Embeddings decode real-time video of what a mouse saw from brain activity
flngr · x · 2026-09-12
A viral post claims researchers reconstructed actual frames of what a mouse saw in real time from its brain activity embeddings — not approximations of activity patterns, but decoded video. The author argues AI labs training on text embeddings are exploring only a small fraction of what embeddings could enable.
More from Research
- Burt Totaro on Terence Tao's blog: what math loses if AI brute-forces the Hodge conjecture — natanielruizg · 2026-09-12
- Michael Levin clarifies: paper does not conflate physical knowledge with mathematical truth — drmichaellevin · 2026-09-12
- LeVJEPA: video pretraining at 5.6-20.8x less compute, matching V-JEPA 2 — Cohere · 2026-09-12
- Continued pretraining vs RAG: an accuracy and performance comparison on Qwen 3.5 4B — funJS · 2026-09-12
- Researcher speculates spatial reasoning leap comes from Blender training data — yoavartzi · 2026-09-12
- tszzl pushes back on Fermi paper: 50-OOM lognormal abiogenesis prior under-justified — tszzl · 2026-09-12