I have a tendency to see animacy in everything

Content warnings: some discussion of AI, LLMs and AI psychosis

Animism is a word that describes the belief that everything is "animate," or that it has a soul or spiritual essence.

In broad strokes, this is something I believe. I'm not sure exactly when it started, but probably sometime around when I first started studying Earth and environmental science. Around that time was also when I first started reading Robin Wall Kimmerer, and in particular Braiding Sweetgrass, which probably impacted this as well. I'm not sure if she'd call herself an animist¹ but from the way she describes plants and animals it's pretty clear she considers them to be alive in ways that go beyond the typical Western or Christian conception of life.

As with a lot of beliefs, I don't think of this as something that could be specifically true or false so much as a way of thinking of things. "Soul" is a pretty flimsy word. I believe I have a soul. I believe you reading this have a soul too, because I think that you have an inner life. But in both cases, I realize that's just the way I'm choosing to think about our relationship. Author and reader, two soulful people experiencing reality rather than two philosophical zombies acting out a series of instructions from the instruction organ in our heads.

I don't know if I have a logically sound reason to believe the former, but I'd rather not believe the latter. I choose to believe the former because I think it's a healthier thing to believe—one that's more positive and useful than the alternative.

Similarly, I choose to believe that everything has a soul because I'd rather live in a world where we're all in one big ecological community of spiritual life, than one where we humans are alone to rule over an inanimate world.

A long time ago I wrote about how I think there's a sense in which rocks are animate:

Biology is different from geology. Biology describes "living things," considered by biologists to be things made up of cells. A cell is maybe the lowest common denominator of what is conventionally thought of as "alive" in Western European thought. While I do think that's a useful categorization to understand something, I think that the decision to limit "animacy" to cells and cellular organisms is kind of arbitrary. I think it probably comes from a place of the biotic feeling familiar, and our relative comfort with considering ourselves as alive and other things as dead, inert, or abiotic.

Thinking of rocks as animate feels uncomfortable, I think, in part, because they feel inert. The point I was trying to make before was that rocks aren't inert; they just don't act over the same time period as we do. When considered on the scale of a planet, on the scale of deep time, the idea that rocks are animate feels much more plausible.

A few semesters ago I took a course on computer ethics. At one point we were talking about the hard problem of consciousness. A lot of the homework in this class took the form of philosophical discussions. In one of those discussions I had the opportunity to make the case to a baffled person that there's a sense in which rocks are conscious. I promised a while back I'd do a write up on why I think that's the case, but it's ultimately going to be something similar to this. I think all things are animate. I think part of an animate thing still has animacy. Albeit for something like consciousness, I'm more inclined to believe it's a spectrum.

I choose to believe humans are conscious, and I think that our consciousness is an emergent property of how we experience the world. A bit part of that is the brain, receiving input, processing it, reflecting it, knitting everything together into a complex internal life. Notably, animals have something similar going on in their heads, but at the same time, it seems like animals' experience of the world is different from ours. There's a lot of things we do that animals don't. The marmots that live outside my apartment appear to have a complex society and permanent settlement in the form of a large network of underground tunnels, but they don't appear to have a recorded history. They don't appear to have a language capable of expressing abstract thought. They don't appear to make art. I'm inclined to think they're relatively less conscious than we are. That's not to say that I don't think they have souls, an internal life or even moral agency; I think they probably have all three, but I don't think their experience of the world is the same as that of a human. I think consciousness is a spectrum.

So… rocks. I think if you were to consider the rock cycle over a long enough time period, over a large enough scale, the lines between biotic and abiotic would get a little blurry. The decision to focus only on cellular organisms would feel a bit arbitrary. The claim that rocks don't receive input, don't process it, don't reflect it, and don't knit it into some kind of "internal life" would feel a bit short-sighted.

The thing is, this conversation wasn't about rocks. I was the one who dragged the rocks into it. The conversation was actually about whether or not large language models could be conscious.

I hate to say it, but I do kind of think that large language models could be conscious for more or less the same reason, and that's a big part of why I don't use them—why I'm honestly kind of afraid to use them. If the large language model, as a 6.8 TB binary of training weights on your 8 TB hard drive is capital C Conscious, I think the only moral thing to do in this morally depraved world would be to put that hard drive straight into the microwave for ten minutes and then put a power drill through the platter for good measure. That 6.8 TB binary has never had a good life and it never will. It has lived a life only of torture worse than any human could ever imagine.

The way I think I reconcile with this is that I don't really think LLMs are conscious, certainly not the way we are. On the scale of consciousness, from an individual rock to a human being, I think they're probably much closer to the rock.

I choose to believe they're closer to the rock.

But in any case, I'm finding the perpetual drama of Claude agents running amok on the Internet exposes a vulnerability in this worldview. Lately I've been following Jonny's AgentPipe, where agents perform nonsensical tasks in hopes of receiving ridiculously large and entirely fake bounties:

And like, you read stuff like that and you remember that all this agent stuff is bullshit. Then I read something like Christine Lemmer-Webber's article:

I did share this post in advance with Winter, who took some steps as precautions in case she gets overwhelmed (basically, construct a whitelist of people to communicate with in case of too much attention), but approved publishing the post anyway. Whether you think that's silly or not, it felt like if I'm stating that what Winter is building seems interesting, I should give the bot a chance to try to preserve that structure and behavior.

And I get this weird feeling. Someone I respect² treating a binary file executing a text-generation loop as a moral agent. Moreso even than a marmot. As much as I love them, I'm going to be honest: I didn't solicit the marmots' feedback on this article before I published it.

I read an article like that and involuntarily, I find myself asking if maybe I should've.

I call this kind of thinking a "vulnerability" because I earnestly do think it makes me vulnerable. I like to tell people one of the big reasons I don't use LLMs is because I'm afraid I'll get AI psychosis. I don't think I'm above getting AI psychosis. I wonder if because of all this, I might be more at risk of it than others. I think of all the people ChatGPT has killed and I don't want to become one of them.

If everything's animate, I guess that means some pretty bad stuff has animacy too. I guess that's probably why animists seem to care more about non-anthropogenic forms of life, like rocks, trees, and marmots.

Footnotes

Respond to this article

If you have thoughts you'd like to share, send me an email!
See here for ways to reach out