🚀 Groundbreaking Critique at CVPR 2025: Saining Xie Warns AI Research Is Stuck in a “Finite Game” 🎮
-
In a powerful keynote at CVPR 2025, NYU Assistant Professor and Young Researcher Award recipient Saining Xie delivered a compelling critique of today’s AI research landscape, drawing intense attention from both industry and academia. 🧑
Finite vs. Infinite Games ♾️
Xie kicked things off by contrasting two paradigms:
- Finite games : short-term, win or lose competitions
️ - Infinite games : open-ended, collaborative exploration
He argued that AI research has drifted dangerously toward the finite model, marked by:
Intense short-term competition
Superficial innovation
Crushing pressure on researchers, especially early career scientists
Personal Experience: From Rejection to Revolution
️
“My most influential papers were first rejected for ‘lack of novelty,’ yet they reshaped generative AI.” Saining Xie
- DiT (Diffusion Transformer) and SiT were initially rejected by CVPR
- Later, DiT became a core of OpenAI’s Sora
️, and SiT powered Stable Diffusion 3 
These turnarounds highlight the limits of current review metrics.
Call for an Infinite Game Mindset 🧭
Xie urged the community to embrace:
- Sustained exploration
️ - Collaborative openness
- Resilience & antifragility 🪄
- Lifelong learning
Setbacks should fuel deeper discovery, and education must go beyond ticking academic boxes.
Why It Matters
- Encourages healthier, sustainable research culture
- Shifts focus from paper counts to long-term impact
- Sparks dialogue on reforming incentives
Final Rallying Cry
“Let’s rethink our incentives, prioritize lasting impact, and play the infinite game together.” Saining Xie
Xie’s speech has already become a landmark moment, urging the AI community to re-evaluate and redefine its values for the years ahead.
Research as an infinith game, slides: https://t.co/rsNleF610k - Finite games : short-term, win or lose competitions
