Researchers debate: learning from unlabeled video will beat text-only scaling, no recipe yet
abursuc · x · 2026-08-23
A spat over whether self-supervised learning (DINO line) needs text. The quoted Meta-side view: text exists, so why not use it—learning representations without text is unnecessarily hard.
Yuki Asano pushes back: text gets you only so far (perhaps far enough for many applications and commercial value), but learning from unlabeled data, especially video, will ultimately be best—huge quantities plus physics, geometry, occlusions. He cautions there is no winning recipe yet; it will take proper research.
More from Research
- Wan_2.2_I2V_14B-Clean trending on Hugging Face — wank3r · 2026-08-23
- New Perspective: Scientific papers need an uncertainty layer for AI — anshulkundaje · 2026-08-23
- Tutorial: How to quantize a Mixture of Experts (MoE) model — cephaloform · 2026-08-23
- Badtheorylabs shares near-lossless 2-bit model quantization method — cephaloform · 2026-08-23
- New KnotBench Benchmark: Pay Topologists $10M Each to Solve AGI — suchenzang · 2026-08-23
- Geoffrey Irving on the Grounding Problem in Character Training and Alignment — geoffreyirving · 2026-08-23