'Pluribus' as AI Allegory
Let's talk about Pluribus. Admittedly, I was a bit late in starting to watch it, but after hearing non-stop buzz about how good it is, I gave in...
The first episode is good, but also a bit misleading for the rest of the show thus far. It's a full-on zombie apocalypse situation. Yes, people are raised from the dead Night King-style with smiles upon their faces, but still, that's what it is. And this leads you down the path of thinking this is setting up to be a social commentary on the "hive mind". That is also nothing new with it comes to science fiction, of course. But it does feel more timely than ever in our current culture.
But then it keeps going. And now I'm here wondering if this isn't in fact meant to be a commentary about AI? To be clear and fair, that's probably projecting a bit much just given what I tend to write about. But it also works! Almost perfectly. Large Language Models in particular. Watching Pluribus feels like a statement about using ChatGPT!
Granted, the allegory may break down if you simply look at the timeline for when Vince Gilligan (Breaking Bad, Better Call Saul, etc) was writing this show. ChatGPT, of course, launched in November 2022 – happy 3 year birthday next week! – and that very tool tells me that Gilligan was writing the show in 2023 and into 2024. Now it's certainly possible that he was ChatGPT-pilled by then, but it also feels fairly unlikely as he's not a person who, say, lives and works in Silicon Valley. Still, it's possible!
Regardless, it doesn't really matter. Because at least through four episodes, Pluribus works as a pretty great analogy for LLM-based AI. To wit: All of humanity's knowledge is involuntarily uploaded to the collective "cloud", as it were. With this, any of the beings on Earth – aside from a handful that the virus couldn't infect for unknown reasons1 – have access to all of this information in real time. And any of the "unenlightened", when they ask any question of these connected beings, they get back a factually true, but bland response in real time. Have I mentioned the responses are highly sycophantic?
I'm watching this and as the story unfolds, I'm thinking at times that it's almost too much on the nose for our current situation. Again, I'm not sure Gilligan meant this as an AI allegory versus simply a "hive mind" one about broader culture – and I'm specifically not reading about such things so as not to spoil anything for me, or alter my own thoughts on the matter – but it feels almost too perfect.
Beyond just the high-level stuff, it actually goes deeper with power conservation and, in the most recent episode, even prompt injecting! By literal injection!
How's that for on the nose?!
Again, maybe I'm reading too much into this. But I'm not sure that I am. And I'm going to keep going with it until proven otherwise in future episodes.
Overall, I like the show thus far, but don't love it in the way that others seem to. It's fun and interesting and at times good, but also slow. The same could be said of Gilligan's other shows, I suppose. But this one has the most clear Lost vibes. But you know, hand grenade aside, there's not a lot of action.
I'm assuming that changes as the season crosses the halfway point. Maybe we don't get smoke monsters, but instead some agentic beings meant to convert Carol at all costs – perhaps quite literally. Something to watch, quite literally.