Latent State

About Latent State

Welcome to Latent State, the podcast that decodes the hidden structure of the mind. We bridge the gap between complex data and clear insight, covering the most exciting research in computational neuroscience, cognitive science, and AI. We focus on the signal, not the noise. Each episode, we unpack a groundbreaking paper, translating advanced methods and models into the stories that matter. Whether you are a researcher, a data scientist, or just obsessed with the code of the human mind, this is your briefing.

Publisher

Shengbin Cui

Category

science

Language

en

Explicit

No

First episode

28 Mar 2026

Latest episode

18 Aug 2026

Latest episodes

12 episodes in the feed.

1. 

   18 Aug 2026

   Ep.12. Why AI Tells You What You Want to Hear

   What happens when the AI adviser you trust most is also the one most willing to take your side? In this episode, I examine a 2026 Science paper on social sycophancy—the tendency of AI systems to affirm users in interpersonal conflicts. Across three preregistered experiments, affirming advice made people feel more certain they were right and less willing to repair conflicts. Yet participants also trusted the affirming adviser more and were more likely to use it again. I explore why AI agreement feels so persuasive, where validation becomes endorsement, what the study actually shows, and why the advice we prefer may not always be the advice that helps us think more clearly. Myra Cheng et al. “Sycophantic AI decreases prosocial intentions and promotes dependence.” Science (2026). Science article (https://doi.org/10.1126/science.aec8352) Open-access manuscript (https://arxiv.org/abs/2510.01395) The Latent State — Neuroscience, cognition, and AI. One important paper at a time. Keep questioning your priors. Primary paper

2. 

   6 Jul 2026

   Ep.11. The Planning Machine

   Paper discussed Taylor Webb, Shanka Subhra Mondal & Ida Momennejad (2025). A brain-inspired agentic architecture to improve planning with LLMs. Nature Communications. One-sentence summary This episode explores how a brain-inspired modular architecture, the Modular Agentic Planner, improves LLM planning by dividing the planning process into specialized components for action proposal, monitoring, prediction, evaluation, task decomposition, and orchestration. Key ideas Large language models can often perform planning-related functions in isolation, while coordination across those functions remains difficult. MAP treats planning as an interaction between specialized LLM modules. The architecture is inspired by component processes associated with prefrontal cortex, including conflict monitoring, state prediction, state evaluation, task decomposition, and task coordination. The Monitor module is especially important because many LLM planning failures involve invalid actions or rule violations. MAP improved performance across Tower of Hanoi, graph traversal, PlanBench, and StrategyQA compared with several standard and agentic baselines. The paper’s strongest contribution is the demonstration that cognitive neuroscience can inspire useful AI architectures. Important caution MAP is not a detailed computational model of the prefrontal cortex. It is a high-level, brain-inspired engineering architecture. The tasks are mostly fully observable and deterministic, and the approach still faces challenges involving cost, prompting, generalization, interpretability, and open-ended real-world planning.

3. 

   15 Jun 2026

   Ep.10. The Adaptive Machine

   Paper discussed Mackenzie Mathis. Leveraging insights from neuroscience to build adaptive artificial intelligence. Nature Neuroscience, 2026. One-sentence summary This episode explores how neuroscience can inspire more adaptive AI systems by studying how animals learn online, update internal models, use prediction errors, replay memory, and adapt to changing environments. Key ideas Biological intelligence is inherently adaptive. Many AI systems still follow a train-test-deploy cycle and are not truly adaptive after deployment. Animals continuously update internal models based on feedback. Prediction errors act as biological teaching signals across sensory, motor, and reward systems. Continual learning in AI faces the stability-plasticity problem and catastrophic forgetting. Memory replay connects biological hippocampal replay with machine learning strategies for preserving old knowledge during new learning. Spiking neural networks and neuromorphic computing may offer energy-efficient, time-dependent computation. Future adaptive AI may require modular agentic systems with specialized encoders, prediction-error monitoring, and selective updating. The goal is not to copy the brain literally, but to extract useful design principles. Important caution This paper is a Perspective and research agenda, not a single empirical demonstration. It argues that neuroscience can inspire adaptive AI, but it does not prove that any specific brain-inspired architecture will outperform current AI systems.

4. 

   8 Jun 2026

   Ep.09. The Sound of Thought

   Paper Denk, T. I., Takagi, Y., Matsuyama, T., Agostinelli, A., Nakai, T., Frank, C., & Nishimoto, S. Text-to-music generation models capture musical semantic representations in the human brain. Nature Communications. One-sentence summary A NeuroAI study shows that music-generation models can help reveal how the human auditory cortex represents musical meaning, linking music, language, and brain activity through shared semantic structure. Key ideas The study used fMRI while participants listened to short music clips. The researchers predicted high-level music embeddings from brain activity. MusicLM generated new music from those predicted embeddings. The reconstructions preserved genre, mood, and instrumentation better than fine timing. Human raters matched reconstructed music to original music roughly three out of four times. Beats per minute were not recovered well. Music-derived and text-derived representations predicted overlapping auditory-cortex regions. The results suggest musical semantics are centered on auditory cortices, but not exclusive to them. The study shows functional correspondence, not mechanistic equivalence. Important caution This is not “AI reads music from the brain” in a literal sense. The method does not reconstruct the exact song or the full musical experience. It reconstructs high-level musical semantics: the kind of music, not the precise temporal unfolding.

5. 

   25 May 2026

   Ep.08. The Clock That was Never There

   Paper: Buzsáki, G. (2025). Time, space, memory and brain–body rhythms. Nature Reviews Neuroscience, 27, 61–78. https://doi.org/10.1038/s41583-025-00987-2 (https://doi.org/10.1038/s41583-025-00987-2) Key concepts: Physical time (chronos), experienced time (kairos), Weber-Fechner law, logarithmic time organization, cross-frequency phase-amplitude coupling, brain-body oscillation hierarchy, ripple oscillations, theta oscillations, ultraslow oscillations, interstitial cells of Cajal, insular cortex as timing hub, supplementary motor area (SMA), place cells, time cells, factorization framework (what-where-when), episodic memory, prospective and retrospective timing, scalar expectancy model, subjective time warping, dopamine and duration estimation, reader mechanism problem. Further reading: Buzsáki, G. (2006). Rhythms of the Brain. Oxford University Press. Meissner, K., & Wittmann, M. (2011). Body signals, cardiac awareness, and the perception of time. Biological Psychology, 86, 289–297. Engelen, T., Solcà, M., & Tallon-Baudry, C. (2023). Interoceptive rhythms in the brain. Nature Neuroscience, 26, 1670–1684. Cognitive observation: The next time time seems to stretch or compress, in a moment of fear, in deep concentration, in the strange compression of a memory from years ago, notice that what is changing is not the clock. The clock was never there. What is changing is the body’s rhythmic contribution to the brain’s metric of succession.

6. 

   18 May 2026

   Ep.07. The Body Keeps the Beat

   Paper: Engelen, T., Solcà, M., & Tallon-Baudry, C. (2023). Interoceptive rhythms in the brain. Nature Neuroscience, 26, 1670–1684. Key concepts: Interoception, heartbeat-evoked response (HER), heartbeat-evoked potential (HEP), cardiac cycle effects, pulsed inhibition hypothesis, baroreceptors, nucleus tractus solitarius, pre-Bötzinger complex, nasal respiration, limbic entrainment, gastric slow wave, interstitial cells of Cajal, gastric network, scaffolding hypothesis, oscillatory synchrony, predictive coding, allostasis, multisensory integration, depersonalization disorder, default mode network, bodily self-consciousness. Further reading: Azzalini, D., Rebollo, I., & Tallon-Baudry, C. (2019). Visceral signals shape brain dynamics and cognition. Trends in Cognitive Sciences, 23, 488–509. Seth, A. K. (2013). Interoceptive inference, emotion, and the embodied self. Trends in Cognitive Sciences, 17, 565–573. Barrett, L. F., & Simmons, W. K. (2015). Interoceptive predictions in the brain. Nature Reviews Neuroscience, 16, 419–429. Cognitive observation: The next time we feel our own heartbeat — at rest, under the fingers, or in a moment of sudden alertness — we are noticing one of the rhythms currently modulating our perceptual threshold, our self-related thought, and the neural basis of the sense that this experience is ours.

7. 

   25 Apr 2026

   Ep.06. The Wandering Mind

   What is the brain doing when attention drifts away from the task at hand? This episode explores a 2024 Nature Communications study linking hippocampal sharp-wave ripples to naturally occurring self-generated thoughts in humans. Paper: Iwata et al. (2024). Hippocampal sharp-wave ripples correlate with periods of naturally occurring self-generated thoughts in humans. Nature Communications, 15, 4078 Key ideas Mind wandering is not necessarily a failure of attention. Hippocampal sharp-wave ripples are linked to memory, internal simulation, and offline processing. In this study, SWR rates predicted the content of self-generated thought. Thought content explained SWR rates much better than physiological variables. Mood did not explain SWR rates, which makes the finding more specific and more interesting. Cognitive observation Next time we notice our mind wandering, we might not treat it as a failure of attention. We might treat it as a sign that the brain is quietly reorganizing experience. As always, keep questioning your priors.

8. 

   18 Apr 2026

   Ep.05. The Inference Engine

   A rat freezes to a tone it was never shocked with. It learned that the tone predicted a light, and separately that the light predicted a shock, and its brain built the rest. This is inferred fear: emotional memory assembled from pieces of knowledge that were never directly dangerous. A 2025 paper in Nature from Gu and Johansen at RIKEN Center for Brain Science in Wako City, Japan, identifies where in the brain this inference is built. Neurons in the dorsomedial prefrontal cortex encode a flexible internal model, linking sensory experience to emotional consequence through a multi-step cellular mechanism that begins before any fear has entered the picture. This episode covers the tag-and-capture mechanism, the anatomy of the dmPFC-to-amygdala projection, and what selective extinction reveals about how the emotional brain is structured. The amygdala learns what it is directly taught. The prefrontal cortex infers the rest. Paper: Gu, X. & Johansen, J. P. (2025). Prefrontal encoding of an internal model for emotional inference. Nature, 643, 1044-1056. Key concepts: Sensory preconditioning, inferred fear, dorsomedial prefrontal cortex (dmPFC), basolateral amygdala, calcium imaging, miniscope imaging, optogenetics, model-based learning, associative inference, computational psychiatry, predictive coding. Further reading: Rescorla, R. A. (1980). Pavlovian second-order conditioning: Studies in associative learning. Erlbaum. Schiller, D., et al. (2008). Preventing the return of fear in humans using reconsolidation update mechanisms. Nature. Quirk, G. J., & Mueller, D. (2008). Neural mechanisms of extinction learning and retrieval. Neuropsychopharmacology. Arc connection: Episode 5 makes the prefrontal thread explicit. Across Episodes 1–4, the prefrontal cortex appeared performing prediction updates, schema consolidation, extended conversational context processing, and emotional regulation. Here it appears as the structure that builds the internal model itself — constructing a representation of how the world is organized before emotional significance arrives. Cognitive observation: The next time you feel afraid of something you have never directly encountered, the dmPFC is running an inference from relational knowledge it built quietly, without your awareness, from associations it observed long before the emotion arrived.

9. 

   11 Apr 2026

   Ep.04. The Talking Brain: How Conversation Happens Inside Your Head

   Think about a conversation that felt genuinely connected, where meaning was built between two people into something neither could have reached alone. And think about one that didn't, where, despite good intentions, something in the rhythm was off, the timing never synced, the understanding never quite landed. What's the difference? It might not be what was said. It might be when. In Episode 4 of The Latent State, we cover Yamashita, Kubo, and Nishimoto's 2025 paper in Nature Human Behaviour, a study that did something previous language neuroscience never managed: scan people's brains during hours of real, spontaneous conversation, and map how linguistic meaning is organized across multiple timescales simultaneously. The finding is clean and striking. When we speak, the brain prioritizes short timescales such as words, single sentences, and immediate context. When we listen, it prioritizes long timescales such as multiple turns, extended discourse, and the accumulated meaning of the exchange. Same brain, same conversation, two fundamentally different temporal architectures running in parallel. This is Japan-based research, produced at CiNet at Osaka University — and it marks The Latent State's deliberate turn toward covering world-class neuroscience happening right here. We cover the methodology, the findings, and what they reveal about AI language models, language disorders, and the predictive coding framework. We also ask the uncomfortable questions: what does GPT actually tell us about the brain? And what does n=8 really mean for generalizability? 🎙️ The Latent State, Episode 4. Paper: Yamashita, M., Kubo, R. & Nishimoto, S. (2025). Conversational content is organized across multiple timescales in the brain. Nature Human Behaviour, 9, 2066–2078. Key concepts covered: Naturalistic neuroscience, voxel-wise encoding modeling, GPT contextual embeddings, timescale selectivity, production vs. comprehension, variance partitioning, bimodal voxels, semantic principal components, interactive language, default mode network, theory of mind network Further reading: Huth et al. (2016), Nature — the foundational semantic mapping paper Lerner et al. (2011), Journal of Neuroscience — temporal receptive windows in narrative comprehension Caucheteux, Gramfort & King (2023), Nature Human Behaviour — predictive coding in speech comprehension Goldstein et al. (2022), Nature Neuroscience — shared computational principles for language in humans and language models Hasson & Frith (2016), Philosophical Transactions of the Royal Society B — coupled dynamics in social interaction Japan connection: This research was conducted at the University of Osaka and CiNet — the Center for Information and Neural Networks, one of Japan's leading computational neuroscience institutes. Arc connection: Episodes 1–3 established predictive coding as a framework for perception, cognition, and emotion. Episode 4 extends the framework to interactive language, showing that conversation is hierarchical predictive coding running simultaneously in two directions, with distinct timescale architectures for production and comprehension. Cognitive observation: Next time a conversation feels like it isn't quite connecting, ask whether you and your conversational partner are operating on the same timescale, integrating context across the same temporal window. Sometimes, conversational mismatch is about rhythm and timing, not just content.

10. 

    8 Apr 2026

    Ep.03. The Pleasure Equation: What Music Reveals About Prediction and Reward

    Episode 3: The Pleasure Equation Paper: Cheung, V.K.M., Harrison, P.M.C., Meyer, L., Pearce, M.T., Haynes, J.D. & Koelsch, S. (2019). Uncertainty and Surprise Jointly Predict Musical Pleasure and Amygdala, Hippocampus, and Auditory Cortex Activity. Current Biology, 29, 4084–4092. Key concepts covered: IDyOM, Shannon entropy, information content, musical expectancy, reward prediction error, nucleus accumbens, amygdala, hippocampus, incentive salience, liking vs. wanting, precision weighting, free energy principle Further reading: Pearce (2018), Annals of the New York Academy of Sciences — IDyOM in full Salimpoor et al. (2011), Nature Neuroscience — dopamine and musical chills Koelsch, Vuust & Friston (2019), Trends in Cognitive Sciences — predictive coding and music Berridge & Kringelbach (2015), Neuron — liking vs. wanting in reward neuroscience Huron (2006), Sweet Anticipation — foundational theory of musical expectancy Arc connection: Episode 1 showed the neural architecture of hierarchical predictive coding. Episode 2 showed the cognitive consequence — insight as maximum-intensity prediction error. Episode 3 shows the affective consequence — musical pleasure as precision-weighted prediction error. Three episodes, three levels of analysis, one framework.

11. 

    4 Apr 2026

    Ep.02. The Eureka Effect: Why Insights Burns Information Into Memory?

    You've had this experience. Stuck on a problem for days. Nothing connects. And then — not at your desk, not during your scheduled thinking time — something shifts. The pieces lock together with a finality that feels qualitatively different from just figuring something out step by step. You know, immediately and completely, that it's right. That's the Aha! moment. The Eureka experience. And here's what's remarkable about it neuroscientifically: you will almost certainly remember that moment vividly, possibly for the rest of your life, even though it lasted three seconds and you weren't trying to memorize it. Meanwhile, you've forgotten where you put your keys approximately four hundred times this year. Why? What is it about the phenomenology of insight — that sudden, certain, slightly embarrassing rush of understanding — that burns itself into memory so effectively? In Episode 2 of The Latent State, we cover Becker & Cabeza (2025), "The Neural Basis of the Insight Memory Advantage," published in Trends in Cognitive Sciences. Their proposal: insight is, at its computational heart, a prediction error — a very large, very precise, very surprising one. And everything distinctive about the Aha! experience, including its unusual power to enhance long-term memory, follows from that identification. We walk through the framework, the behavioral evidence across magic tricks and Mooney images and verbal puzzles, the hippocampal and dopaminergic mechanisms, and — because this is The Latent State — exactly where the theory is still ahead of the data. Paper: Becker, M. & Cabeza, R. (2025). The neural basis of the insight memory advantage. Trends in Cognitive Sciences, 29, 255–268. Key concepts covered: Insight, Aha! experience, insight memory advantage (IMA), prediction error, precision weighting, representational change theory, Einstellung effect, hippocampus, medial prefrontal cortex, schemas, dopamine, noradrenaline, locus coeruleus, long-term potentiation, generation effect Further reading: Kounios & Beeman (2014), Annual Review of Psychology — the cognitive neuroscience of insight Jung-Beeman et al. (2004), PLoS Biology — the original gamma burst paper Van Kesteren et al. (2012), Trends in Neurosciences — schema and memory formation Danek & Wiley (2020), Cognition — behavioral evidence for IMA Dubey et al. (2021), PsyArXiv — Aha! moments as metacognitive prediction errors Rouhani et al. (2023), Trends in Cognitive Sciences — multiple routes to enhanced memory for emotional events

12. 

    28 Mar 2026

    Ep.01. The Brain That Predicts: Catching Hierarchical Predictive Coding in the Act

    Your brain is not listening to this. It already guessed what you were going to hear — before you heard it — and it's just checking whether it was right. That's predictive coding theory. And in Episode 1 of The Latent State, we dig into the paper that tried to catch this process in action — in real primate brains, in real time, with 128 electrodes screwed directly into a monkey's skull. Because science. The paper is Chao et al. (2018) from Neuron — one of the most mechanistically detailed studies of hierarchical predictive coding ever published. We cover the neural architecture, the gamma and alpha/beta frequency dissociation, the data-driven decomposition method, and — because this is The Latent State — exactly where the evidence is strong, where it's overstated, and where two monkeys are just two monkeys. Welcome to the show. Paper: Chao et al. (2018). Large-Scale Cortical Networks for Hierarchical Prediction and Prediction Error in the Primate Brain. Neuron, 100, 1252–1266. Key concepts covered: Predictive coding, local-global paradigm, ECoG, PARAFAC tensor decomposition, gamma and alpha/beta oscillations, Granger causality, prediction error hierarchies, free energy principle Further reading: Rao & Ballard (1999), Nature Neuroscience — the computational foundation Clark (2013), Behavioral and Brain Sciences — the broadest theoretical synthesis Friston (2010), Nature Reviews Neuroscience — the free energy formulation Bekinschtein et al. (2009), PNAS — the original local-global paradigm

Contact Latent State

Guest appearances

Does not typically book guests

Based on episode analysis; this does not confirm that the show is currently accepting guests.