Continuous Thought Machines - an architecture inspired by biological brains

=== ENTRY DETAILS ===

- QR code with entry information is available below, in the "Photos" section of this event page.
- Gate closes at 18:15 - no late entries.

======================

Continuous Thought Machines (CTM) are a novel neural network architecture that is pushing modeling closer to how real biological brains work. They have larger compute costs than the current best in class models (MLP / attention...), yet they are an interesting attempt to create innovative, general-purpose models for a wide range of tasks - building on learnings from neuroscience. The core idea of the model is to engage temporal dynamics on the foundational level of the architecture.

We'll walk through the 2 main ideas introduced by the authors (1: neuron-level time processing and 2: neural synchronization). Also, we'll review some of the tasks the model can be trained to solve.