Unsupervised Thinking

Despite the fact that the brain is full of them, glial cells don't get much attention from neuroscientists. The traditional view of these non-neurons is that they are supportive cells---there to silently help neurons do what they need to do. On this episode we start by describing this traditional view, including types of glial cells and their roles. Then we get into the more interesting stuff. How do glia communicate with each other and with neurons? Turns out there are many chemical messages that get sent between these different cell types, including via the energy molecule ATP! We then talk about the ways in which these messages impact neurons and reasons why the role of glia may be hard for neuroscientists to see. In particular, glia seem to have a lot to say about the birth and control of synapses, making them important for scientists interested in learning. Finally we cover some of the diseases related to glia, such as multiple sclerosis and (surprisingly) depression. Throughout, we ask if glia are important for computation, and relatedly, how the hell do we define computation? Also Grace is weirded out that glia are everywhere but nobody is talking about (or drawing) them.

For decades, neuroscientists have explored the ways in which neurons update and control the strength of their connections. For slightly fewer decades, machine learning researchers have been developing ways to train the connections between artificial neurons in their networks. The former endeavour shows us what happens in the brain and the latter shows us what's actually needed to make a system that works. Unfortunately, these two research directions have not settled on the same rules of learning. In this episode we will talk about the attempts to make artificial learning rules more biologically plausible in order to understand how the brain is capable of the powerful learning that it is. In particular, we focus on different models of biologically-plausible backpropagation---the standard method of training artificial neural networks. We start by explaining both backpropagation and biological learning rules (such as spike time dependent plasticity) and the ways in which the two differ. We then describe four different models that tackle how backpropagation could be done by the brain. Throughout, we talk dendrites and cell types and the role of other biological bits and bobs, and ask "should we actually expect to see backprop in the brain?". We end by discussing which of the four options we liked most and why!

This very special episode of Unsupervised Thinking takes place entirely at the IBRO-Simons Computational Neuroscience Imbizo in Cape Town, South Africa!

Computational neuroscience is a very interdisciplinary field and people come to it in many different ways from many different backgrounds. In this episode, you'll hear from a variety of summer school students who are getting some of their first exposure to computational neuroscience as they explain their background and what they find interesting about the field. In the second segment of the episode, we go into a conversation with the teaching assistants at the Imbizo about what could make training in computational neuroscience better in the future and what we wish we had learned when we entered the field. Finally, we throw it back to the students to summarize the impact this school had on them and their future career plans.

In the past few years, we've noticed researchers making more explicit efforts to engage with scientists in other countries, particularly those where science isn't well-represented. Inspired by these efforts, we took a historical dive into the international element of science with special guest Alex Antrobus. How have scientists viewed and communicated with their peers in other countries over time? To what extent do nationalist politics influence science and vice versa? How did the euro-centric view of science arise? In tackling these issues, we start in the 1700s and work our way up to the present, covering the "Republic of Letters," the Olympic model of scientific nationalism, communism, and decolonization. We end by discussing the ethical pros and cons of mentoring and building academic "outposts" in other countries. Throughout, we talk about the benefits of open science, the King of Spain's beard, and how Grace doesn't do sports.

Cerebellum literally means "little brain," and in a way, it has been treated as a second-class citizen in neuroscience for awhile. In this episode we describe the traditional view of the cerebellum as a circuit for motor control and associative learning and how its more cognitive roles have been overlooked. First we talk about the beautiful architecture of the cerebellum and the functions of its different cell types, including the benefits of diversity. We then discuss the evidence for non-motor functions of the cerebellum and why this evidence was hard to find until recently. During this, we struggle to explain what cognitive issues someone with a cerebellar lesion may have and special guest/cerebellum expert Alex Cayco-Gajic tests our cerebellar function. Finally, we end by lamenting the fact that good science is impossible and Alex tells us how the future of neuroscience is subcortical!

Reinforcement learning is important for understanding behavior because it tells us how actions are guided by reward. But the topic also has a broader significance---as an example of the happy marriage that can come from blending computer science, psychology and neuroscience. In this way, RL is a poster child for what's known as Marr's levels analysis, an approach to understanding computation that essentially asks why, how, and where. On this episode we first define some of the basic terms of reinforcement learning (action, state, environment, policy, value). Then we break it down according to Marr's three levels: what is the goal of RL? How can we (or an artificial intelligence) learn better behavior through rewards? and where in the brain is this carried out? Also we get into the relationship between reinforcement learning and evolution, discuss what counts as a reward, and try to improvise some relatable examples involving cake, cigarettes, chess, and tomatoes.

In part two of our conversation on what counts as an explanation in science, we pickup with special guest David Barack giving his thoughts on the "model–mechanism–mapping" criteria for explanation. This leads us into a lengthy discussion on explanatory versus phenomenological (or "descriptive") models. We ask if there truly is a distinction between these model classes or if a sufficiently good description will end up being explanatory. We illustrate these points with examples such as the Nernst equation, the Hodgkin-Huxley model of the action potential, and multiple uses of Difference of Gaussians in neuroscience. Throughout, we ask such burning questions as: can a model be explanatory if the people who made it thought it wasn't? are diagrams explanations? and, is gravity descriptive or mechanistic?

As scientists, we throw around words like "explanation" a lot. We assume explaining stuff is part of what we're doing when we make and synthesize discoveries. But what does it actually take for something to be an explanation? Can a theory or model be successful without truly being one? How do these questions play out in computational neuroscience specifically? We bring in philosopher-neuroscientist David Barack to tackle this big topic. In part one of the conversation, David describes the historical trajectory of the concept of "explanation" in philosophy. We then take some time to try to define computational neuroscience, and discuss "computational chauvinism": the (extremist) view that the mind could be understood and explained independently of the brain. We end this first half of the conversation by defining the "3M" model of explanation and giving our initial reactions to it.

Machine learning has been making big strides in a lot of straightforward tasks, such as taking an image and labeling the objects in it. But what if you want an algorithm that can, for example, generate an image of an object? That's a much vaguer and more difficult request. And it's where generative models come in! We discuss the motivation for making generative models (in addition to making cool images) and how they help us understand the core components of our data. We also get into the specific types of generative models and how they can be trained to create images, text, sound and more. We then move onto the practical concerns that would arise in a world with good generative models: fake videos of politicians, AI assistants making our phone calls, and computer-generated novels. Finally, we connect these ideas to neuroscience, asking both how can neuroscientists make use of these and is the brain a generative model?

Because of the sheer number of neurons in the gut, the enteric nervous system is sometimes called the second brain. What're all those neurons doing down there? And what, or who, is controlling them? Science has recently revealed that the incredibly large population of microorganisms in the gut have a lot to say to the brain, by acting on these neurons and other mechanisms, and can impact everything from stress to obesity to autism. In this episode, we give the basic stats and facts about the enteric nervous system (and argue about whether it really is a "second brain") and cover how the gut can alter the brain via nerves, hormones, and the immune system. We then talk about what happens when mice are raised without gut microbes (weird) and whether yogurt has any chance of curing things like anxiety. Throughout, we marvel at how intuitive all this seems despite being incredibly difficult to actually study. All that plus: obscure literary references, Josh's hilariously extreme fear of snakes, multiple misuses of the word "species," and DIY feces transplants! (Apologies for minor audio issues on this one!)