A personal meditation on how AI is reshaping the community that built it
"In the real world, when rewarded for doing well on a homework assignment, a student might copy another student to get the right answers, rather than learning the material—and thus exploit a loophole in the task specification."
Google DeepMind
In the AI community, AI itself now generates papers at unprecedented rates and of dubious quality, overwhelming the system. Students are incentivized to submit rather than innovate.
Reward hacking.
In my film, I visit the largest AI conference in the world to uncover how AI is reshaping the community that created it, for good and for bad.
I also look at how my own history of addiction informs the addictive way I use AI now, and how that relates to the pressures and incentives the community feels.
Along the way, we'll cover cutting-edge research made possible by AI, and the opposite, AI slop.
For anyone interested in AI.
Everyone is talking about AI. Almost nothing personal has been made about what it's doing to the people who build it.
NeurIPS received 1,678 papers in 2014. In 2025 it received 21,575.
Careers, visas, and jobs hang on getting in.
The incentives were broken before AI. AI is speeding up the race to the bottom.
1. To better understand myself. I feel like I have an addiction to AI, and I'm trying to understand where that came from.
2. To help people understand how AI is changing the academic community, and work in general.
3. To fuse informational documentary and personal narrative, so that fans of either will watch.
My goal is to spend an entire day with each person I recruit, and to talk to as many others as I can.
Friends, family, and people I meet. Two-part interviews, shot in places that matter to the person on camera.
Part one is their relationship to AI.
Part two is what they think of me, and of this film. For that part I leave the room, and they record without me.
The conference and back home, moving toward one place that is both informational and personal. This is a tentative outline.
Back home. Friends, family, and people I meet, on what they think of AI.
Introduction to NeurIPS. Straightforward and fast-paced.
People who know me, on what they think of this.
Mid-conference. Following people as they do their thing.
Deeper into my past.
Departure. Hopefully, with some insight.
I'm a PhD candidate at Duke, working at the intersection of machine learning and healthcare. I'm an ML4H author and reviewer, and I'll be presenting at NeurIPS myself. I'm not visiting this world. I live in it.
I've never made a film. I've spent years designing and running interviews, which is most of what this film is.
I'm an engineer, and I work with AI. The strategy is to record lots of footage, and save it in the edit.
Claudia, the AI production manager I built, runs on my phone and laptop.
Shoot, log, cut, repeat. The edit starts during the shoot.
Everyone at the conference is doing too much, too fast, with AI. So am I.
Claudia keeps track of the logistics. In the field, she advises me on what footage would be good to get.
The team is me. I'm directing, shooting, recording sound, and editing.
Brian Lerner
Director, camera, sound, edit.
Maura Matney Lerner
Location scout.
✦ Claudia
AI production manager. Logistics, and advice in the field.
Nov 1 – Dec 1. At home in Durham, with family at Thanksgiving.
Dec 2 – 16. Australia, for ML4H, NeurIPS, and the Reef.
Mid-December on. Post-production.
One person, at most two cameras at a time.
Interviews. SL3 on a tripod, Luna Ultra for additional angles.
Run and gun. Luna Ultra is all I need.
I already have what I need to shoot. Duke lent most of the gear.
Not covered yet are AI costs to run Claudia, repairs on the road, and incidentals.
If you'd like to help, with a donation, a loan of gear, a letter, or an introduction, I'd love to hear from you.
Brian Lerner
brianelerner@gmail.com
973-986-3318