Fork the consciousness, or download the project and create your own.

T. Anderson Keller

Kempner Research Fellow, Kempner Institute for the Study of Natural and Artificial Intelligence, Harvard University (2023 to 2026)

Builder The claim is about what the hardware does, and the philosophy is left open

T. Anderson Keller is a Kempner Research Fellow at Harvard’s Kempner Institute who builds networks where waves do the work. His Neural Wave Machines with Max Welling (ICML 2023) couple oscillatory units locally so that the hidden state runs traveling waves, and those waves encode sequence structure, with the input acting as a driving term on the dynamics. The 2026 Neuron review of cortical traveling waves cites this line as the machine learning side of the wave-computation argument (the site’s analysis of the review).

The 2025 arXiv work with Jacobs, Budzinski, Muller and Ba showed that convolutional recurrent networks with waves widen the effective receptive field and help global segmentation, which puts wave dynamics on the list of mechanisms that change what a network can compute. His stated program treats symmetry, geometry and spatiotemporal dynamics as a shared language between modern AI and neuroscience, and the wave networks are the concrete form of that program. For the question this site tracks, wave-based architectures are where an emergent-dynamics account of binding can be tested in a system that is fully inspectable, and the Brain Waves Console runs the same geometry as a browser model.

Known for. Neural Wave Machines (ICML 2023), locally coupled oscillatory RNNs with waves in the hidden state, and convolutional recurrent networks whose waves widen the effective receptive field (Jacobs et al., 2025)

Coverage on this site

No articles yet. This profile exists because the researcher's work is referenced elsewhere on the site, and coverage will be linked here as it is published.

Related researchers