ECCV debates: are explicit 3D representations obsolete as video generative models take over?
qixing_huang · x · 2026-09-12
At ECCV, the author observed serious workshop debates in 3D vision over whether explicit 3D representations are still needed for tasks like novel-view synthesis and robotic action. An increasingly influential view holds that video generative models are now powerful enough to dispense with explicit 3D entirely.
Key points:
- He largely agrees for benchmark/practical purposes: with limited explicit 3D training data, approaches leveraging far larger 2D/video data sources currently win.
- But echoing parallel debates among mathematicians, he argues science is not just about metrics — explicit structure may carry long-term scientific value beyond leaderboard performance.
The post frames a broader tension across AI fields: as scaling + massive data beats structured approaches, how much explicit representation should research preserve?
More from AGI Musings
- Fly brain to neural net: translating synapse counts into LLM parameter equivalents — alvelda · 2026-09-12
- The paranoid style in AI safety: an adversarial frame can create the adversary you fear — sebkrier · 2026-09-12
- How Far Are We From the "ChatGPT Moment" for Embodied Intelligence? — globalsouthworld · 2026-09-12
- What If P=NP Is Solved? Crypto Breaks, Breakthroughs Unlocked — and Emad Mostaque Would 'Lie Down' — CedricMakes · 2026-09-12
- Nat Lambert: frontier lab grind suggests we're approaching RSI, not in it — natolambert · 2026-09-12
- Hugo Bowne: imagine how mathematicians feel about AI slop — hugobowne · 2026-09-12