Brain Space Time Podcast Akseli Ilmanen

Uri Hasson runs a lab in Princeton, where he investigates the underlying neural basis of natural language acquisition and processing as it unfolds in the real world. As Uri visited Tübingen (where I am doing my master's), we were able to meet in person. Originally, I planned to talk about his idea of temporal receptive windows, and how different brain regions (e.g. default mode network) operate at different timescales. However, we ended up talking more about Wittgenstein, evolution, and ChatGPT. An underlying thread throughout the conversation was that (for both biological and artificial agents), language is not clever symbol and rule manipulation but a brute force fitting to statistics across (Wittgensteinian) 'contexts'. This view is best articulated in Uri's Direct Fit paper. We also connect this to transformers and discuss what's missing in AI. The answer here is multimodal integration, episodic memory, and interactive sociality). At the end, I ask Uri about his 1000 days project, talking to crows, and "understanding" in neuroscience/AI.



For Apple Podcast users, find books/papers links at: https://akseliilmanen.wixsite.com/home/post/pod08




Uri's Website
Twitter: @HassonLab
Uri's publications & talks:

Hasson et al., 2015 - Hierarchical process memory: memory as an integral component of information processing Temporal receptive windows paper
Hasson et al., 2020 - Direct Fit to Nature: An Evolutionary Perspective on Biological and Artificial Neural Networks paper
Yeshurun et al., 2021 - The default mode network: where the idiosyncratic self meets the shared social world paper
Goldstein et al., 2022 - The Temporal Structure of Language Processing in the Human Brain Corresponds to The Layered Hierarchy of Deep Language Models preprint
Nguyen et al., 2022 - Teacher student neural coupling during teaching and learning paper
Goldstein et al., 2022 - Shared computational principles for language processing in humans and deep language models paper

Also mentioned:

Podcast episode with Tony Zador on Genomic Bottlenecks link





My Twitter @akseli_ilmanen
Email: akseli.ilmanen[at]gmail.com
Brain Space Time Podcast, my blog, other stuff
Music: Space News, License: Z62T4V3QWL



Timestamps:

(00:00:00) - Intro

(00:04:52) - Studying language in the real world

(00:07:57) - Wittgenstein

(00:11:10) - Evolution and the default mode network

(00:20:54) - Overparameterized deep learning works

(00:25:02) - Direct Fit paper and generalization

(00:39:37) - Episodic memory and sociality in language models

(00:47:15) - 1000 days project and talking to crows

(00:52:14) - "Understanding" in neuroscience


---

Send in a voice message: https://podcasters.spotify.com/pod/show/akseli-ilmanen/message

Kevin Mitchell is an Associate Professor of Genetics and Neuroscience at the Trinity College Dublin. He recently published his second book, "Free Agents: How Evolution Gave Us Free Will." It's a rigorous defense for why we (and other living systems) have free will, arguing all the way from quantum indeterminacy, to C. elegans, to how humans can form abstracted meanings over very long timescales. We also go beyond the book, exploring how free will links to unresolved questions in physics about the discrepancy of microscopic laws being time-invariant and macroscopic laws having a time asymmetry (entropy increase over time). And how the 'present' does it exist and how its duration might differ for a fly vs a human. Kevin also does a great job of explaining why top-down causality and meaning are not just some mythical concepts, but how it scientifically makes sense to speak of neural activity in terms of 'what this means for the orgasm', and how coarse-gaining allows hierarchical control structures to do causal work on this 'meaning-level'. In the end, we also talk about what kind of research Kevin would like to see and advice on learning across disciplines.


For Apple Podcast users, find books/papers links at: https://akseliilmanen.wixsite.com/home/post/pod07


Kevin's Website
Twitter: @WiringTheBrain
Kevin's publications & talks:

Mitchell, 2020 - Innate: How the Wiring of Our Brains Shapes Who We Are book
Potter et al., 2022 - Naturalising Agent Causation paper
Mitchell, 2023 - The origins of meaning – from pragmatic control signals to semantic representations preprint
Mitchell, 2023 - Free Agents: How Evolution Gave Us Free Will book
Mitchell et al., 2023 - Robert Sapolsky vs Kevin Mitchell: The Biology of Free Will | Philosophical Trials #15 YouTube
Mitchell, 2023 - Reflections on “Systems – the Science of Everything” Blog

Other papers/books mentioned:

Smolin et al., 2021 - The quantum mechanics of the present preprint


           
Neuroscience and Philosophy Salon website




My Twitter @akseli_ilmanen
Email: akseli.ilmanen[at]gmail.com
Brain Space Time Podcast, my blog, other stuff
Music: Space News, License: Z62T4V3QWL





Timestamps:

(00:00:00) - Intro

(00:02:40) - The Free Will skeptics

(00:12:56) - Quantum indeterminacy, the weather, and living systems

(00:23:09) - C. elegans and how evolution exploits noise

(00:38:08) - The arrow of time and the quantum mechanics of the present

(00:43:50) - 'How long' is the present for flies vs humans

(00:52:14) - Top-down causality on the biological implementation level

(01:00:03) - Meaning as functional (not epiphenomenal) and Robert Nozick's pleasure machine

(01:05:34) - Interdisciplinary science and education




---

Send in a voice message: https://podcasters.spotify.com/pod/show/akseli-ilmanen/message

Kate Jeffery is the head of the school of psychology & neuroscience at the University of Glasgow (formerly at UCL). This episode is all about grid cells (background info), which Kate was already recording in the 1990s. We discuss how grid cells' rate maps differ when the rats climb in 3D spaces. Here we cover anything from cross-species comparisons (bats, birds), to self-organizing dynamics, and symmetry breaking. Kate also shares her (maybe unpopular) thoughts that the hexagonal grid regularity is not functional but a by-product. We also get physics-y by discussing entropy, evolution, complexity and how they link to memory and the arrow of time. At the end there is career advice and some thoughts on climate change.


For Apple Podcast users, find books/papers links at: https://akseliilmanen.wixsite.com/home/post/pod06


Not familiar with place, grid or head direction cells? Here is my 5min primer.




Kate's Website
Kate's publications:

Jeffery et al., 2015 - Neural encoding of large-scale three-dimensional space—properties and constraints paper
Casali et al., 2019 - Altered neural odometry in the vertical dimension paper
Jeffery et al., 2019 - On the Statistical Mechanics of Life: Schrödinger Revisited paper
Jeffery et al., 2020 - Transitions in Brain Evolution: Space, Time and Entropy paper
Grieves et al., 2021 - Irregular distribution of grid cell firing fields in rats exploring a 3D volumetric space paper
Jeffery, 2022 - Symmetries and asymmetries in the neural encoding of 3D space paper
Rae et al., 2022 - Climate crisis and ecological emergency: Why they concern (neuro)scientists, and what we can do paper



Other reading mentioned:

Cheng, 1986 - A purely geometric module in the rat's spatial representation paper
My article on Michel Foucault and climate change deniers





My Twitter @akseli_ilmanen
Email: akseli.ilmanen[at]gmail.com
The Embodied AI Podcast, my blog, other stuff
Music: Space News, License: Z62T4V3QWL



Timestamps:

(00:00:00) - Intro

(00:02:14) - Missing out on a Nobel Prize

(00:11:05) - Place cells & grid cells interactions

(00:15:19) - Grid cells and rats climbing in 3D

(00:27:24) - (Spatial) ecological niches of rats, bats and birds

(00:32:55) - Self-organizing dynamics

(00:35:36) - 'Speed' in navigating physical vs abstract spaces

(00:40:19) - 3D = 2D planes stitched together?

(00:46:22) - Symmetry breaking in

(00:50:20) - 'A purey geometric module' (Cheng, 1986)

(01:01:24) - Why are grid cells grid-like?

(01:05:22) - Kate's (grid cell) secrets

(01:08:18) - Entropy, evolution, and complexity

(01:17:45) - Memory as metastable states

(01:22:07) - Entropy, memory & the arrow of time

(01:25:03) - Career Advice

(01:28:35) - Climate change & sociology

(01:38:07) - New position in Glasgow


---

Send in a voice message: https://podcasters.spotify.com/pod/show/akseli-ilmanen/message

Two weeks ago, I visited the Bernstein conference in Berlin. I had lots of fun, particularly at the poster sessions, where I met William, Movitz, and Shervin. I met with each of them later and recorded the following conversations (on bark benches again^^).

William Walker (Gatsby Computational Neuroscience Unit, London) had a poster on 'Representations of State in Hippocampus Derive from a Principle of Conditional Independence'. We discuss how current deep learning struggles with generalization, lacks priors, and could benefit by learning latent conditionally independent representations (similar to place cells).

Movitz Lenninger (KTH Royal Institute of Technology, Stockholm) had a poster on 'Minimal decoding times for various shapes of tuning curves'. He was puzzled why neurons with periodic tuning curves (such as grid cells) are so rare in the brain considering their superior accuracy. He posits there may be a trade-off between accuracy and encoding time.

Shervin Safavi (Max Planck Institute for Biological Cybernetics, Tübingen) had a poster on linking efficient coding and criticality. We introduce those concepts and talk about why noise is a feature, not a bug. Shervin is also starting a new lab at TU Dresden, where he wants to understand the computational machinery of cognitive processes and he is looking for interdisciplinary-minded applicants!


For Apple Podcast users, find books/papers links at: https://akseliilmanen.wixsite.com/home/post/pod05


Not familiar with place, grid and head direction cells? Here is my 5min primer.






William's publications:

Walker et al., 2023 - Unsupervised representation learning with recognition-parametrised probabilistic models preprint
Walker et al., 2023 - Prediction under Latent Subgroup Shifts with High-Dimensional Observations preprint



Movitz's LinkedIn


Movitz's poster from another conference:


Movitz's publications:

Lenninger et al., 2022 - How short decoding times, stimulus dimensionality and spontaneous activity constrain the shape of tuning curves: A speed-accuracy trade-off preprint
Lenninger et al., 2023 - Are single-peaked tuning curves tuned for speed rather than accuracy? paper

Shervin's Website
Twitter: @neuroprinciples
For Shervin's new lab: interest mailing list
Shervin's publications:

Safavi et al., 2022 - Multistability, perceptual value, and internal foraging paper
Safavi et al., 2023 - Signatures of criticality in efficient coding networks preprint

Synchronization of metronomes video


My Twitter @akseli_ilmanen
Email: akseli.ilmanen[at]gmail.com
The Embodied AI Podcast, my blog, other stuff
Music: Space News, License: Z62T4V3QWL



(00:00:00) - Intro

(00:02:53) - William Walker

(00:32:53) - Movitz Lenninger

(00:55:04) - Shervin Safavi


---

Send in a voice message: https://podcasters.spotify.com/pod/show/akseli-ilmanen/message

An episode with my favourite podcast host, Paul Middlebrooks. Paul and I met in Berlin, and talked about his journey away from (and back into) academia and why he started his podcast BrainInspired. Yes, there is a lot of podcast meta-talk in this episode. For example, how science podcasts give you a glimpse into another field (as an outsider) and some advice for fellow podcast hosts. We also get into productivity, self-learning and some big-picture questions on what's holding neuroscience back.


For Apple Podcast users, find books/papers links at: https://akseliilmanen.wixsite.com/akseli-ilmanen/post/pod04




Paul's Podcast BrainInspired
His NeuroAI course with topics ranging from the history of ANNs to explaining variational autoencoders
Twitter: @pgmid






Epistimones podcast by Paco Chow and Megan Lee






My Twitter @akseli_ilmanen
Email: akseli.ilmanen[at]gmail.com
Brain Space Time Podcast, my blog, other stuff
Music: Space News, License: Z62T4V3QWL




Timestamps:

(00:00:00) - Intro

(00:01:51) - Interesting conversations Paul had at the conference

(00:07:57) - The why and how of podcasting

(00:11:16) - Changing one's mind in science

(00:19:34) - Paul's NeuroAI course

(00:20:46) - Podcasts for self-learning & productivity fallacies

(00:26:15) - Podcast advice

(00:30:58) - Paul is back in academia

(00:38:48) - Neuroscience needs theory (beyond manifolds)

(00:45:50) - Saying thank you to Paul


---

Send in a voice message: https://podcasters.spotify.com/pod/show/akseli-ilmanen/message

A couple of weeks ago, I visited the ESI SyNC 2023 conference in at the Ernst Strüngmann Institute (ESI) in Frankfurt, Germany. Their topic was "Linking hypotheses: where neuroscience, computation, and cognition meet".

During the conference, I got talking to Yossi Yovel (Tel-Aviv University) about how different bat species navigate, what their vocalizations tell us about language evolution, and discussed his recent paper on whether we will ever be able to talk to animals. On the last point, I have some strong thoughts - thoughts including Wittgenstein and crows (see my own article here).

I also chat with Francisco Garcia-Rosales (ESI) on his poster about oscillations in the bat auditory and frontal cortex, and how bats and marmosets are really good animal models for speech (and maybe language).

Sarah Robins is a philosopher at Purdue University. Based of fMRI studies, many neuroscientists have grouped memory and imagination as a single phenomena. Sarah has been busy disentangling the two and we discuss how constructivist accounts of memory might have gone too far when abandoning memory traces.

David Poeppel (ESI) has a lab on auditory cognition, music, speech and language and how they map to neurobiology. Yet, going beyond that David has some intriguing thoughts on what's missing in neuroscience more generally. We dig deep into why we need a theory of memory storage/retrieval ("engram renaissance") and how to do interdisciplinary science.



For Apple Podcast users, find books/papers links at:

https://akseliilmanen.wixsite.com/home/post/pod03




Yossi's Website
Twitter: @YovelBatLab
Yossi's talk available here in October-ish
Mentioned books/papers:

Genzel et al., 2018 - Neuroethology of bat navigation paper
Yovel et al., 2023 - AI and the Doctor Dolittle challenge paper
Amit et al., 2023 - Bat vocal sequences enhance contextual information independently of syllable order paper
Khait et al., 2023 - Sounds emitted by plants under stress are airborne and informative paper
My article: Talking to a crow will be possible in 50 years



Francisco's LinkedIn
Twitter: @fgarciaro92
Mentioned books/papers:

García-Rosales et al., 2023 - Oscillatory waveform shape and temporal spike correlations differ across bat frontal and auditory cortex preprint



Sarah's Website
Twitter: @SarahKRobins
Sarah's talk available ⁠here⁠ in October-ish
Mentioned books/papers:

Robins, 2022 - Episodic memory is not for the future book chapter
Ménager et al., 2022 - Modeling human memory phenomena in a hybrid event memory system paper
Robins, 2023 - The 21st century engram paper
Brigard, 2023 - Counterfactual Thinking paper



David's Website
Twitter: @davidpoeppel
Mentioned:

Gallistel, 2021 - The physical basis of memory paper
Poeppel et al., 2022 - We don’t know how the brain stores anything, let alone words paper
Recent talk⁠ by Hessam Akhlaghpour on an RNA-Based Theory of Natural Universal Computation





My Twitter @akseli_ilmanen
Email: akseli.ilmanen[at]gmail.com
The Embodied AI Podcast, my blog, other stuff
Music: Space News, License: Z62T4V3QWL





(00:00:00) - Intro

(00:01:55) - Yossi Yovel on bat navigation, calls & talking to animals

(00:44:45) - Francisco on calls and oscillations in bats and marmosets

(00:59:35) - Sarah Robins on engrams, memory & imagination

(01:39:00) - David Poeppel on why we need a theory of memory storage and retrieval


---

Send in a voice message: https://podcasters.spotify.com/pod/show/akseli-ilmanen/message