A particularly interesting part that I did not expect from the title:
> Before the rats encountered the detour, the research team observed that their brains were already firing in patterns that seemed to "imagine" alternate unfamiliar mental routes while they slept. When the researchers compared these sleep patterns to the neural activity during the actual detour, some of them matched.
> “What was surprising was that the rats' brains were already prepared for this novel detour before they ever encountered it,”
Suppose we simplify the scenario and think of experiences as draws from a discrete probability distribution, e.g. p=[0.1, 0.1, 0.7, 0.1].
Suppose further that all events are a draw of type 1, 2, 3, or 4, and that our memory kept a count and updated the distribution - it is essentially a frequency distribution.
When we encounter a stimulus, we have to (1) recognize it and (2) assign a reward valence to it. If we only ever observed '3', the distribution would become very peaked. Correspondingly, this suggests that we would recognize '3' events faster and be better at assigning a reward valence to those events.
Then if we ever encounter a non-3 event, we would recognize it more slowly - it is well-established that recognition is tied to encounter frequency - and do a poorer job assigning reward valence to it. Together this means that we would do a bad job selecting the appropriate response.
Perhaps this scenario-based dreaming keeps us (and rats) primed so we're not flat-footed in new scenarios.
The question then becomes - if these scenarios are purely imagined, where are they being sampled from? If we never observe 1, 2, and 4...how do we know that these are the true list of alternative scenarios?
Yeah this part was pretty weird. How do they know that was caused due to the rats' brains firing because they were imagining unfamiliar routes, vs something completely unrelated to the maze routes at all? Just because the hippocampus flash patterns matched doesn't mean that's what the rats were thinking about while sleeping I'd think
In effect, my position is that biological systems maintain a synchronized processing pipeline: where the hippocampal prediction system operates slightly “ahead” of sensory processing, like a cache buffer.
If the processing gets “behind” the sensory input then you feel like you’re accessing memory because the electrical signal is reaching memory and sensory distribution simultaneously or slightly lagging.
So it means you’re constantly switching between your world map and the input and comparing them just to stabilize a “linear” experience - something which is a necessity for corporeal prediction and reaction.
I think we should be careful about materialistic reductions of awareness. Because some rats dreamed detours that ended up being correct in waking rat life, it does not follow that all instances of deja vu are misfirings. It's a tempting connection to draw, but it does not actually explain how the detours were dreamt to begin with, and this points to a deeper question about awareness in general. If I were pressed for an analogy, I might say something like "just because all books have ink does not mean that all ink lives in books." You know what I mean? There's a superset of experiences that cannot be easily explained away by caching, as tempting as it might be.
Wondering if you have any ideas on this, which can be quite jarring when it happens?
You are thinking about something, and then walk through a doorway into another room and suddenly completely lose track of what you were thinking.
The closest idea I've seen for that is: Jeff Hawkins in his Thousand Brain Theory of Intelligence made a statement that learning is a function of navigation and the world models we construct are set in the context of location we create them.
--------
Edit: Just read your piece on Faith:
"Faith, as it’s traditionally understood, is trivial bullshit compared to the towering, unseen faith we place in the empirical all day everyday."
Absolutely correct, and the traditional understanding of Hebrews 11:1 I don't believe reflects what the author (supposedly Paul) was trying to convey.
πίστις: Pistis can be translated as confidence, as in: I'm confident this chair won't collapse when I sit on on it. Much stronger than belief or faith.
ὑπόστασις: Hupostasis is also a much stronger word than assurance, it conveys substance, as in your past experience backs up your confidence.
The way it is phrased, looks like a pre computed model confronted to real data.
So... our current AIs except we have incremental continuous training (accumulated experience)?
And dreams are simulation-based training to make life easier, decision-making more efficient?
> The same brain networks that normally help us imagine shortcuts or possibilities can, when disrupted, trap us in intrusive memories or hallucinations.
There is a fine line between this an wisdom. The Default Mode Network (DMN) is the brain's "simulation machine". When you're not focused on a specific task, the DMN fires up, allowing you to daydream, remember the past, plan for the future, and contemplate others' perspectives.
Wisdom is not about turning the machine off; it's about becoming the director of the movie it's playing. A creative genius envisioning a new world and a person trapped in a state of torment isn't the hardware, but the learned software of regulation, awareness, and perspective.
Wisdom is the process of learning to aim this incredible, imaginative power toward flourishing instead of suffering. Saying "trap us in intrusive memories or hallucinations" is the negative side where there is also a positive side to it all.
Wisdom is an arbitrary concept. The drive to avoid suffering is built from sensory and affective affinities and networks funnlled into the cog-mapping motor systems. Calling this wisdom is simply a simplistic narrative.
>A creative genius envisioning a new world and a person trapped in a state of torment isn't the hardware, but the learned software of regulation, awareness, and perspective.
No, it's hardware. There is no amount of 'wisdom' bootstraps pulling that will make you not schizophrenic.
My theory is that this darting is the mechanism of consciousness. We look inward and outward in a loop, which generates the perception of being conscious in a similar way to how sequential frames of film create the illusion of motion. That "persistence of vision" is like the illusion of persistent, continuous consciousness created by the inward-outward regard sequence. Consciousness is a simple algorithm: look at the world, then look at the self to evaluate its reaction to the world. Then repeat.
It's a mechanism of intelligence, not consciousness. Intel is built up from path-integration, short-cuts, vicarious trial and error that begins in very tiny local areas and expands to landmark and non-landmark navigation. This switching between vision and hippocampus has always been theorized about as the fundamental sharp wave ripple threshold of how intelligence is built as most mammals can do this, so it's not the "algorithm of consciousness".
But why does that feel like anything? I could write a program that concurrently processes its visual input and its internal model. I don't think it would be conscious, unless everything in the universe is conscious (a possibility I can't, admittedly, discount).
Interesting idea called transparent self model by Thomas Metzinger, author of The Ego Tunnel where he explains it further.
The gist from my memory of 15+ years ago is that the brain needs to model the world and then itself within the world, creating a model that is transparent to itself, situated in the world.
From what...my friend say, this becomes evident with LSD.
They said it clearly amplified the internal part of some visual perception loop, in fairly straightforward ways. For example, intentionally trying to see something as it wasn't (like a shadow as a snake) would make it be seen that way (the shadow would take on a clear snake appearance, and even move a bit).
Some simple examples are all the face optical illusions (Thatcher, reverse mask, etc), that show our perception of a face is in no way direct.
This takes me to Zen and the Art of Motorcycle Maintenance. Your physical experience of something has to be analysed in accordance with your mental model of it in order to attain a diagnosis (in the book it was a motorcycle engine).
My take on this is, especially in regard to debugging IT issues, is that you have to constantly verify and update your mental model (check your premises!) in order to better weed out problems.
There was a neural net paper like this that generated a lot of discussion on HN, but that I haven't been able to find since (I probably downloaded it, but that teaches me to always remember to use Zotero because academic paper filenames are terrible.)
It was about replacing backprop with a mechanism that checked outcomes against predictions, and just adjusted parameters that deviated from the predictions rather than the entire path. It wasn't suitable for digital machines (because it isn't any more efficient on digital machines) but it worked on analog models. If anybody remembers this, I'd appreciate the link.
I might be garbling the paper because it's from memory and I'm not an expert, but hopefully it's recognizable.
It would be interesting to move beyond rats and into humans binned via navigating their local area through understanding of the street networks independent of any tooling, and those that can't get down the street without mapping software telling them what to do.
Anecdotally it is striking to see the contrast as a member of the former group talking to people of the latter. They have truly no idea where places are or how close they are to other places. It is like these network connections aren't being made by them at all. No sense of scale either of how large a place is or how far away another place might be. I imagine this dependency on turn by turn navigation with no spatial awareness leads to quite different outcomes in terms of modes of thinking.
I mean, when I think about going to a place I am constructing a mental map of the actual city map. I am considering geography, cardinal directions, major corridors and their connectivity along the route, rough estimates of distance, etc. My CPUs are being used no doubt. Others though it is like a blankness in that wake. CPUs idle. Follow the arrows. Who knows where north is? What is a mile?
This is something that I've wondered about when it comes to things like self driving cars, and the difference between good and bad drivers.
When I'm driving I'm constantly making predictions about the future state of the highway and acting on that. For example before most people change lanes, even without using a signal they'll look and slightly move the car in that direction, up to a full second before they actually do it. Or I see two cars that are going to end up in a conflict state (trying to take the same location on the highway) so I pivot away from them and the recovery they will have to make.
Self driving cars for all I know are reactionary. They can't pick up on these things beforehand at this time and preemptively put them self in a safer position. Bad/distracted/unaware drivers are not only reactionary, they'll have a much slower reaction time than a self driving car.
The more you drive the more you notice things. After a few years in the taxi, while going through the tunnel in central Phoenix, I pointed at the cars in the far-left lanes of the freeway and said to my passenger, "you see that car right there? It's going to change into the next lane, and that other guy is going to have to slam on his brakes." My passenger was amazed when exactly this happened seconds later.
Yea but what are you "doing" when you're imagining stuff into the future - you're running some sort of prediction based on the patterns you see in the present. You're planning out moves before you execute them. The biggest difference is accuracy and breadth of simultaneous tracking, and it is relatively difficult to tell when its a good time to slow react or fast react.
A curious thing happened with the link of the article :
Clicking on it brought me to a 404. I found afterwards that the "correct" article does not mention "darting" but "flickering", even in the url. Why are some people mentionning "darting" when it appears nowhere in the text except on HN ?
Am I having a mandela effect ? /S
34 comments
[ 1.9 ms ] story [ 50.0 ms ] thread> Before the rats encountered the detour, the research team observed that their brains were already firing in patterns that seemed to "imagine" alternate unfamiliar mental routes while they slept. When the researchers compared these sleep patterns to the neural activity during the actual detour, some of them matched.
> “What was surprising was that the rats' brains were already prepared for this novel detour before they ever encountered it,”
Suppose further that all events are a draw of type 1, 2, 3, or 4, and that our memory kept a count and updated the distribution - it is essentially a frequency distribution.
When we encounter a stimulus, we have to (1) recognize it and (2) assign a reward valence to it. If we only ever observed '3', the distribution would become very peaked. Correspondingly, this suggests that we would recognize '3' events faster and be better at assigning a reward valence to those events.
Then if we ever encounter a non-3 event, we would recognize it more slowly - it is well-established that recognition is tied to encounter frequency - and do a poorer job assigning reward valence to it. Together this means that we would do a bad job selecting the appropriate response.
Perhaps this scenario-based dreaming keeps us (and rats) primed so we're not flat-footed in new scenarios.
The question then becomes - if these scenarios are purely imagined, where are they being sampled from? If we never observe 1, 2, and 4...how do we know that these are the true list of alternative scenarios?
https://kemendo.com/Deja-Vu-Experiment.html
I think it also supports my three loops hypothesis as well:
https://kemendo.com/ThreeLoops.html
In effect, my position is that biological systems maintain a synchronized processing pipeline: where the hippocampal prediction system operates slightly “ahead” of sensory processing, like a cache buffer.
If the processing gets “behind” the sensory input then you feel like you’re accessing memory because the electrical signal is reaching memory and sensory distribution simultaneously or slightly lagging.
So it means you’re constantly switching between your world map and the input and comparing them just to stabilize a “linear” experience - something which is a necessity for corporeal prediction and reaction.
Wondering if you have any ideas on this, which can be quite jarring when it happens?
You are thinking about something, and then walk through a doorway into another room and suddenly completely lose track of what you were thinking.
The closest idea I've seen for that is: Jeff Hawkins in his Thousand Brain Theory of Intelligence made a statement that learning is a function of navigation and the world models we construct are set in the context of location we create them.
--------
Edit: Just read your piece on Faith: "Faith, as it’s traditionally understood, is trivial bullshit compared to the towering, unseen faith we place in the empirical all day everyday."
Absolutely correct, and the traditional understanding of Hebrews 11:1 I don't believe reflects what the author (supposedly Paul) was trying to convey.
Ἔστιν δὲ πίστις ἐλπιζομένων ὑπόστασις, πραγμάτων ἔλεγχος οὐ βλεπομένων
πίστις: Pistis can be translated as confidence, as in: I'm confident this chair won't collapse when I sit on on it. Much stronger than belief or faith.
ὑπόστασις: Hupostasis is also a much stronger word than assurance, it conveys substance, as in your past experience backs up your confidence.
And dreams are simulation-based training to make life easier, decision-making more efficient?
What kind of next level machinery is this?! ;D
There is a fine line between this an wisdom. The Default Mode Network (DMN) is the brain's "simulation machine". When you're not focused on a specific task, the DMN fires up, allowing you to daydream, remember the past, plan for the future, and contemplate others' perspectives.
Wisdom is not about turning the machine off; it's about becoming the director of the movie it's playing. A creative genius envisioning a new world and a person trapped in a state of torment isn't the hardware, but the learned software of regulation, awareness, and perspective.
Wisdom is the process of learning to aim this incredible, imaginative power toward flourishing instead of suffering. Saying "trap us in intrusive memories or hallucinations" is the negative side where there is also a positive side to it all.
No, it's hardware. There is no amount of 'wisdom' bootstraps pulling that will make you not schizophrenic.
The gist from my memory of 15+ years ago is that the brain needs to model the world and then itself within the world, creating a model that is transparent to itself, situated in the world.
They said it clearly amplified the internal part of some visual perception loop, in fairly straightforward ways. For example, intentionally trying to see something as it wasn't (like a shadow as a snake) would make it be seen that way (the shadow would take on a clear snake appearance, and even move a bit).
Some simple examples are all the face optical illusions (Thatcher, reverse mask, etc), that show our perception of a face is in no way direct.
Consciousness could still be the self reaction to this sub-conscious predictive/generative function.
Who's doing the looking?
I often find myself lost in my mental maps in daily life (Living inside my head) unless I'm in a nice novel environment. Meditation helps, however.
My take on this is, especially in regard to debugging IT issues, is that you have to constantly verify and update your mental model (check your premises!) in order to better weed out problems.
It was about replacing backprop with a mechanism that checked outcomes against predictions, and just adjusted parameters that deviated from the predictions rather than the entire path. It wasn't suitable for digital machines (because it isn't any more efficient on digital machines) but it worked on analog models. If anybody remembers this, I'd appreciate the link.
I might be garbling the paper because it's from memory and I'm not an expert, but hopefully it's recognizable.
Anecdotally it is striking to see the contrast as a member of the former group talking to people of the latter. They have truly no idea where places are or how close they are to other places. It is like these network connections aren't being made by them at all. No sense of scale either of how large a place is or how far away another place might be. I imagine this dependency on turn by turn navigation with no spatial awareness leads to quite different outcomes in terms of modes of thinking.
I mean, when I think about going to a place I am constructing a mental map of the actual city map. I am considering geography, cardinal directions, major corridors and their connectivity along the route, rough estimates of distance, etc. My CPUs are being used no doubt. Others though it is like a blankness in that wake. CPUs idle. Follow the arrows. Who knows where north is? What is a mile?
When I'm driving I'm constantly making predictions about the future state of the highway and acting on that. For example before most people change lanes, even without using a signal they'll look and slightly move the car in that direction, up to a full second before they actually do it. Or I see two cars that are going to end up in a conflict state (trying to take the same location on the highway) so I pivot away from them and the recovery they will have to make.
Self driving cars for all I know are reactionary. They can't pick up on these things beforehand at this time and preemptively put them self in a safer position. Bad/distracted/unaware drivers are not only reactionary, they'll have a much slower reaction time than a self driving car.
Clicking on it brought me to a 404. I found afterwards that the "correct" article does not mention "darting" but "flickering", even in the url. Why are some people mentionning "darting" when it appears nowhere in the text except on HN ? Am I having a mandela effect ? /S