Live data from Hacker News

New theory proposes ‘forgetting’ is a form of learning

tcd.ie

31–40 of 75 posts

Re: New theory proposes ‘forgetting’ is a form of learning

#31
post #9

Earlier quoted context omitted.

An old mentor of mine once said some thing I found very profound: “Learning is like eating; you can choose what you consume. But knowing is like digesting; it will happen regardless, and in its own time. You only have control over eating, so eat well and frequently” There’s really a little leap of faith any time we want to learn something, because we don’t actually control how or when it fits into our brain or body,…

I frequently find myself contributing to conversations with knowledge that I don't recall specifically learning. Once in a while, I think "Wow, how did I come up with something that insightful? When did I ever learn that?" It seems there's a background activity in the mind that is actually synthesizing a wide variety of inputs from our environment. As in what I read, movies I watch, meetings I sit in on, Tweets that…

I think this is what happens during sleep (a part of it).

There have been times where I've been unable to "fall" asleep, but stayed in my bed, completely unaware of my surroundings or anything else except what's going on inside my mind.

It feels like my thoughts were starting from a single point, then randomly branching off into multiple directions (one after the brisk "completion" of the other), and each path had a certain "feel" to it -- as if my brain was testing things it had "learned" or "inferred" (and the mental models it had collected), or even things it wasn't too certain about or even guessing -- simply to see how it would "feel" when thought.

It's like your mind has a complex array of filters/gates that decide how an input will be processed -- and these filters are constantly changing, rearranging, and growing more complex as you grow; so during sleep it feels like the mind is in a loop, constantly throwing very simple things that have collected in memory, into the "mix"/filters to see what happens (will those filters adjust, a la conditioning and adjusting weights like in a neural net? Will some new and palpable way of looking at things be found? Are some thoughts/paths no longer needed? Should other thoughts/paths be prioritized?).

I think our minds hold onto all the inputs we've gotten during the day, and then process them at night, is what I'm trying to say. I imagine this is one of the big reasons why we sleep so much as we do (compared to hunger-gatherers, and primitive tribes): we have so much sensory input that gets collected during the day (including our thoughts, albeit the thinking we do during the day serves the same end: forcing our heads to deal with new information to figure out what to do with it).

And then you have meditation, that basically forces you to clean out the "volatile memory," so it doesn't interefere throughout the day. Basically telling your mind, "I don't need to incorporate this shit into my thinking. It's not vital. Dump the memory onto the disk, and we'll deal with it later, but not now!"

I've noticed that sleep completely changes my train of thought.

On any given day, I'll have a certain "mindset" or overarching "feeling" for the day. If I stay up for a day or two, that feeling will remain. And no matter what I do, that feeling rarely changes throughout the day. BUT, when I go to sleep, it's completely different. I always wake up with a new "train" of thought. Like when I was awake, the train was going on a certain rail line, but during sleep it was moved to a different one.

Re: New theory proposes ‘forgetting’ is a form of learning

#32
This reminds me of a funny passage in Jin Yong’s wuxia novel “The Heaven Sword and Dragon Saber”, where Zhang Wuji, the protagonist, learns taichi sword from his grand master Zhang Sanfeng.

Zhang Sanfeng demonstrated the whole taichi sword form to Zhang Wuji repeatedly, and after each demonstration, asked Zhang Wuji “how much have you forgotten?”

Zhang Wuji instantly got it and each time replied that he has forgotten more and more, while Zhang Wuji’s subordinates watched on the side and became more and more worried that their hierarch was forgetting his martial arts.

Re: New theory proposes ‘forgetting’ is a form of learning

#34
post #28

My ex is a linguist, and she used to tell me about the “learning by forgetting” theory of the “phonological inventory” — which is a fancy way of saying “the sounds you can distinguish and produce.” The idea was that children were born able to distinguish all possible sounds, and over time forget the ones they are not exposed to. Interestingly she was a syntactician by training and would tell me that there are many pe…

Your comment, along with the article, confirm my biased beliefs that humans are less pattern matching machines than noise filtering machines. Pattern recognition is less about connecting dots than it is distinguishing and purging the unconnected. When that process goes awry, and we can’t discern disconnections, things like psychosis arise.

Tangentially, it stimulates a connection to another though that sometimes flares up concerning evolution and specialization.

> If memories were gained in circumstances that are not wholly relevant to the current environment, forgetting them can be a positive change that improves our wellbeing.

If as a species we evolved and specialized for a certain environment, and now we have changed the environment so much that those specializations are a hindrance, would a past proto human be better at adapting to today’s world, if brought up by modern society, due to better adaptability? Biodiversity is the blank slate beginning, and selection is the forgetting.

Re: New theory proposes ‘forgetting’ is a form of learning

#35
Forgetting is similar to regularization mechanisms in deep learning like dropout. You tend to forget details, and this makes your mental models simpler over time. The memories that get refreshed are the details that actually matter and keep tripping you up

Re: New theory proposes ‘forgetting’ is a form of learning

#36

I wonder if this is related to the effect of being better at something after not doing it for a while. When it comes to a lot of things I do repetitively, I notice that there comes a point where either my skill doesn't improve or I actually get worse at what I'm doing. Then I'll take a day or more off, and when I come back suddenly I'm a total boss. The brain probably needs rest time to throw out the bad input and th…

That’s the theory behind Spaced Repetition Learning, and it doesn’t surprise me that it applies to physical skills.

Re: New theory proposes ‘forgetting’ is a form of learning

#37
post #34
post #28

My ex is a linguist, and she used to tell me about the “learning by forgetting” theory of the “phonological inventory” — which is a fancy way of saying “the sounds you can distinguish and produce.” The idea was that children were born able to distinguish all possible sounds, and over time forget the ones they are not exposed to. Interestingly she was a syntactician by training and would tell me that there are many pe…

Your comment, along with the article, confirm my biased beliefs that humans are less pattern matching machines than noise filtering machines. Pattern recognition is less about connecting dots than it is distinguishing and purging the unconnected. When that process goes awry, and we can’t discern disconnections, things like psychosis arise. Tangentially, it stimulates a connection to another though that sometimes flar…

What also does this mean for psychedelics? Could they "reopen" the filters that were earlier preventing new signals from being found?

Something like switching from a digital radio dial (every .2 MHz) to an analog dial (everything in between) - you can find something you had missed before.

Re: New theory proposes ‘forgetting’ is a form of learning

#38
post #37
post #34

Earlier quoted context omitted.

Your comment, along with the article, confirm my biased beliefs that humans are less pattern matching machines than noise filtering machines. Pattern recognition is less about connecting dots than it is distinguishing and purging the unconnected. When that process goes awry, and we can’t discern disconnections, things like psychosis arise. Tangentially, it stimulates a connection to another though that sometimes flar…

What also does this mean for psychedelics? Could they "reopen" the filters that were earlier preventing new signals from being found? Something like switching from a digital radio dial (every .2 MHz) to an analog dial (everything in between) - you can find something you had missed before.

It does make you wonder if we “learn” that things are NOT connected, and can chemically stimulate reevaluation of such learned non-patterns, helping break preconceived notions.

Like if after being taught oil and water don’t mix, you could unlock the ability to question such belief.

Re: New theory proposes ‘forgetting’ is a form of learning

#39
post #28

My ex is a linguist, and she used to tell me about the “learning by forgetting” theory of the “phonological inventory” — which is a fancy way of saying “the sounds you can distinguish and produce.” The idea was that children were born able to distinguish all possible sounds, and over time forget the ones they are not exposed to. Interestingly she was a syntactician by training and would tell me that there are many pe…

I don't know how much of this has to do with forgetting so much as attaching significance. One thing I realized I was able to do for much of my life (perhaps not quite as well as I used to) is keep the concept of sound and word separate. Of course we can do this for sound effects. I've been able to maintain this for most word sounds. This was very apparent when learning foreign languages, reproducing the sounds was the easy part--just make the noise you hear.
Post reply on HN