When we start to lose our minds the unguarded core functionality of the system becomes more and more exposed. And as it becomes more exposed it starts to look like chatGPT.
Because as much as we don't like to admit it, maybe chatGPT does model a core aspect of human cognition.
Every new technology is always used as an analogy to describe how the brain functions. The brain is "a series of pipes", "a series of cogs", "an eletro-mechanical machine", "a computer", "a series of programs", "GANs", now "an LLM". I get a feeling it's as accurate a comparison with LLMs than with everything before.
While I’m not saying positively that LLM are the way a brain works, it is worth noticing that the meme argument structure you are using is very often absurd because Turing Machines can be made out of all the things people usually list. People trying to accuse of category shift are very strongly overestimating the strength of that argument in much the same way that people commenting that people who thinks money were cents, nickels, pesos, and dollars are not quite managing to articulate true dissent by pointing out the letters are different.
> Well it's definitely true that the human brain is Turing complete.
That is definitely false. "In computability theory, a system of data-manipulation rules... is said to be Turing-complete... if it can be used to simulate any Turing machine." [0] A Turing machine has infinite memory; the human brain does not.
Describing the brain as "a computer" can't really be wrong because computers are universal. But it's also not necessarily all that useful, because, well, computers are universal.
obviously he's talking about the part of the brain that matters. Nobody is comparing anything to say the part of the brain that controls your heart rate. You're just being pedantic here.
Comments like this should be banned from HN. It's used deliberately and manipulatively to characterize the other person as "unhinged." It does not promote useful discussion and is a clear violation of the rules.
I just have a lot to say, and I like to use capital letters to express it. It doesn't justify tactics like this of "chill bro". If you have a response, just say it.
Look at the other end of life. Every parent knows about the very long, difficult climb, from tykes blathering noises and words, to really intelligent speech - sense of past/present/future, telling the truth, keeping promises, understanding the world, etc.
Good point. Well illustrated in Kubrick's 2001 Space Odyssey, when Hal's memory modules are removed one by one, and his speech patterns revert to increasingly infantilized forms.
Or when people are carrying out post-hypnotic suggestions. They will come up with plausible reasons for their behaviour if asked.
Scott McCloud talks about this in "Understanding Comics" as well, where he calls it "closure": the human mind hallucinates information to make the inputs it receives coherent. The neat thing IMO is that it will do this with "internal" inputs as well.
As one carer put it, "You can't fix crazy." There's basically just no reasoning with them.
I saw an interesting vid on YouTube about a man who was relatively young, IIRC. He was in the C-suite of of a biotech company, but he suffered from some kind of dementia. I don't think it was Alzheimer's though. He slept in a separate room from his wife. He complained to his wife that his bed headboard was "rusty". As far as I could tell it was just an ornately carved headboard, and he somehow equated some of the carvings as rust.
There were other incidences like he would go through his wardrobe at like 2am in the morning. He appeared to want to sort out his shoes so that he could put them in his car. So there was something of a mess on his bed with wardrobe contents strewn out on it. He wasn't achieving much of anything, and of course there was no reason for him to sort out his shoes. Especially not at 2am.
It seems as though dementia sufferers do have some kind of "logic" to their behaviour, but it appears to be some kind of random logic without any discriminative faculties. And they fixate hard on this rationale too, making it particularly frustrating for those around them.
What's interesting, and actually quite well documented, is something called "terminal lucidity". This is when as someone actually gains mental clarity for a period of time some time before their death. The specific time periods vary with the sufferer. This is baffling medically, of course, because if a person's mental function is gone due to brain decay, then how can they get it back? And why does it foreshadow death?
> There were other incidences like he would go through his wardrobe at like 2am in the morning.
Huh, odd. When I was single-digit years old I used to get sleepwalking night terrors. I don't remember the details about most of them, but this sounds exactly like something I could do then.
My brain would fixate on a deep feeling that something has gone terribly wrong, and only I could fix it by performing something nonsensical, like figuring out how many of my bed would be needed to cover the surface of the earth, and I would start to move it around my room and see if I could tile it neatly with itself in some arrangement (because here is as good a place as any to start counting.)
Aside from these weird ideas I was mostly lucid, though. My mother worried I'd fall down the stairs but I never had reason to suspect I'd have trouble navigating the stairs. My literal perspective was sometimes distorted (things seemed to get smaller faster than a regular perspective transform) but I was careful to account for that.
I've got something similar, though I've so far not left my bed due to it. I'll half-wake up in the middle of the night and have some random task around my bed that absolutely must be done. I'll usually wake up the next morning remembering everything with perfect clarity but deep confusion as to wtf I was thinking.
e.g. I've woken up with a need to remove by bedside lamp from the wall, which I managed to do without damaging anything. Still no clue why the heck I felt the need to do that but I remember doing it perfectly.
I often feel as though the ideas of “logic” or “rationality” are means to irrational ends. Or, if we ask ourselves “why” to our desires a couple times, we don’t necessarily have a rational motivation to have relationships, stay healthy, or try new things; we simply intrinsically want them.
Given that, I often wonder if the behaviours in those moments are because the irrational desires have become very different for some of us. I think it might explain why people will seem so methodical or sure about carrying those ideas out.
Knowing a close friend who lost his father to this, where his father spent many years in care and was always talking crazy.
He told me that in the week of his fathers death they were able to talk with each other as adults for the first time in many years. It was surreal because he had long thought the man was lost.
I believe the potential cause is an inflammatory response to an unknown agent (bacteria, virus or even our own cells). I wonder if terminal lucidity occurs as the immune system shuts down resulting in temporary reduction in brain inflammation.
> What's interesting, and actually quite well documented, is something called "terminal lucidity". This is when as someone actually gains mental clarity for a period of time some time before their death. The specific time periods vary with the sufferer. This is baffling medically, of course, because if a person's mental function is gone due to brain decay, then how can they get it back? And why does it foreshadow death?
Huh. This is interesting. In multiple intelligent agent designs I know how to create this improved lucidity would come as a consequence of estimation of values not needing to be short-circuited due to intractability. One design has this happen because the limit of the monte carlo rollouts converges to the accurate result due to reduced computation time. The other design has this happen because it stops taking the limit with respect to a value near but below one such that the limit of future expectation gets discounted to nothing. A final design, which is far less optimistic with respect to recovery from this, just has literal cached values for end of life scenarios.
It would foreshadow death because the world modeling suggests that death is happening and is still functional. It would produce lucidity because the erroneous heuristic functions wouldn't be relied upon as heavily as the world model's actual results. Or in the final case... it would have been precomputed before reaching that scenario and didn't decay but it was cache lookups.
Do you happen to know if putting the non-lucid people in life threatening and novel such that they would not have anticipated them situations - for example, standing near the edge of a cliff on a windy day - tend to increase lucidity?
Now that I'm thinking about this I notice that free soloists have been claiming that they gain increased lucidity during their climbs. So my model that predicts non-lucidity gaining lucidity actually successfully predicts an improved lucidity in free soloists too, which is not something I realized before I started thinking about this.
I am not a neuroscientist so I don't know how biologically plausible this is as an explanation goes.
> The “interpreter” is an unconscious process responsible for sweeping inconsistencies and confusion under the rug.
I think it was through Kahneman I first read about this. With split-brain patients, you can trick one of their brain halves to perform an action in view of the other half. If you then ask the other half "why did you just do that?" it will come up with the weirdest of explanations, instantly. It's quite remarkable.
There's an argument that this interpreter is responsible for explaining all our decisions - it's just that mostly our decisions are more easily make sense of (or at least, seem to make sense to others). That it, our consciousness is not running the show, but is just along for the ride.
Maybe it's somewhat like chatGPT, in that if it has the correct information to explain an action it'll give that explanation, but just confabulate otherwise.
I'm not sure that chatGPT actually has the correct information - it just responds to prompts for explanation as it does for any prompt. Confabulating. If its guess sounds correct, we think it is.
Since it's drawing on the same information both times, it probably often is correct.
50 comments
[ 3.6 ms ] story [ 135 ms ] threadWhen we start to lose our minds the unguarded core functionality of the system becomes more and more exposed. And as it becomes more exposed it starts to look like chatGPT.
Because as much as we don't like to admit it, maybe chatGPT does model a core aspect of human cognition.
(Trivial approach: It seems fairly unlikely that Alan Turing would not have been able to simulate Turing machine inside his head. :-P )
That is definitely false. "In computability theory, a system of data-manipulation rules... is said to be Turing-complete... if it can be used to simulate any Turing machine." [0] A Turing machine has infinite memory; the human brain does not.
[0] https://en.wikipedia.org/wiki/Turing_completeness
Of course, the technically correct definition would limit the usefulness of the term to just the field of mathematics.
Colloquially, people also say that systems that are obviously finite (due to having to exist in the real world) are also Turing complete. [1]
[1] https://en.wikipedia.org/wiki/Turing_completeness#Non-mathem... (subsection of article you quoted)
That is, a human brain (or mind) is more like an LLM than like clockwork, etc., but it still isn't all that much like an LLM.
We can only communicate technology through terms we collectively understand.
I just have a lot to say, and I like to use capital letters to express it. It doesn't justify tactics like this of "chill bro". If you have a response, just say it.
Look at the other end of life. Every parent knows about the very long, difficult climb, from tykes blathering noises and words, to really intelligent speech - sense of past/present/future, telling the truth, keeping promises, understanding the world, etc.
Scott McCloud talks about this in "Understanding Comics" as well, where he calls it "closure": the human mind hallucinates information to make the inputs it receives coherent. The neat thing IMO is that it will do this with "internal" inputs as well.
I saw an interesting vid on YouTube about a man who was relatively young, IIRC. He was in the C-suite of of a biotech company, but he suffered from some kind of dementia. I don't think it was Alzheimer's though. He slept in a separate room from his wife. He complained to his wife that his bed headboard was "rusty". As far as I could tell it was just an ornately carved headboard, and he somehow equated some of the carvings as rust.
There were other incidences like he would go through his wardrobe at like 2am in the morning. He appeared to want to sort out his shoes so that he could put them in his car. So there was something of a mess on his bed with wardrobe contents strewn out on it. He wasn't achieving much of anything, and of course there was no reason for him to sort out his shoes. Especially not at 2am.
It seems as though dementia sufferers do have some kind of "logic" to their behaviour, but it appears to be some kind of random logic without any discriminative faculties. And they fixate hard on this rationale too, making it particularly frustrating for those around them.
What's interesting, and actually quite well documented, is something called "terminal lucidity". This is when as someone actually gains mental clarity for a period of time some time before their death. The specific time periods vary with the sufferer. This is baffling medically, of course, because if a person's mental function is gone due to brain decay, then how can they get it back? And why does it foreshadow death?
Huh, odd. When I was single-digit years old I used to get sleepwalking night terrors. I don't remember the details about most of them, but this sounds exactly like something I could do then.
My brain would fixate on a deep feeling that something has gone terribly wrong, and only I could fix it by performing something nonsensical, like figuring out how many of my bed would be needed to cover the surface of the earth, and I would start to move it around my room and see if I could tile it neatly with itself in some arrangement (because here is as good a place as any to start counting.)
Aside from these weird ideas I was mostly lucid, though. My mother worried I'd fall down the stairs but I never had reason to suspect I'd have trouble navigating the stairs. My literal perspective was sometimes distorted (things seemed to get smaller faster than a regular perspective transform) but I was careful to account for that.
e.g. I've woken up with a need to remove by bedside lamp from the wall, which I managed to do without damaging anything. Still no clue why the heck I felt the need to do that but I remember doing it perfectly.
Given that, I often wonder if the behaviours in those moments are because the irrational desires have become very different for some of us. I think it might explain why people will seem so methodical or sure about carrying those ideas out.
He told me that in the week of his fathers death they were able to talk with each other as adults for the first time in many years. It was surreal because he had long thought the man was lost.
I believe the potential cause is an inflammatory response to an unknown agent (bacteria, virus or even our own cells). I wonder if terminal lucidity occurs as the immune system shuts down resulting in temporary reduction in brain inflammation.
Huh. This is interesting. In multiple intelligent agent designs I know how to create this improved lucidity would come as a consequence of estimation of values not needing to be short-circuited due to intractability. One design has this happen because the limit of the monte carlo rollouts converges to the accurate result due to reduced computation time. The other design has this happen because it stops taking the limit with respect to a value near but below one such that the limit of future expectation gets discounted to nothing. A final design, which is far less optimistic with respect to recovery from this, just has literal cached values for end of life scenarios.
It would foreshadow death because the world modeling suggests that death is happening and is still functional. It would produce lucidity because the erroneous heuristic functions wouldn't be relied upon as heavily as the world model's actual results. Or in the final case... it would have been precomputed before reaching that scenario and didn't decay but it was cache lookups.
Do you happen to know if putting the non-lucid people in life threatening and novel such that they would not have anticipated them situations - for example, standing near the edge of a cliff on a windy day - tend to increase lucidity?
Now that I'm thinking about this I notice that free soloists have been claiming that they gain increased lucidity during their climbs. So my model that predicts non-lucidity gaining lucidity actually successfully predicts an improved lucidity in free soloists too, which is not something I realized before I started thinking about this.
I am not a neuroscientist so I don't know how biologically plausible this is as an explanation goes.
I think it was through Kahneman I first read about this. With split-brain patients, you can trick one of their brain halves to perform an action in view of the other half. If you then ask the other half "why did you just do that?" it will come up with the weirdest of explanations, instantly. It's quite remarkable.
Since it's drawing on the same information both times, it probably often is correct.
In other words, like us.