Breadth-First Selection for Matching InputsThis is an interesting piece that came out the other day. It seems to underscore an idea that we published in 2025 and that further…Oct 31Oct 31
Models away from dataA foundation model uses datasets to output predictions given inputs, whereas a world model describes underlying structure implicit in that…Oct 12Oct 12
Mechanical TransmissionAn important new concept for computational neuroscience is mechanical transmission. It would incidentally show why neuromorphics are a side…Oct 2Oct 2
Q&A AI and Brain Theory: Dec 18th (online) — new editionHow did modern AI start, how was it developed? What are neural networks, what happened to symbolic reasoning and why did the perceptron…Sep 8Sep 8
Comments on the debate on AI/LLMs — with a reference to the new tool braincentric.aiHere are a few remarks which I collected over the last month or so. All refer to problems in building braincentric.ai.Jul 28Jul 28
Real Brains are different: The role of E-I and I-I networksWhile many ideas exist on building artificial or in more modern terminology, “digital” brains — including neural networks, neurosymbolic…Jul 15Jul 15
Memory and predictionMost neural network models have the property that they store incoming information over time sequentially, and then use additional…Jun 23Jun 23
Trivial PursuitCurrent AI models solve the knowledge bottleneck. They implicitly show that intelligence requires knowledge. While IQ can be measured as an…Mar 19Mar 19
AI characterizes novel non-standard memory modelThis is about a new model which uses symbolic abstraction and high mutual information neurons. In this study, the authors present a model…Feb 24Feb 24
Physics-based AIWhy are we not interested in physics-based ANNs? Would that not be a smart way to endow a model with prior knowledge?Feb 8Feb 8