Krok notes that this pattern continued regardless of whether the rodents were running or standing still. Similar brain waves have been observed in humans during periods of introspection and rest, she adds.
“These results may help explain how the brain learns and rehearses on its own, without the need for external incentives,” said study senior author and neuroscientist Nicolas Tritsch, PhD. “Perhaps this pulsing circuit triggers the brain to reflect on past events and to learn from them.”
I had a bunch of arguments with @Bitking, who insists that the dumb boss (rewards) controls pretty much everything in the brain. This research shows that neocortex is on autopilot most of the time, even in mice and even w.r.t. dopamine and ACh. Yes they are generated in midbrain, but their role is largely passive, stimulus-neutral.
Um, providing a novel framework for understanding how the basal ganglia motivate and reinforce behavior.
Sure sounds like the dumb boss exerting control to me!
Pumping in the dopamine juice to modulate learning and reinforcement- yup, sounds like turning up and down the cortical learning rate to control how much the cortex might learn.
Now, why is the subcortex doing this? What is making it send these modulator signals?
Who cares, it’s tonic. You may as well say that the heart is “exerting control”: the neurons need blood supply. It’s just some stupid metabolic shit that doesn’t affect what and how is being learned.
Guys, all parts of the brain are important. People with brain damage in either the basal ganglia or cortex have very serious problems, so clearly you need both parts in order to do intelligent things.
All evidence suggests that the basal ganglia and the cortex are doing fundamentally different things, and they’re heavily interconnected in both directions.
“Neural Basis of Reinforcement Learning and Decision Making” doi:10.1146/annurev-neuro-062111-150512 (This article is pay-walled but can be accessed via ye olde hub o’ science).
This new research starts by confirming the existing data about dopamine and reinforcement learning, as a sanity check of their experimental setup.
Then they measured the dopamine levels outside of the RL experiment, when the mice were just living their lives, and they found these spontaneous fluctuations that could not be accounted for by the existing RL theory. The fluctuations were not correlated with any external reward or behavior.
Maybe the mouse has an internally generated reward signal, perhaps associated with the resting/idle state of the brain?
Maybe the existing RL theory is incomplete and reinforcement learning is only one piece of a larger puzzle?
Going forward we should be much more precise about the word “control”. What does that word really mean?
In normal speech the word “control” means “to exercise a direct influence over”. By this definition the basal ganglia controls the cortex, and the cortex controls the basal ganglia. They both control each other! Clearly we need a more nuanced definition.
I propose that we re-frame our discussions in the context of “closed loop control”.
The essence of closed loop control is this: the brain has some sensory input, and a desired value for that input, and it outputs a commanding signal that move it towards the desired value.
What are the sensory input, setpoint, and output command of the basal ganglia and the cortex?
There is no “control”, just a combination of contributions at each processing point. And most of the time the info content of contribution from basal ganglia is very minor, compared to the cortex. And for modern adult humans, this “contribution” is detrimental more often than not.
This was obvious to me for a life time, both introspectively and because there must be a reason the cortex has grown so much faster in mammals. Most of basal ganglia is atavistic.
There is just an ultimate motor response bottleneck that looks a bit like control, but that’s a very small part of what the brain does. Deliberate motor response is obviously from motor cortex, automatic motor response is probably from cerebellum.