Yuste C1

Question 1

When did the first nervous systems appear? What does this reveal about the purpose of the brain?

Answer 1

The first nervous system emerged about 750 million years ago. This was right before the split between cnidarians and bilaterians (animals with bilateral symmetry). Alongside this development in the tree of life, we see organisms that can move. The NS is associated with the ability to move (only those species with an NS can move). In order to survive, being able to move means being able to predict the future (what/who in around you in the new environment). Movement = future.

Note: very small organisms can move without NS. But, once you grow to a certain size you need to be able to coordinate activity across muscle cells that are too far apart to communicate among themselves.

Question 2

Learning

Answer 2

The brain as a predictor for action:
Extrapolate your internal model of the world, generated by memories and past experiences, into the future to predict what will happen. Then, you act, physically entering the future and perceiving what happens. Accordingly, by comparing what you expected to happen vs. what did, you make adjustments to your internal model.

Question 3

Traditional way of thinking about the brain

Answer 3

Input-output machine. Based on how reflexes work. In response to an input – sensory stimulus – the brain outputs behaviour. Proponent of this model is Sherrington.
Stimuli –> perception –> decision and behaviour.

Question 4

Implications of the brain being a predictor

Answer 4

You can run the internal program without inputs and choose whether or not to generate behaviour.
The brain is a biological machine that mentally manipulates the world using symbols. Our perception of the world is likely all constructed and internally generated.
This concept of the brain being a predictor necessitated the created of the self, for we needed to be able to mentally manipulate ourselves in relation to our environment in order to have consistent behaviour that promotes survival.

Question 5

Challenges to Sherrington model

Answer 5

The existence of spontaneous brain activity. AKA internal activity not triggered by any input, “the senses”.

Question 6

CPGs

Answer 6

Endogenous patterns of activity in the brain. Named so by Graham Brown, who argues that these could be used by the spinal cord to intrinsically generate walking even in the absence of sensory inputs. –> Cf. today = Bayesian models of predictive coding, surprise minimization, free energy and reinforcement leaning.

Question 7

Bayesian models - general agreement

Answer 7

Animals generate internal statistical models of their environments and of themselves in relation to their environment in order to accurately predict the future, minimizing “surprising” or life-threatening conditions. Animals continually change this model to statistically maximize their model evidence and minimize prediction error. Thus, comparisons are made between intrinsic brain states with incoming sensory information.

Question 8

How the internal model of the world is created

Answer 8

Through neural circuits and according to seven principles: ensembles, hierarchy, wiring, maps, control, learning, and optimization.

Question 9

Neuronal ensembles

Answer 9

Neurons work in groups, firing together. Thoughts = symbols = ensembles of neurons. Thoughts must be able to activate themselves, so that they can exist internally in the absence of outside stimulus. Purpose of ensembles: build a large assortment of internal states that represent the world as symbols. Ultimately –> generates our minds.

Question 10

Hierarchy

Answer 10

Neuronal ensembles are organized in a hierarchy. Each ensemble is a module, together forming a more complex structure. They are organized in layers, with increasing abstraction the higher the layer. The higher order ensembles could represent combinations of lower order ensembles.

Question 11

Wiring

Answer 11

2 different proposed wiring diagrams (of neuronal circuits and ensembles).

1. Labelled lines – wires carry specific information. (Eg neuron coding for sweetness sends this info to other neurons involved in sweetness detection etc).
2. Distributed connectivity model – neuron is indiscriminately connected to many other neurons without any specificity. The more connected you are the less important. Distributed wiring can be either complete or random – they are mathematically equivalent.

Both are present in the brain.

Question 12

Maps

Answer 12

Most specific wiring diagrams are organized into maps. Including topographic maps and likely computational maps, whose orderly organization of info in space likely key to enable fast and efficient processing and computations.

Question 13

Control

Answer 13

Control theory is the mathematical way in which you compare an output with an expectation, compute an error signal, and fine tune your output accordingly for greater future accuracy.

Strategies:

1. Areas of the brain built to detect novel stimuli –> how you compare sensory input with expectations and adjust internal models.
2. Computation of Bayesian probabilities.
3. Reinforcement learning. Error signal can be fed back into the system so that it strengthens certain synapses at the expense of others.

Question 14

Learning

Answer 14

Enables us to improve our predictions based on the result of our actions, and in doing so, build better models.
Hebb’s rule – neurons that fire together work together. They will generate an ensemble. One fires, so does the other. –> Associative learning.

Question 15

Optimization

Answer 15

Brain is an organic computer anchored in the physics and chemistry of its hardware, using it as a computational device.