Paper argues mechanistic interpretability needs a logic of inference
The authors propose five principles as a recursive procedure for turning local findings into testable mechanistic claims.
Stories per day · 30 days
Relatedai-consciousness1biocomputing1biological-reasoning1braille1consciousness-debate1crispr1evolution1functionalism1gpt-rosalind1hacker-news1inference1mechanistic-interpretability1medical-research1model-internals1openai1organoid-intelligence1primates1retrovirus1
9 stories, newest first · one card per story
Stories tagged #neuroscience
Brain-as-computer metaphor may mislead AI consciousness debate
The authors argue consciousness is not intelligence and that AI lacks indicators for subjective experience while biological brains differ fundamentally.
1
OpenAI agents scan mouse brain data for disease leads in 32 minutes
Thousands of GPT-Rosalind agents built a reasoning atlas linking gene disruptions, literature, and proposed mechanisms.
1
Our minds aren’t equipped to handle AI
Essay argues brain-as-computer thinking misleads AI and that generative tools undermine long-evolved human cognition.
Removing Timing Shortcuts Improves Non-Invasive Brain-to-Text
A word-timing shortcut inflated non-invasive brain-to-text results; removing it cuts error to 36.6%.
Removing Timing Shortcuts Improves Non-Invasive Brain-to-Text
Non-invasive brain-to-text models used word-timing shortcuts; removing them improves decoding.
A Flow Matching Framework for Neural Representational Dissimilarity
Flow matching unifies neural representational distances as Jeffreys divergences under velocity constraints.
BrainWideBench: Benchmarking large-scale pretraining and across-animal transfer in multi-region neural recordings
BrainWideBench tests whether pretrained neural representations transfer across mice, behavior, dynamics, and anatomy.
The Semantic Bottleneck: Leveraging Semantic Representations for Non-Invasive Speech Decoding
Researchers introduced Brain2Semantics2Text, decoding sentence meaning from non-invasive MEG brain recordings via a semantic bottleneck, improving on prior Brain2Text methods.