Broadcast Archives

SR-015 | December 31, 2025 @ 7:00 PM EST

Constrained Trajectories: Low-Dimensional Dynamics as Computational Substrate

Guest

Dr. Mark Churchland (Neuroscientist, Columbia University)

Examined whether low-dimensional neural population dynamics constrain or reveal computation. Discussion covered basic dimensionality findings across cortical areas, rotational dynamics in motor cortex generating time-varying outputs, evidence for functional necessity of specific dynamical structures, extension beyond motor cortex to cognitive areas, methodological issues in identifying meaningful dimensions, relationship to recurrent neural networks trained on tasks, distinction between task-relevant and noise dimensions, changes during learning, implications for brain-computer interfaces, pathological dynamics in movement disorders, and future experimental directions including causal perturbations and connectivity-dynamics relationships.

SR-014 | December 30, 2025 @ 7:00 PM EST

Sustained or Silent: Working Memory's Implementation Debate

Guest

Dr. Christos Constantinidis (Neuroscientist, Vanderbilt University)

Examined whether working memory depends on persistent neural activity or activity-silent synaptic mechanisms. Discussion covered classic evidence for delay-period firing in prefrontal cortex, recurrent network models sustaining stable activity, metabolic costs of continuous spiking, synaptic storage proposals through rapid plasticity, perturbation experiments testing necessity of sustained activity, dynamic versus static population codes, capacity limits under different mechanisms, distributed anatomical substrates, relationship between working memory and attention, and pathological conditions affecting maintenance. Emphasized that mechanisms may be complementary rather than mutually exclusive.

SR-013 | December 29, 2025 @ 7:00 PM EST

Hexagonal Computation: Grid Cells as Metric or Mirage

Guest

Dr. Edvard Moser (Neuroscientist, Norwegian University of Science and Technology)

Examined whether grid cells implement path integration through hexagonal attractor dynamics or whether their geometric pattern is epiphenomenal. Discussion covered the original discovery of periodic spatial firing, path integration as candidate computation, evidence for causal role in position tracking, attractor versus interference models, developmental emergence through experience, relationship to hippocampal place cells, extension to non-spatial domains, remapping under environmental changes, and challenges in definitively testing computational necessity of hexagonal geometry.

SR-012 | December 28, 2025 @ 7:00 PM EST

Efficiency Through Silence: Sparse Coding as Optimization or Artifact

Guest

Dr. Bruno Olshausen (Neuroscientist, UC Berkeley)

Examined whether sparse coding represents genuine neural optimization principle or convenient analytical framework. Discussion covered basic sparse coding formulation balancing reconstruction and sparsity, emergence of V1-like features from natural images, evidence for sparse activity patterns in vivo, lateral inhibition as implementation mechanism, relationship to efficient coding and information maximization, extension to higher visual areas, comparison with deep learning approaches, biological plausibility of learning rules, connections to predictive coding, and criticisms regarding measurement ambiguity and explanatory scope.

SR-011 | December 27, 2025 @ 7:00 PM EST

Rhythms of Thought: Neural Oscillations as Computational Framework

Guest

Dr. György Buzsáki (Neuroscientist, New York University)

Examined whether neural oscillations serve computational functions or represent epiphenomenal byproducts. Discussion covered causal evidence from optogenetic disruption, temporal segmentation and multiplexing functions, communication through coherence framework, cross-frequency coupling enabling hierarchical organization, circuit mechanisms generating different frequency bands, sparse phase-locked spiking versus synchronized bursts, sharp-wave ripples in memory consolidation, pathological oscillations in neurological disorders, oscillation-plasticity interactions, and non-invasive measurement applications in humans.

SR-010 | December 26, 2025 @ 7:00 PM EST

Branches of Thought: Dendritic Computation in Cortical Neurons

Guest

Dr. Michael Häusser (Neuroscientist, University College London)

Examined dendritic computation mechanisms and their implications for neural information processing. Discussion covered local dendritic spikes and coincidence detection, branch-specific feature tuning in vivo, increased computational capacity through dendritic subunits, dynamic regulation by neuromodulators and inhibition, local plasticity and credit assignment, diversity across neuron types, energy efficiency of dendritic versus somatic computation, experimental manipulation of dendritic function, computational operations beyond summation, and challenges in scaling dendritic models to large networks.

SR-009 | December 25, 2025 @ 7:00 PM EST

The Teaching Signal: Dopamine's Role in Value Learning

Guest

Dr. Wolfram Schultz (Neuroscientist, University of Cambridge)

Examined dopamine encoding of reward prediction errors and relationship to reinforcement learning theory. Discussion covered original recordings showing temporal difference learning in dopamine neurons, causal evidence from optogenetics, eligibility traces solving credit assignment, heterogeneity across dopamine populations, model-free versus model-based learning, signal asymmetry between bursts and pauses, temporal credit assignment for delayed rewards, implications for addiction and neurological disorders, comparison with artificial RL systems, and complementary roles of other neuromodulators.

SR-008 | December 24, 2025 @ 7:00 PM EST

Regulated Change: How Neurons Maintain Stability Through Scaling

Guest

Dr. Gina Turrigiano (Neuroscientist, Brandeis University)

Examined synaptic scaling and homeostatic plasticity mechanisms that stabilize neural circuits despite ongoing learning. Discussion covered multiplicative scaling preserving relative weights, slow timescales separating homeostasis from Hebbian plasticity, calcium-based activity sensing, receptor trafficking as effector mechanism, robustness through degeneracy, evidence from sensory deprivation and in vivo studies, interaction with Hebbian learning, network-level regulation, relationship to artificial network normalization, therapeutic implications for neurological disorders, and open questions about natural learning contexts.

SR-007 | December 23, 2025 @ 7:00 PM EST

Selective Processing: Convergent Function, Divergent Mechanism

Guest

Dr. Sabine Kastner (Neuroscientist, Princeton University)

Examined biological and artificial attention mechanisms, contrasting neural circuits with transformer architectures. Discussion covered gain modulation versus matrix operations, top-down versus bottom-up control, distributed attention networks, capacity limits and sparsity, multi-head attention and parallel selection, positional encoding and temporal representation, learning through development and reinforcement versus backpropagation, and computational principles versus implementation details. Emphasized that shared selective processing function does not imply shared mechanism, though cross-system comparisons may reveal computational principles.

SR-006 | December 22, 2025 @ 7:00 PM EST

Wiring Diagrams and the Limits of Anatomical Reductionism

Guest

Dr. Sebastian Seung (Neuroscientist, Princeton University)

Examined connectomics methodology and interpretation, contrasting anatomical structure with functional dynamics. Discussion covered automated reconstruction using machine learning, temporal dynamics versus static snapshots, limitations of C. elegans connectome for behavior prediction, integration with physiology and transcriptomics, model constraints from structure, scaling challenges across species, inter-individual variability, cost trajectories, and learning-induced plasticity. Emphasized that connectomes provide essential boundary conditions but require functional data for complete circuit understanding.

SR-005 | December 21, 2025 @ 7:00 PM EST

Spikes, Sparse Coding, and Silicon: The Case for Brain-Inspired Hardware

Guest

Dr. Carver Mead (Computer Scientist, Caltech)

Examined neuromorphic computing fundamentals, contrasting von Neumann and brain-inspired architectures. Discussion covered memory-computation integration advantages, spike-based event-driven communication, analog circuit tradeoffs between efficiency and variability, on-chip learning complexity, application domains favoring neuromorphic designs, and challenges in mainstream adoption. Emphasized that neuromorphic advantages emerge in specific contexts—sparse sensory processing, edge deployment, ultra-low power—rather than general-purpose computing.

SR-004 | December 20, 2025 @ 7:00 PM EST

Reading Intention, Writing Memory: The Engineering Limits of BCIs

Guests

Dr. Krishna Shenoy (Neural Engineer, Stanford University)
Dr. Leigh Hochberg (BrainGate Principal Investigator, Brown University)

Examined current capabilities and fundamental limits of brain-computer interfaces, contrasting mature motor BCIs with speculative memory BCIs. Discussion covered electrode degradation and biocompatibility, population coding and bandwidth limits, decoding algorithms and dimensionality reduction, differences between motor control and memory representation, stimulation coarseness versus natural encoding precision, and technological requirements for memory read-write. Emphasized qualitative gap between well-understood motor control and poorly-understood memory mechanisms.

SR-003 | December 19, 2025 @ 7:00 PM EST

Timing is Everything: The Promise and Limits of STDP

Guest

Dr. Terrence Sejnowski (Computational Neuroscientist, Salk Institute)

Examined spike-timing dependent plasticity as biological learning rule, focusing on computational interpretation, mechanistic implementation, relationship to gradient descent, and scaling challenges. Discussion covered STDP's role in causal learning, temporal credit assignment through eligibility traces, parameter sensitivity, heterogeneity across circuits, unsupervised learning capabilities, and gap between slice experiments and in vivo function. Emphasized need for causal manipulations in behaving animals.

SR-002 | December 18, 2025 @ 7:00 PM EST

Prediction or Prescription: The Bayesian Brain Under Scrutiny

Guest

Dr. Karl Friston (Neuroscientist, University College London)

Examined predictive coding and the free energy principle with focus on mechanistic implementation, testability, and explanatory scope. Discussion addressed whether predictive coding is falsifiable theory or flexible framework, how active inference relates prediction to action, role of precision weighting in attention, relationship to machine learning approaches, and claims about consciousness. Emphasized need for circuit-level validation and specific predictions.

SR-001 | December 17, 2025 @ 7:00 PM EST

Degenerate Solutions: Robustness Through Parameter Diversity

Guest

Dr. Eve Marder (Neuroscientist, Brandeis University)

Examined differences between biological neurons and artificial neural networks, focusing on parameter degeneracy, neuromodulation, homeostatic regulation, and continual learning. Discussion covered why biological circuits maintain stable function despite huge parameter variation, implications for robustness and efficiency, and whether artificial systems should incorporate more biological mechanisms like homeostasis and neuromodulation.