29 Nov 2025

I’m tired of hearing about AI. I’m tired of “tech” podcasts like Hard Fork making 30-minute segments every time Sam Altman posts something cryptic.¹ I hear nails on a chalkboard every time someone says “AI is being used to supercharge advancements in pharmaceuticals!” as if Alphafold has anything to do with Large Language Models. I love KataGo: a Go engine made possible by Deepmind’s AlphaGo. I hate that Deepmind’s AlphaFold won the Nobel in Chemistry. A lot of people in my department feel the same. I hate that I hate that AlphaFold won the Nobel: I like the people who worked on it, I like what they do with it. I still think Demis Hassabis is cool…

I miss when people who talked about AI actually cared, when they were genuine, when it was an academic interest, when it was gatekept by those with formal educations, when I hadn’t yet received a formal education. I know people who care still exist. Most programmers I encounter have nuanced opinions on the use of LLMs to write code. I think their voices are drowned out by the accelerationists, proompters, people who make fun of proompters, those who fear AI will destroy humanity, the next model, the bubble.

I miss being excited about SIGGRAPH papers that used generative AI.² I loved EbSynth when it came out.³ I would have old laptops chugging away on videos overnight. I miss training my own GAN on photos I took, when the idea of training models locally hadn’t been collectively overwritten. Last summer, I downloaded the weights to a photo-to-\(\LaTeX\) model from Hugging Face, but I could never get it to run entirely offline despite my best efforts. It had to connect to the internet to run on my own computer. It felt like I had lost something.

I wonder what it’s like to the vandals of Friend advertisements in the NYC subway,⁴ thinking you were making a stance about humanity and the importance of social connections. To see them, not you, get time on TV, to see yourself become their pawn. They spent a million dollars to put the ads underground, and you could have left it there. Worst is, I doubt any real social connections were formed from a mutual hatred of Friend.

I wonder what it’s like to be one of the girls tormented with non-consensual p✨rn generated from photos they took.⁵ The real bits representing you are painted over, corrupted into a parody of the female body. These models were trained by people. They turn innocuous photos into blackmail, normal girls into traumatized girls, fantasies into nightmares. It doesn’t matter if it’s not realistic. While you’re lecturing about the importance of being able to determine what’s real online, there are boys making CSAM in the back of class.

I wonder what it’s like to be one of the 13 or 14 year-old boys who faced felony charges. Arrest records say they used “an artificial intelligence application” to make fake nudes of their peers at school.⁶ I don’t have to wonder what it’s like to grow up around men misogynistic enough they’re tolerated; men who found that line before the internet democratized this flavor of sexual harassment. It’s a fact of life on the internet that you will be exposed to explicit pop-ups, banner ads, and redirects. It becomes noise, it doesn’t matter if it’s real. That is, unless you’re in Florida, where it’s only a misdemeanor to share real revenge-porn. “It is really strange to me that you impose heftier penalties for fake nude photos than for real ones,” said lawyer Mary Anne Franks.

I’ve used these models. I wanted to see what it would do to my body. I wanted to see it confused without the input of a prepubescent girl’s mirror selfie. I noticed the model weights setting in a Github repository, I saw websites who would even run it for you, free apps. I saw it put a pair of fully developed breasts over my chest. I guess photos of shirtless men weren’t in the training data.

I don’t hate normies. I hate when people get overpromised and underdelivered. Somehow, it never tanks a company when they release a half-baked product only to add the features promised months⁷, a year,⁸ or even a decade⁹ later; it’s normalized. People still pay their subscription.¹⁰ I hate how absurd it is. I hate how acknowledging how absurd it is is part of the game. Hate is intoxicating, it gives them eyeballs, it takes oxygen from the room.

Recently, I’ve been having breathing difficulties. Occasionally, without reason, I get the feeling that I’m not getting enough oxygen, but everything is fine. I don’t wheeze or cough. The only evidence is that I feel persistently bad for an hour. I can’t find help online, I don’t care enough to sift through the slop.

Every time it happens, I can’t help but think about the xAI data center that spawned less than 15 miles away from where I sleep, my job, all the stores I go to, where I’m writing this. “Colossus”: claimed to be the worlds largest AI supercomputer,¹¹ a.k.a. “MechaHitler” is thought to be powered partially by gas turbines emitting 3000-5000kg of NOx every day.¹² In undergrad, I modelled secondary organic aerosols formed, in part, by compounds like NOx. I told a supercomputer to predict what would happen in different environments. Now, a much larger supercomputer is being used to predict tokens that best own the libs, and those aerosols are in the air for real this time.

TLDR: Just watch this Internet of Bugs’ video on fears about AI. It’s very well spoken, concise, and worth your time if you every worry about these things.

Epilogue #

Sam Altman says “we know how to build AGI” and if they don’t, his company still gets money and attention. Sam Altman says AI will probably kill us all, and he gets an article on CNN. Sam Altman posts “i love summer in the garden” and gets articles in Tech Radar, Business Insider, Mashable, Medium. While looking up these quotes, I saw this “AGI by 2025” video thumbnail (2025 isn’t over yet, have you even used the new models?) and it somehow doesn’t feel any more outlandish to proclaim now than it did in 2024. Whether AI automates all work away, kills everyone, or makes us go crazy, it still claims the title of “disruptive technology.” Why do we let it disrupt our lives and our thoughts?

I’m tired of these hypotheticals. AI isn’t going to destroy humanity. I’ll direct you to a video that makes this point better than I can. Many other fears only serve AI companies in that they distract from real problems that exist for real people right now. If you’ve made it this far, you probably already know the state of LLMs. You may believe (as I do) that the transformer architecture has reached its limit. What good does it do to stay plugged in all the time? The problems we are facing now are real, tangible, conceivable, and addressable. All I’m suggesting is that we direct our attention away from the big problems, many of which are prompted by Big LLM, and focus on the ones that are real and have obvious solutions.

P.S. It’s insane that there are still people that think machines are physically incapable of being biased or influenced by humans. That needs to stop. Seriously.