This isn’t your dusty old academic get-together. The numbers alone are staggering. Back in 2015, NeurIPS drew around 3,850 people. This year? A colossal 24,500 descended on the city, a testament to an industry that has exploded into the mainstream. But beneath the surface of poster sessions and technical papers, a bizarre and frankly worrying disconnect is playing out, one that says everything about the state of AI today.
From Niche to Notorious: The Conference Gold Rush
So, why the ten-year surge from a niche workshop to a stadium-sized spectacle? The answer is simple: money and talent. The objective of these conferences has morphed. Whilst collaboration and innovation are still the official currency, the real trade is in people. The conference floor is now the most expensive recruitment fair on the planet.
This is where the sheer tech industry excess comes into focus. You want to hire the best minds in AI? You don’t just offer a competitive salary anymore. You throw a party on a decommissioned aircraft carrier. That’s not hyperbole; Cohere, a major AI player, actually hosted an event on the USS Midway. Not to be outdone, Abu Dhabi’s Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) took over a steakhouse for a lavish buffet. It’s a feeding frenzy, with companies dangling starting packages of, as one insider put it, “a million, a million five” (mostly in equity, of course) to lure PhDs away from their rivals.
This NeurIPS extravagance is like watching the dot-com boom all over again, but with higher maths. The flashiness serves a purpose: it signals power, momentum, and inevitability. It creates a reality distortion field where the primary goal isn’t just to advance science, but to win the war for talent before the other guy does.
Armageddon on the Agenda, Crickets in the Labs
And what is everyone talking about amidst the canapés and cocktails? The end of the world, naturally. The dominant conversation, amplified by luminaries like Max Tegmark, is about the existential threat of ‘artificial general intelligence,’ or AGI. This is the hypothetical point where an AI becomes so smart it could, as Tegmark warns, “precipitate the end of human life.” You hear whispers of it everywhere, a constant drumbeat of AGI speculation that casts the entire industry in a heroic, high-stakes drama.
Here’s the rub, though. It’s almost entirely theatre. As detailed in a brilliant piece by The Atlantic, when you look at the actual research being presented, the story crumbles. Out of 5,630 peer-reviewed papers at NeurIPS, how many do you think had “AGI” in their title? Two. Just two.
The disconnect is staggering. A poll of researchers found that over a quarter of them didn’t even have a clear definition for AGI. Even OpenAI’s Sam Altman, whose company narrative is built on this grand quest, admits AGI is a “weakly defined term.” So what are we doing here? We have an entire industry fixated on a sci-fi bogeyman that its own researchers are barely working on. It’s like the world’s top chefs gathering to discuss how to cook on Mars, whilst ignoring the fact that everyone’s kitchens are on fire back on Earth.
The Real Risks Nobody Wants to Talk About
This glaring contradiction was thrown into sharp relief by keynote speaker Zeynep Tufekci, one of the few voices of reason in the room. She cut through the noise with a simple, devastating question. Why, she asked, is everyone obsessed with mass unemployment or human extinction scenarios that are “three to 10 or 20 years away,” as Yoshua Bengio puts it, whilst ignoring the clear and present dangers?
Where are the packed sessions on AI-driven addiction pathologies? The urgent debates on the tidal wave of misinformation that’s already poisoning our information ecosystem? The serious reckoning with the copyright and labour issues that are hollowing out creative industries? Tufekci noted, “I don’t really see these discussions. I keep seeing people discuss mass unemployment versus human extinction.”
This isn’t just an academic oversight; it’s a colossal failure of conference ethics. The community is so enthralled by its own myth-making—this grand, “holy grail” quest as Rich Sutton calls it—that it’s willfully ignoring the damage its current, far less intelligent, systems are already causing.
Follow the Funding: The AGI Narrative is Big Business
So why does the AGI narrative persist? Because it’s incredibly profitable. Ben Thompson of Stratechery would call this strategic positioning. The quest for AGI is the story that unlocks billions in AI research funding.
Companies like OpenAI, Anthropic, and xAI aren’t just building technology; they are selling a narrative of destiny. They need the existential stakes to be high to justify their astronomical valuations and cash-burn rates (OpenAI, for instance, reportedly expects to remain unprofitable until 2030). The focus on AGI frames them not as mere software companies, but as humanity’s vanguards, either leading us to utopia or saving us from oblivion. It’s a marketing masterstroke.
This shapes the entire ecosystem. Venture capital flows towards the grandest vision, and the grandest vision is AGI. Consequently, research and talent are pulled into its orbit, leaving more immediate, less glamorous problems under-resourced and under-discussed. The NeurIPS extravagance isn’t a bug; it’s a feature of this narrative-driven funding model. You have to look like you’re winning the future to get the money to build it.
The entire AI conference culture has become a self-reinforcing loop. The AGI narrative fuels the funding, the funding pays for the extravagant recruitment, and the extravagant recruitment secures the talent needed to keep the narrative alive. The actual science? It’s becoming a side-show.
The question for the AI community is whether it can break this cycle. Can it pivot from navel-gazing about hypothetical doomsdays to addressing the real-world messes it’s already making? Or has the industry become so captivated by its own story that it can no longer see the world outside the convention hall?
What do you think? Is the focus on AGI a dangerous distraction, or a necessary long-term vision? Let me know your thoughts below.
