Last night I went to a local lecture series on “neat science“ stuff. There were some interesting things on the bill, but ultimately it was a big letdown. It was more revelatory about the people and the culture of LLMs than the tech itself. In that regard my experience ranged from mild diversion to actual, physical nausea.
To keep it succinct, one of the presenters had a fairly interesting project to create an LLM religion. One of the saving graces here was that it involved physical materials being shaped by actual humans to create religious iconography. The aesthetic quality of the imagery was late ‘80s/early ‘90s computer art, so make that what you will, at least it was a try. The tech gimmick was that the LLM intakes various spiritual texts, and then generates text/vocalized outputs responding to questions. The LLM was designed to have an oracular tone to the vocabulary strings it produced, essentially producing little fortune-cookie remixes of vaguely Zen vocabulary that had enough meaning to sound relevant. Mostly it was just recasting the question in algorithmically predicted vocabulary according to a set of interpretation prompts (“mirror” the question, use supportive or reflective language, etc.). Like all LLM generated text, it had a certain coherence based on the predictability of words ordered according to what word was most likely to come next.
Personally, I would take a real fortune cookie, a magic eight ball, or better still, a tarot deck, or (if I was a religiously affiliated person) an actual spiritual text and get way more useful narrative and symbology to interpret and think about. Whatever, the folks making this were having fun.
Then there was a rich fellow who had made money over the last few decades doing tech stuff. The first 75-80% of the presentation was about how cool he was. He threw the best parties, he hung out with the best tech innovators at key moments in recent history, etc. The actual presentation was supposed to be about using AI to hack agriculture and improve food systems. I’m sure the agroecologists in the room would be able to provide more substantive feedback on that stuff—my take on it was that somewhere within the wall of self-aggrandizement there is a sincere, if self-absorbed, effort to produce better human-centric efficiencies in the dehumanized, ecologically destructive industrial agriculture system. So I’ll give him that.
Truth be told, you don’t need LLMs to hack the industrial food system, you can feed the world sustainably with everything we already have, there’s just this bullshit economic model (“capitalism”) turning everything it touches into shit.
For me what the presentation was really about was the amazing ego and heedlessness of this fellow. The obsessive certainty, and the financial empowerment to enact that certainty. I do actually think he could make industrial food systems “better,” it’s just, are industrial food systems as they are comprised today what’s actually needed.
My breaking point was when—in keeping with his habit of personal digressions—he began talking about his dead friend. He was talking in general about how great LLMs are, and he had a dead friend who he loved, and he had many many hours of video of this person, so he decided to feed it all into an LLM to create an interactive video avatar so he could talk to his friend — of course, it wasn’t his friend, it was a remixable collection of digital impressions designed to give the illusion of interacting with a dead person.
At this point, I actually felt a little surge of physical nausea. I took a minute to let it pass, not quite believing what was happening, but recognizing that this was a clear exemplar of the monstrous maelstrom of heedless tech-baron self-absorption.
I don’t know if he got permission from his friend’s family, or even from his friend, to make an LLM memento mori, my impression was, he just did it because he could. Certainly there are folks who would be fine with this I suppose. (I’m not among them.) Whether this was happening on a context of posthumous or pre-planned consent, I dunno.
Then he activated the avatar, it chirpily announced that it wasn’t a “morning person” and hadn’t had coffee, and I felt that flash of nausea again, and had to flee the room. It made me want to actually puke.
As I walked back up to Mission Street an empty robot taxi passed in front of me. No passengers, and of course no driver. It was blank, innocuous, ready to receive whatever humanity put into it.
Exactly.
Leave a Reply