DeepMind's Andrew Lampinen: why would only non-language input grant understanding?
AndrewLampinen · x · 2026-09-28
DeepMind researcher Andrew Lampinen pushes back on the claim that only non-linguistic inputs (images, sensor readings) can grant a model genuine understanding.
He questions why non-language things get a 'magical meaning-ness' that can't even in principle be present in language, noting that much of what an image or sensor reading conveys could just as well be said in a chat.
More from AGI Musings
- Indie dev on AI: your game gets cloned within days and slop buries originals — ring_hyacinth · 2026-09-28
- AI Personal Assistants May Quietly Replace Much of Normal Human Interaction — GarrisonLovely · 2026-09-28
- Open AI labs aren't burning money — they're dismantling the 'intelligence tax' — thatgodzillaguy · 2026-09-28
- Paras Chopra: AI's Pull Comes From Looming Human Economic Irrelevance — paraschopra · 2026-09-28
- Jan Kulveit: Stop Overcorrecting Toward 'Power-Seeker' Readings of Frontier Models — jankulveit · 2026-09-28
- Silicon Valley Parenting: Kids' Best Years Consumed by Sports and Accelerated Classes — vaibhavbetter · 2026-09-28