EEG vs implanted BCIs: researchers clash over physical limits and AI decoding potential
burny_tech · x · 2026-11-10
A technical debate over non-invasive brain-computer interfaces:
- JTRobinsonLab's critique: EEG cannot measure what implanted BCIs capture. Spatial averaging means signals above 70 Hz (research-grade) or 20 Hz (consumer) fall below the noise floor at the scalp, while implanted electrodes resolve up to 200 Hz (epidural), 600 Hz (ECoG), and 2 kHz (penetrating). "Inferring high-frequency brain signals from EEG is like guessing lyrics from the bassline."
- kanair's rebuttal: while EEG faces fundamental physical measurement limits, we are nowhere near the limits of what AI can decode from it — scaling EEG decoding with AI still has far more potential to unlock.
The core question: can algorithmic and compute progress partially overcome the physical bottleneck of signal acquisition?
More from AGI Musings
- Andrew Gelman asks: can peer review survive unrestricted AI use by reviewers? — JessicaHullman · 2026-11-10
- McCulloch's one question: 'What is a number that a man may know it?' — forestmars · 2026-11-10
- We're in a sweet spot where AI detectors work but slop producers don't know it yet — paulnovosad · 2026-11-10
- "Phones and AI killed music": boredom and time made great art, and both are gone — threepointone · 2026-11-10
- Martin Casado: people now use "Bitter Lesson" to mean "a frontier lab spent lots of money" — ZeYanjie · 2026-11-10
- Professors' new vibe toward AI: 'I feel extremely melancholy,' scholar says — _akpiper · 2026-11-10