Sort by: Top
Feed post
Senior Lecturer, CIBM Center for Biomedical Imaging & Department of Clinical Neuroscience, University of Geneva
2h •
The hype around EEG foundation models just got hit with a cold
The 260,000-hour benchmark introduced in NeuroAtlas provides an interesting reality-check:
EEG-specific pretraining shows no consistent advantage over generic time-series models.
Time to step out of the hype and refocus on mathematical rigor, proper data preprocessing, and true clinical validation.
Paper: https://lnkd.in/eehwruZ8
Neurofeedback paper too
“We analysed every resting-state psychedelic dataset we could find. None of that default mode network stuff held up under 250 subjects. That story needs to go to hell.”
“This whole default mode network narrative became huge because Michael Pollan wrote about it. When we put the data together, the decrease in connectivity under psychedelics was an extremely unreliable finding.”
This might be the most sceptical conversation about psychedelics I’ve ever had on the show, cognitive neuroscientist Manoj Doss is live
. It got picante
