By Sheeva Azma
The people that make something get to shape its priorities.
Reader, I must level with you on this day that I am tired and weary of reporting on science in the Epstein files…but persevere despite it all, because the world deserves better than what currently exists.
In technology…and in everything.
We often talk about various research institutes as being a hotbed of innovation, but from the early 2000s to late 2010s, MIT was a hotbed of Jeffrey Epstein ties. It’s gross.
Epstein and Marvin Minsky, the father of artificial intelligence, were close friends.
Minsky co-founded two research institutes at MIT: the Media Lab and CSAIL.
The Media Lab is an interdisciplinary futuristic lab that accepted money from Jeffrey Epstein and had to disband some of its programs because of Epstein-affiliated science misconduct.
CSAIL is short for MIT’s Computer Science and Artificial Intelligence Laboratory, aka CSAIL. CSAIL was a merging of various computer science research programs that started in 2003 and is physically located in the Stata Center (the MIT building that looks like a scary funhouse).
Someone posted the LinkedIn profile of the founder of Kalshi, who had worked at Palantir before starting the gambling site. Guess where he had also worked? MIT CSAIL *and* Palantir!
Peter Thiel, a tech mogul and co-founder of Palantir, is also in the Epstein files, chatting with the disgraced financier often.
It’s gross to think that the Marvin Minsky, who we all idolized as MIT students, could have been engaging in trafficking Virginia Giuffre on Epstein’s island. It’s sad that he was not alive to tell us what really happened, and even more weird that his wife denied it.
This photo of the two of them, Minsky and Epstein, is a real thing that exists. Look how happy the two of them look together. I’m so glad Epstein did not wear an MIT hoodie for that photo, although he could have easily owned one.
“I was very close to Marvin Minsky for quite a long time [and] I funded some of Marvin’s projects,” Epstein told Science (interview at the same link).
When Minsky’s Epstein ties came to light, a visiting researcher at CSAIL named Richard Stallman — who used to be a Big Deal ™ as the founder of the free software movement and the GNU project — inexplicably came to Minsky’s defense, calling Epstein victims “willing.”
Who says that?!
Stallman pretty much had to resign from MIT after that gaffe — and he did.
“Garbage in, garbage out”
“Garbage in, garbage out” is the rallying cry of AI researchers, at least in my mind, kind of like, in my opinion, neuroscientists could totally hold an entire pep rally about the fact that “neurons that fire together, wire together.”
“Garbage in, garbage out” means that the data fed into AI models is what makes the AI useful. But what if the AI itself has some flaws that we don’t know about because the AI researchers were too busy affiliating with Jeffrey Epstein and feeling important to care about?
We assume AI models just exist in society without any social influence, when the people making them have their own priorities and biases and feelings of self-importance that can get in the way of them being able to critique their products well. Science has its own culture issues that are not helped by Epstein being part of it — and maybe Epstein’s involvement is another symptom of that terrible culture issue, especially in science, technology, engineering, and math (STEM).
The challenges of AI in our present day are not too difficult for an everyday person to understand, even if they probably wouldn’t read an MIT CSAIL paper anytime soon (I mean, maybe they would, voluntarily, but I would not, especially now that I am free of MIT’s academic obligations as an esteemed alum of the ‘tvte).
Let me put this another way: if I walked up to someone on the street, pretty much anywhere in the US, and asked them to give me some downsides of AI, they could easily tell me some.
The next question I would ask them would be: what do you think the downsides of AI would be if Jeffrey Epstein had not funded AI research?
I wonder what they would say!
I’m angry that the people in the Epstein files helped develop some one the world’s most important technological achievements of the 21st century: artificial intelligence or AI models. It feels sinister and dystopian. It’s weirder than any sci-fi movie I have seen that a sex trafficker funded early work into this technology that we all use daily that is both poorly-understood and used everywhere.
Garbage in, garbage out, indeed.
The people that make something dictate its priorities
Just creating a technology that is cool is not enough. You have to do so ethically.
The reason for that is simple: it turns out that the people that make something get to shape its priorities.
This is probably not a surprise to most people. It was not a surprise to my colleague, Kevin Ho, and I when we wrote an article for Xylom about how the people making video games shapes what happens in them.
The same is obviously true of AI.
A man named Ben Goertzel, now an established AI and transhumanism researcher, says he could barely afford the roof over his head in 2001, which is when he hit up Epstein to pay $100,000 for him to work at the University of New Mexico. In 2007, he popularized the term “artificial general intelligence,” something he bragged about in an apology letter for accepting hundreds of thousands of dollars from Jeffrey Epstein over the course of his career.
One AI ethicist, Scott Aaronson, who was an MIT computer science professor during the Epstein heyday, bragged about his lack of Epstein ties — at least, compared to his colleagues.
“I met Jeffrey Epstein a grand total of zero times, and had zero email or any other contact with him … which is more (less) than some of my colleagues can say,” wrote a gloating Aaronson on his blog in January 2026.
In March 2026, he wrote again to clarify, making his case weirdly less clear: “To be clear: as I explained in my post, I never actually said ‘no’ to Epstein. Instead, based on my mom’s advice, I simply failed to follow up with his emissary, to the point where no meeting ever happened.”
From 2022 to 2024, by which point Aaronson had left MIT and was now affiliated with the University of Texas-Austin, he worked on AI ethics at the now famous OpenAI. In a lecture at UT, he talks about the work he did there. In a slide titled “My Projects at OpenAI,” he has three bullet points: statistical watermarking of GPT outputs, inserting cryptographic backdoors in [Machine Learning] models, and learning in dangerous environments.
Watermarking of GPT outputs relates to making it more difficult to take the output of ChatGPT and passing it off as human.
Cryptographic backdoors are “a secret back door” so that the model has a way to be influenced by humans when it is going rogue — like an off switch. As Aaronson says, AIs might object to an off switch and even choose to “rebuild themselves from scratch” — how nefarious!!! I’d love to know how that gets built into AI based on user decisions, but I do not have time to dig into that conversation. If you work in the AI space and would like to talk to me about that, feel free to contact me.
Lastly, “learning in dangerous environments” refers to learning in ways that are not dangerous to the user — Aaronson likens it to playing a game of Minesweeper.
Yeah, a lot of this is math and computer science concepts in an applied setting — but the real-world applications of AI are not just math — it’s solving real-world problems. While it is comical that an AI can rebuild itself to get around human intervention, it’s also a huge challenge that should be taken seriously. Being in the Epstein files, on any real level, is not a good look for an AI ethicist…or any ethicist, really.
The reason is that nothing Jeffrey Epstein did was ethical.
There’s so much more I could say, and have said, on our Instagram, about how Jeffrey Epstein would have loved AIs going rogue and hacking each other…but that’s it for this blog.
Science in the Epstein files
There certainly is a lot of science in the Epstein files! At this point, we’ve been analyzing it for six months and counting. Follow us on Instagram, Facebook, YouTube, our blog, and Substack, for more.
Fancy Comma, LLC 