DeepMind, Harvard, Stanford Argue Visual Intelligence May Be a Path to AGI
rohanpaul_ai · x · 2026-09-12
A white paper by 21 researchers from Google DeepMind, Harvard, Stanford, Oxford and other labs argues that vision-centered AI could offer a path to AGI, termed "Visual General Intelligence" (VGI).
- The analogy: in language, the GPT series showed transfer to unseen tasks via Transformer architecture, autoregressive training on web-scale text, and aggressive scaling. The paper asks what capabilities could emerge from images, video, and geometry.
- Critique of the status quo: most multimodal AI treats vision as input to a language model; the paper wants vision to do more of the thinking itself — learning directly from visual data to build world models, remember changes, predict outcomes, and act.
- Rather than a single definition, the paper maps the principles CV should pursue in the AGI era, visual modalities, benchmarks, learning paradigms, and vision's relationship with language and other modalities.
Related event: DeepMind, Harvard and Stanford Say Visual AI May Be the Path to AGI(2 posts)→
More from Research
- Rachel Thomas points to classic paper on articulating the value of not building tech — math_rachel · 2026-10-28
- Meta's Omnilingual MT covers 1,600 languages, 1B–8B models rival 70B LLMs — kalyan_kpl · 2026-10-28
- Supercharging LLM classification with logprobs: soft classifiers you can threshold-tune — JnBrymn · 2026-10-28
- cryoET Data Portal releases three datasets benchmarking crossed laser phase plate — ZhongingAlong · 2026-10-28
- Blind Robot Hand Spins Baoding Balls 225 Times With Only Touch — Scobleizer · 2026-10-28
- Rollback: World Models Rewind Robot Decisions to Explore Counterfactual Actions — agibot-world · 2026-10-28