The current AI conference model is facing a crisis. With the rapid growth of AI research, the number of papers being published has more than doubled over the past decade, with individual contributors publishing over 4.5 papers annually. This exponential growth is projected to continue, with some contributors expected to publish over one paper per month by the 2040s. But this growth comes at a cost – literally. The carbon footprint of AI conferences is staggering, with NeurIPS 2024’s travel emissions exceeding 8,254 tCO₂e, surpassing Vancouver’s daily citywide footprint.
But it’s not just the environment that’s suffering. The mental health toll of AI conferences is also a growing concern, with over 71% of Reddit threads on AI conferences being negative and 35% mentioning mental-health concerns. And to make matters worse, the AI research lifecycle is outpacing conference schedules, often rendering results outdated before presentation.
So, what’s the solution? It’s clear that the current model is unsustainable, and it’s time for a change. Whether it’s adopting more sustainable conference practices, finding new ways to present research, or rethinking the entire conference model, something needs to be done. The future of AI research depends on it.