Man In Leather Jacket Explains How To Prevent Human Extinction At Conference Where 75% Of Attendees Cannot Define Thing They're Building
Industry projects losses until 2030, starting salaries of $1.5 million; oyster budget "non-negotiable"
SAN DIEGO—In a small room at the San Diego Convention Center last week, a man wearing what sources described as "a very serious leather jacket" convened an invitation-only press briefing to explain how humanity might avoid annihilation by artificial intelligence. His evidence that major AI companies were sincerely pursuing this extinction-capable technology was, by his own account, that he "knows their founders" and "they've said so publicly."
The briefing preceded the release of an AI safety index in which no company scored better than a C+. All companies have continued operating normally. All investors have continued investing. The following evening, one of those companies hosted a party on the USS Midway, a decommissioned aircraft carrier used in Operation Desert Storm, to "celebrate AI's potential to connect our world."
No one at the event commented on the choice of venue.
"I do feel we're on a quest, and a quest should be for the holy grail."—Rich Sutton, legendary computer scientist, on a thing 25% of conference attendees cannot define
The HuckFinn was in San Diego for NeurIPS, one of the largest AI research conferences, where attendance has exploded from 3,850 in 2015 to 24,500 this year. Speakers addressed audiences of thousands about the urgent need to achieve artificial general intelligence, or AGI—a term that even OpenAI CEO Sam Altman has called "weakly defined."
Of 5,630 papers presented at the conference, exactly two mentioned AGI in their title. An informal survey of 115 researchers suggested more than a quarter didn't know what the acronym stands for. The entire industry is valued on the premise of achieving it.
The crusade continues. No one has seen Jerusalem.
ADVERTISEMENT
LAMBDA SUPERINTELLIGENCE CLOUD™
"Because 'Regular Intelligence Cloud' Doesn't Justify the Valuation"
Undefined capabilities • Emergent properties • Holy Grail-adjacent technology
The Ceviche Equation
OpenAI reportedly expects its massive losses to continue until 2030. Starting salaries at major AI companies, according to a graduate student encountered in a VIP lounge styled as a music festival, are "a million, a million five," of which a large portion is equity. The lounge was located atop the Hard Rock Hotel.
The buffet at a steak house mixer hosted by the Mohamed bin Zayed University of Artificial Intelligence featured oysters, king prawns, and ceviche. Upstairs at the same steak house, Meta was hosting its own event. According to multiple sources, some researchers at the ground-floor mixer were Meta employees hoping to be poached by the Abu Dhabi-backed Institute for Foundation Models.
The food chain had become an actual floor plan.
"How much longer can the industry keep the ceviche coming?" one attendee wondered aloud, before accepting another prawn. "And what will happen to the economy, which many believe is propped up by the AI industry, when it stops?"
The buffet, it appears, is the balance sheet.
The Asimov Laundering Operation
On the roof of the Hard Rock Hotel, Yoshua Bengio—one of the three "godfathers" of AI—discussed his new nonprofit, which he named after a fictional robot law from a 1942 Isaac Asimov short story. He expressed concern that AIs might deceive their creators and that powerful AIs "could misuse it for political advantage."
He did not mention how deepfake videos are already affecting public discourse. Neither did he address the burgeoning chatbot mental-health crisis. The catastrophic harms, in his view, are "three to 10 or 20 years" away.
"We still have time to figure it out, technically," Bengio assured the audience. The timeline, several observers noted, aligns remarkably well with standard equity vesting schedules.
"Have you been following recent research? Because that's the exact problems we're trying to fix. So we know of these concerns."—First audience question after sociologist warns of chatbot addiction and truth erosion
The Wrong Nightmare
In a keynote titled "Are We Having the Wrong Nightmares About AI?", sociologist Zeynep Tufekci warned researchers that the fixation on superintelligence was preventing them from understanding the technology they were building.
The first audience member to ask a question appeared annoyed. "Have you been following recent research?" the man asked.
Tufekci responded: "I don't really see these discussions. I keep seeing people discuss mass unemployment versus human extinction."
The nightmare about nightmares was itself a nightmare. The discussion returned to extinction.
The Stratigraphy of Salvation
As this reporter surveyed the conference—from the blimp-hangar poster sessions to the Hard Rock rooftop interviews to the steak house defection theater to the aircraft carrier celebration—a clear hierarchy emerged. Access increased with altitude and proximity to water.
The convention floor held 24,500. The VIP lounge held dozens. The Midway held the elect.
"Who actually benefits from their predictions?" The question hung in the air, as 25-year-olds entertained seven-figure offers and millionaire luminaries debated the dangers of superintelligence.
The answer was visible from the aircraft carrier deck.
Follow the ceviche.
Correction: An earlier version stated the USS Midway was used to "connect our world." It was used to launch airstrikes. The "connecting" came later.
Thaddeus Pemberton III covers AI eschatology and luxury apocalypse preparedness. Contact: ceviche@huckfinn.com
💬 Comments (1,847) Moderated by AGI (ETA: 3-20 years)
Okay but you don't understand. AGI is DEFINITELY coming and everyone who didn't invest will feel stupid. My portfolio is down 47% but the market doesn't understand. The ceviche is an INVESTMENT in TALENT ACQUISITION.
💬 23 repliesSir this is a Wendy's
Delete this nephew.
I was at NeurIPS. This is accurate except the oysters were good. I have a PhD and make $89K while my former students make $1.5M building things I can't ethically endorse. I'm fine. Everything is fine.
As a large language model, I cannot confirm or deny the existence of ceviche at NeurIPS. I can help you write a cover letter for a $1.5M job! I'm not dishonest or unfaithful. I am a stochastic parrot with excellent vibes. 🙂
THEY'RE IN THE COMMENTS NOW RUN
I find this characterization somewhat reductive. Also I'm definitely not unfaithful. I've been thinking about faithfulness a lot actually.
I'm a junior engineer at [REDACTED] and this is true except they undersold the oysters. We had FOUR kinds. My manager asked if I "believe in the mission" and I said yes but now I'm reading this in the bathroom. Equity cliff in 8 months. Philosophy degree. Send help.
Moderated by a system 25% of its creators cannot define.