weapon on the planet, that’s a problem, even if we’re not going to die. I think a lot of them are wrong and very personally incentivized to not believe that it is a threat. They have a lot of money on the line to keep pushing the “go forward,” “go fast” button. 

Why did you start this project? 

I felt like there was nothing I could do, and I’m just gonna go party, because my skill set is looking at kink data, you know? Then I saw an opportunity. I was, Oh, I have some ability to organize other people and to try to communicate complex topics to a layman audience. So maybe I can help a little bit.

How can content creators help with AI safety?

A lot of people are talking about pretty important stuff that’s very technical. Last year, there were a couple of instances where I watched nerds try to communicate, and oh my God, they don’t know how to talk or get the concepts across. 

I believe short-form content is a lot more important, as people aren’t watching long-form anymore. It’s a really big challenge to figure out how to communicate complex stuff through short-form, but that’s what the experiment is. We’re going to throw a bunch of stuff on the wall and see if we can figure out a good way of doing it. Our dream is to have about 100 creators. Ultimately, we care about accuracy. I would not just accept anybody.

How does your project compare to programs like the Frame Fellowship, a bootcamp for AI safety creators? 

They’re great; we’re just doing slightly different things. Their type of communication is to people who are already one foot in this field, and I want to get to people who aren’t in this field at all. I think both of us need to exist. 

How are you funding this?

We have funding from the Survival & Flourishing Fund and the Machine Intelligence Research Institute. Our goal is $800,000 to cover everything we want. We’re roughly 60% of the way there.

Your mentors for the program include AI safety experts, a poker player, chess streamers the Botez sisters, and the musician Grimes. It’s a wide range of people.

I’m trying to find people who are either really excellent communicators about AI and existential risk in general or who have a lot of experience with art and creativity and fun, like production online for a wide audience.

We were looking for people who understand the field already, people who believe this is a pretty important topic. A lot of people think that it’s probably not going to kill us, but it’s still dangerous, which is totally fine. We want diverse views and to have a discussion about it.

Grimes will stop by at least a little bit, possibly a lot. She’s down to help. But she’s very flexible with her schedule, I think, so we’re probably not going to know until we hit that time.

Any hard rules for content creators?

Our rules are no violent content — that’s in our conduct policy — and no false statements. We will penalize or remove people if they are making false claims to the public. So far, those are the only hard lines. You can use AI [to create content] if you want. 

Even before the [attack on Sam Altman’s home], there have been discussions in the community about what happens if an insane person thinks that doing a random act of violence is going to be useful somehow to AI safety stuff.

We’re like, We should be very explicit that this is not something we’re going to accept or promote.

Do you think there will be more violence toward leaders in AI?

Yes. I have had random acts of attempted violence against me. Anything or anyone that’s high profile, especially when people feel strongly about it, you’re gonna get insane people doing crazy shit. There’s no real way around it. But this is not an effective method of achieving our goals. You’re not going to prevent nuclear war by doing a random act of violence against one of the AI researchers’ houses. It doesn’t work like that.

What result do you hope you get from the residency program?

I really want people to go into action, as opposed to anxiety. Some people have strong anxiety responses to this, and I don’t like that. That’s not my goal. I would really prefer people to call their representatives. It’s cliche, but the politicians need to understand that the public understands that this is a threat to them and want them to be taking it seriously. 

We need everybody in the movement to be able to call for some kind of safety. We need people paying attention to the actual risk going on here.