What Rust helps with is memory safety, which is stuff like dangling pointers and use-after-free. Memory leaks are not a safety issue, they're an efficiency issue.
> Why are people so crazy about Rust?
The whole memory safety thing isn't actually the reason so many people love Rust. In the Rust 2024 community survey, 82% of users agreed that "Rust allows us to build relatively correct and bug free software." Memory safety is part of that, but there's a lot more to it.
So, for example, while Rust does not guarantee the lack of memory leaks, it does make them relatively hard to produce by accident in the first place. This is due to the intersection of a few different parts of Rust's overall design.
How do you know that is where you got the memory? Early birthdays can be tricky because it is common for families to record them on video and if you saw such a video a few years later that could create a memory that you would think was from the birthday itself.
Personally I don't trust most memories I have from before around 4 because my parents had around 20 reels of home movies from that time and I know that I saw several of those movies a few years later.
I fell off a countertop and got a concussion when I was 2.5, it's a very vivid and painful memory. I also moved to a different state when I was 5.
I have a very clear delineation of before and after the age of 5, and I have hundreds of memories from before the move. Most are rather intense events, but I even remember the layout of my house, construction sites, my daycare, neighbors, friends houses, holidays, first bike ride, first lizard I caught, my pets, even some dreams I had, etc...
I didn't have a good idea of time yet, but I can retroactively tell when things occurred based on facts I later learned, like when certain neighbors moved, even my sister was born, etc...
We didnt have anything to film with & there are no pictures from that day.
I remember only a specific moment: we went to pick up my grandpa and i recall the shirt he was wearing very precisely.
I can’t guarantee that the memory is genuine, but it is very specific and I had to reconstruct my age from discussing the memory with my mom.
Ontologically, you can not say that definitively, you can only say that op's claim is highly unlikely. Only OP knows what OP knows and no amount of theory can disprove it, nor can any theory lay greater claim to ontological accuracy than he can.
I'd go as far as saying, memory fabrication happens all the time - we recompute our memories when we reference them, and that result is affected by all the other experiences and memories we accumulated between subsequent recalls. Or, in other words, humans always confabulate (in the exact same sense as LLMs "always hallucinate", and I'm invoking this comparison on purpose) - the difference between "correct" and "false" memory is a matter of degree.
One side says, "I remember this thing from when I was 1 year old".
Another says, "You cannot remember it, you must have reconstructed it from being told or from pictures."
The memory itself can certainly be falsified: if you remember your mom using an iPhone in 1998, then that memory is false. If you remember talking to your uncle but he died before you were born, the memory is false.
How do we know if the origin of the "memory" is the person's actual experience, or their imagination of it based on descriptions or pictures?
Well if it's an event that had pictures taken of it, if it's the kind of event which grown-ups talk about, then the simplest explanation is that it was reconstructed.
But my son, almost 5, is always coming out with random things that happened when he was 1 or 2, which are absolutely of no importance to us, and which we would have forgotten long ago if he didn't keep remembering them. Once, for example, my wife put a pair of his shoes on top of the car while she put him in the car seat, forgot they were there, and drove off -- obviously at some point they fell of and were completely lost. Every six months or so my son talks about that incident completely unprompted -- something of absolutely no significance to us, but obviously something that struck him. And about the time I lost my temper and sprayed him in the face with water -- definitely not a memory I'm eager to revisit and bring up.
Someone dogmatic person may still say, "You must be bringing that up yourself somehow." That is certainly an unfalsifiable assertion -- there's no way, other than my assertion and probability, to prove that I'm haven't talking about those shoes on a regular basis. But I think any person with an open mind is likely to agree that "he's remembering them" is a more likely scenario than "the grownups are talking about those lost shoes all the time".
Probably these "revisitations" are his brain's way of refreshing the memory as his brain grows, which probably means the memory as they exist in his brain will have been shifted over the years. But that's still an original memory -- that's how adult memories work as well.
If at any point he misses a "refresh", they'll probably be gone; but the memories he manages to do this for will probably stay with him into adulthood. If at any point he "misses" a refresh, they'll probably be gone forever.
The flip side of this, of course, is that many of the memories we think we have as adults are heavily edited too. A few years ago someone recounted to me an emotional conversation that they'd had with someone on their deathbed (as an adult, only a few years prior to that). As it happens, I was in the other room when this conversation happened, and my memory of that conversation was significantly different. Obviously at least one of us is "remembering" something untrue. As I had little "skin in the game" about the content of the conversation, and the memory this person has seems to me very consistent with this person's narrative about their life, I'm inclined to think my memory is more accurate. But who can tell.
This matches my experience also. Kids can "juggle" some of these extremely early memories into more permanent memories but the vast majority are dropped. It's only because of the early and possible frequent recall that the memories end up winning their mythic permanence.
I had memories as a toddler that I confirmed after meeting my mother for the first time in decades a couple of years ago. Specific things about where I lived, events that had happened and how someone else in our family had died. I had another conversation with an aunt that confirmed other things, including a word from another language that I knew from when I was little but didn't realize wasn't a made up word.
Memory is fallible and we have an imperfect understanding of it, but I know for a fact from personal experience that people can remember events from very early in life if the emotional impact is deep enough.
I have the same experience confirming memories with my parents. A few years ago they finally got garbage pick up service again and mentioned it to me.(Tiny area, roughly 1,000 people, so that service was not feasible until recently.) To which I replied, "Oh yeah, it's been like, 30 years since you last had it." They asked how I knew that. "I remember you carrying me up the driveway to drop the dirty diapers in the bin." They were both surprised that I remembered that and could confirm it.
However, the time that a giant plate glass mirror fall off a wall at a department store and crashed through me when I was about two years old? No idea. My parents had to tell me about that one later in life.
I believe you. I have a few memories around my second birthday. People are often impressed with my recall of people and places. I have reasons to believe because of corroborations later in life that my memories were true.
I genuinely do have at least one memory from the age of two, because it was a completely banal event that no one told stories about (my grandmother moved to a new office, lol), nor certainly photographed, that I (because I remembered it) assumed had taken place when I was three. Turns out (crowd-sourced family chronology agreed, and documentary proof later corroborated) took place either 5 or 6 months after I turned two. (Memories differed as to the exact month, and the earliest letter we had was from the corroboratory year, but from a couple of months after when she would have moved.)
The matter came up because some time when I was in my twenties I said, apro pos of something or other, "[grandma] moved to [office] in year X", and they said "No, it was year X-1", and I said "well how come I remember [mundane sense-memory detail], if I was only two?" And they said, "yeah, [detail] is correct, but it was definitely year X-1", rinse-repeat, until they set out to prove me wrong about the year, which it turned out I was, which was a win for them, but then both I and my aunt with PhD in child-development were forced to conclude that I had a memory from when I was two.
Yeah. It blew my mind, too. I didn't think human memory worked like that, either.
Something-something only a Sith deals in absolutes. Alternatively, don't be such a jerk to that other guy. It's not really within the HN ethos.
You are way over confident. I definitely have memories from 2 onwards and I know I remember them because I remember relaying them to people at age 4 onwards, by which point my memory was very well formed.
Around 2 tends to be the earliest limits that have been confirmed.
Memory scientists don't claim that no memory persists from infancy, its just that it hasn't been confirmed, and there's lots of evidence of infants failing to remember events from months back in infant memory experimental paradigms, and highly plausible scenarios of memory reconstructions, from pictures, or from stories.
5 year later show the same two balls and say you will give the kid an ice cream if it picks the same color. Obviously give the ice cream no matter the pick ...
Heh. I've got memories from ages two and three that are pretty solid, and there were not photos or other records to prompt them. Some people really can do this. Some of my early memories were "filled in" by later experiences when I learned something that "explained" the earlier memory, and it was then reinforced, but the early memories are independent from the later ones.
My earliest memory was President Kennedy's assassination, but of course I had no idea of what a president was, life, death, etc. I was 16 months old. What I do remember was a bunch of strange people in and out of the house with the TV on and everyone very concerned, but what I remember most of that day was that it was the day I learned how to shake my crib and move it around the room while I was inside. I remember my parents not liking the fact that I was able to do that. I've never discussed that memory with them, and they're both gone now, so I never will.
I've got memories from when I barely knew how to talk, and didn't understand what the people around me were saying. My memory isn't good enough to decode what was said after I learned the language. I can clearly remember sitting in my high chair eating Alphabet cereal, and trying to make words with the letters I had not yet learned. I get sad when I think about how my parents would brush me off, probably thinking that I was too young to learn much. (My parents were both school teachers!)
Same. I'm always kind of shocked when people don't. My wife hardly remembers anything before she was 8.
We moved a couple months before I turned three and I have a bunch of memories that predate that. I remember being potty trained and lying to my parents about it so I could get a treat. I remember my dad coming home from the family farm and having injured his hand on some machinery. I remember stealing a pack of orange gum from the grocery store and chewing it all up underneath the piano in our basement. I could keep going and going.
My earliest memory was before I was two. Somewhere around 19-20 months. I know this because I had an ear infection and that's when it occurred. I'm sitting in my crib, crying due to the pain in my ear, and screaming "mom" at the door in the room. The room is mostly dark, but the door has been left open a crack to let in a little bit of light. From my vantage, looking through the slats in the crib, the door is open on the left. My mom comes in once, finally, and then leaves.
I also feel like I have an earlier one, being carried in a room which may have been my nursery, looking down on a changing table. I'm pretty sure I'm being held by my father, but I can't say for certain that this is real.
I wonder if it has to do with the snapshot timing resolution being asymptotically high when you have not yet experienced enough passage of time.
As you know we generally experience timescales logarithmically with age, i.e. your incremental experience of time is always compared in reference to dividing by the total passage of time experienced, which is why children tend to get bored much more quickly than adults, because waiting for ten minutes constitutes a much larger percentage of their current experienced life compared to that of an adult's proportion.
Since a baby/toddler has only experienced a tiny amount of time passage thus far, their tiny reference "yardstick" would result in their memory being snapshotted at an untenable timining resolution thanks to division-by-almost-zero. So perhaps the brain does some form of filtering to prevent the entirety of experienced memory being dominated by the super-early ages, or perhaps there is some equivalent of an overflow in their internal counter.
Keep in mind that the above are purely metaphorical as a functional description, and not to be treated as a literal hypothesis on the mechanistic operation of memory.
An interesting idea; although I think somewhat unlikely. This is one of those instances where I would assume the brain is doing something carefully calibrated. But it'd be easiest to just not record as many experiences for the first few years rather than do a complete flush.
My guess has always been there is some optimal approach to learning that works by developing a really basic schema (what does a person look like, which ones are mine, roughly how does this body thing work, etc) then flushing all the training data and starting again. I vaguely expect the machine learning people develop some sort of similar process where they get a lot of value lightly conditioning a model before sinking compute into doing a full train.
Basically, my wild guess is there is a lot to learn in a baby's first few experiences but the risk of mis-encoding the lessons is so high that the brain uses the data to bootstrap but then throws it out as too unreliable and starts again once it orients.
Exactly what I was going to say. Go travel through strange lands for 3 months and compare the feeling of the passage of time to 3 months doing your normal routine.
N of 1, but my experience doesn't fit with this common explanation. Big life changes, travel, having kids... none of it has done a thing to disrupt a steady progression to a point where (around age 40) a year feels about subjectively as long as maybe three months did in high school. Started being noticeable around age 25 IIRC, and time's done nothing but keep getting faster. This doesn't revert, at all, when things get shaken up, not even temporarily. It's still "you know that place we went the other day... oh, shit, that was six weeks ago".
> which is why children tend to get bored much more quickly than adults, because waiting for ten minutes constitutes a much larger percentage of their current experienced life compared to that of an adult's proportion
I think boredom is the brain's hunger more than its watch. It's about the needs of a developing brain, rather than plain time.
Right after the bulk of neurogenesis the brain goes into synaptic pruning where it strengthens some synapses and eliminates others. Developmental pruning is experience dependent and the scale of the process in childhood needs a lot of experiences to fuel it.
Those experiences are literal "food for thought". Building the body takes more food, building the brain takes more experiences.
I've always found two things interesting about memory. First, it often seems many more memories can be stored than recalled. This becomes apparent when you can't remember something to the point of being certain you've forgotten, only to later recall what you were trying to remember. (My definition of a good trivia question is one you can't answer until you're told the answer and you realize you knew all along.) And of course people who have the ability to apparently remember everything indicate memory capacity can be enormous.
Second, I've often thought I can correctly remember a scene from a movie or TV show I saw decades ago, but when I see it again I realize I got a tremendous number of details wrong, indicating that particular stored memory is very general in nature and not accurate at all.
Re trivia, I often find myself thinking "I would have definitely gotten that if it was multiple choice".
I agree with your second paragraph too. I often think of a certain tone of voice or pacing with which a line was delivered, and it ends up being completely different when I rewatch it. And yet there are still Mandela Effect people out there insisting their memories are 100% infallible and it's more likely there's a global time travel conspiracy to change a dot on a KitKat package than that they misremembered something small.
Since LLMs/AI are all the rage these days - I also found a similar behavior with LLMs.
If I ask a difficult recall question (usually some very niche medical knowledge) it will get it wrong or just hallucinate. But if I ask it in the form of a multiple choice question, it often gets it right. If memory works better when prompted with the information this way - I wonder how similar LLMs and humans actually encode memory?
Because the apparatus for remembering is emergent, derived from accumulated training data. Only when the person's brain has gpt-2 level training data is what we call memory available. Of course actual memory as in persistent storage, is there from before birth.
I figured the reason why we cannot remember such a part of our lives is because our brains grew enough that the 'memory' neuron groups wound up converted to 'network connections', the connection and this is pretty much a result of a geometric constraints. Theoretically we could have a process of the brain relocate the memories before conversion, but it was probably selected against because it would increase calorie consumption by 5%.
>There is debate between memory experts as to the role of language in infantile amnesia. Human researchers suggest memories may be limited by an inability to give language to early experiences. “But there must be something more fundamental that also plays a role because we see this same [infantile amnesia] effect in non-linguistic animals like rats,” says Prof Rick Richardson of the University of New South Wales.
That opposition doesn't quite make sense to me. If you are examining the linguistic view of it, then what we're talking about is that memory requires symbolic understanding. Not just language (a fairly sophisticated symbolic tool), but also just the ability to schematize your various sensory impressions into conceptualized objects. My earliest memory is sitting in a high chair playing with Duplos (large Legos for young children) being frustrated that I couldn't connect two pieces. That requires a few concepts such as a high chair, the tray sitting in front of me, the pieces I was playing with, and emotions like anger. Symbolic understanding isn't necessary for any of that to happen, but it does seem necessary to store it for recall decades later.
The problem with comparing this to rat memory is that I'm not convinced that rats have anywhere near the same kind of symbolic understanding, so how can they be compared to humans in this capacity? Rats can remember things, but babies can also be taught sign language starting at around 6 months, which seems much closer to rat memory (the direct linking of a perception with an emotion, unmediated by a symbolic transformation).
I had a friend who swore she had memories from before she could speak, like remembering specific events with extreme frustration about not being able to express herself.
There are people without internal monologues, and most people can also visualize something and remember scenery without words. There are also animals that (likely) don't have language that apparently have memories.
My dog sometimes buries a bone when we give him one. Later he digs it up and chews on it some more. He forms a memory of where he buried it without the use of language (I'm pretty sure).
From experience, I am skeptical of this hypothesis. Little kids will absolutely recall memories from before they knew how to speak.
Just last week, my two-year-old spied the freezer pops in storage. She pointed out back and said, "Eat on deck!" Clearly, she remembered eating freezer pops on the deck, but the last time she did that was last summer (northern hemisphere) when she didn't know how to talk at all, let alone say "deck".
That's interesting, because my earliest memory was wordplay related. Though I doubt I could have strung a sentence together at the time, there were definitely words involved. I have a few relatively early memories, but none that are totally pre-language.
I can't counterexample the first half, but my only vivid memory from being a 2-year-old is primarily visual (though it does involve surprise at the reality contrasting with a prior description in words).
Animals have no meaningful language but clearly have memory. Also the separation between thinking and language becomes quite clear as you learn another language and realize that asking what language you think in is not such an easy question to answer.
In her autobiography, Helen Keller describes memories from her childhood, before she was taught to communicate, so it would seem language isn't necessarily a prerequisite for memory formation.
Are they remembering a memory or a memory of a memory? And in the end is there any difference?
I took lots of pictures of my kids. Those pictures trigger memories, but it's unclear if they would have remembered those things without the pictures. And there's no real way to tell if those memories are real, or just narratives attached to the trigger.
All memories after the first recall is a memory of a memory. And after second recall, it's a memory of a memory of a memory. And so on. That's how the hippocampus works.
My mother was shocked when I recalled the bathroom tiles, layout and song she sang to me when I was a baby. There's no photos of the bathroom, it wasnt discussed as its refit was banal. The song she sang, she never mentioned to me past being a baby.
As an adult I have very good visual and audio memory, as well as perfect pitch. They're not as useful as they sound.
I remember a lot of my toddler years. I also remember coming out of what, looking back, feels like a dream-like state that was much simpler emotionally and cognitively but had almost a hallucinogenic feel before "waking up" as a toddler.
I believe I remember trying to say a word and being frustrated that parents wouldn't wait and gave up just as I was about to utter the sounds, or something along those lines, it's really a vague memory of having a memory at this point.
Memories aren’t an objective tape recording. Memories are the brain’s model of the world. You spend your life refining your model of the world. Your earliest memories were initial rough approximations (misinterpretations, if you will) that are later superseded by better (more predictive) approximations. So you forget (they don’t activate as predictions) the old inaccurate (less predictive) memories in favor of the newer (more predictive) memories.
If we go a bit old-school on AI and reason in the "connectionist" framing:
Let's say neural memories are encoded in some high-dimensional vector space.
And so memory recall is an associative process that entails constructing a query vector and issuing it across the neural memory space.
And the brain is constantly learning, and that learning entails some changes in the structure of the high-dimensional memory space.
And let's say that re-encoding of a neural memory happens upon recall, and only upon recall.
Then it could be that all experience is in fact stored, but because of changes due to learning, those memories become inaccessible. The machinery constructing query vectors has updated its structure enough that its encoding of those query vectors is sufficiently dissimilar from the encoding of the stored memory vectors (which use the encoding from the last recall).
Doesn't really explain why it happens universally and why this doesn't happen after other major changes in lifestyle (people who move to a radically different country don't lose all memories of their life beforehand).
Age 2: Can point to their own body parts; hold something in one hand while doing something with the other hand
Age 4: Changes behavior based on where you are; can draw a person with more than 3 distinct body parts
There's a huuuge amount of learning that happens through this period. Your brain is learning things like 3-dimensional space, temperatures exist and I don't like some of them, I-have-two-arms, things fall when dropped, I must engage my big toe to stay upright while walking, other people appear to have feelings, other people appear to believe that I appear to have feelings.
And in any case, the difference between 2 and 4 is only relevant to the question of whether a 4 year old can remember being 2, not what this article is about, which is adults not remembering being <4.
>There's a huuuge amount of learning that happens through this period. Your brain is learning things like 3-dimensional space, temperatures exist and I don't like some of them, I-have-two-arms, things fall when dropped, I must engage my big toe to stay upright while walking, other people appear to have feelings, other people appear to believe that I appear to have feelings.
Many of those things are completely innate. Walking for example, while people use the word "learn" in casual speech, is something that is innate. I just don't think the original comment is well-grounded in what we know about infant's cognition. And in any case, a 2 year old definitely understands 3D space.
... walking absolutely must be learned... They will automatically learn it without explicit teaching but indeed it must be learned. A child prevented from standing or walking for 5 years and then stood on their feet for the first time will not be able to walk.
That is simply not true. There are many cultures which greatly restrict infants' ability to move (e.g. traditional rural communities in Northern China or the Ache in Paraguay) and the children in these communities still learn how to walk. Not only that, but the basic neural mechanisms that are used in walking are innately specified (central pattern generators), not learnt (https://www.sciencedirect.com/science/article/abs/pii/S09594...). Now, there is a degree of "fine-tuning" that is learnt that makes the walking more fluent and precise, but the basic principles of walking are innate.
One only needs to see a foal walking less than an hour after birth to be convinced of this.
Part of the problem is that humans are born so premature that people confuse natural maturation with learning. Just as we don't learn puberty, we don't learn how to walk.
>Which cultures completely restrict their infants from attempting to walk?
The ones I mentioned in my comment.
>Did you read the paper you linked? It describes all the immense amount of learning that actually happens.
Again, as I said, there is a degree of fine-tuning but the core mechanisms are innate.
Some examples:
> In particular, the core premotor components of locomotor circuitry mainly derive from a set of embryonic interneurons that are remarkably conserved across different
species
>Detailed EMG recordings in chick embryos during the final week of incubation showed that the profiles of EMG activity during repetitive limb movements resemble those of locomotion at hatching
> In addition, human fetuses exhibit a rich repertoire of leg movements that includes single leg kicks, symmetrical double legs kicks, and symmetrical inter-limb alternation with variable phase.
I don't think you read the article, or else you think that "development" means learning.
>Have you actually seen a foal walking? They are very visibly learning how to do it!
They can walk right away, but they get better at it. It's innate, but you can fine-tune it. Like I said.
I think your definition of innate is counter to the common definition of innate. The common definition of innate is that there is no thought behind full understanding and capacity to perform- for example, snakes do not generally need to learn how to move without legs or how to open the mouth large enough to consume big food. There isn’t a try/fail cycle while they understand the capacities of their body. I fed my pet snake a baby quail for the first time in its life and it clearly had to learn how to eat it (tried and spat out the leg, wing, etc) even though the core mechanism of big mouth big swallow is there was clearly innate in it. Just because there is a core mechanism to walk existing in babies doesn’t mean the baby doesn’t still need to learn how to perform the behavior voluntarily, on command, consciously according to their own will.
What you just described for the baby applies equally to the snake. It's obviously difficult to neatly segment things into innate and non-innate, but the idea that walking is a matter of maturation rather than "learning" is the mainstream view among scientists and has been for a century.
Again, I conceded that you have to "fine-tune" to get good at walking. But the contrast that with say, playing golf. That's something that categorically has to be learnt, we don't see fetus practicing their drive in utero.
No, that is actually exactly what I was describing. If it was innate, they wouldn't need to trial and error their way (i.e. learn) to proficiency. But indeed they do.
Your analogy was puberty, which in fact happens with development regardless of trial and error (i.e. learning).
The two developmental processes are clearly distinct. The distinction is that one is a process of learning and the other is not.
I'm talking about being able to walk, you're talking about being proficient. I've said repeatedly that fine-tuning to get better is not incompatible with innateness.
My 2 year old went on a mental breakdown of a temper tantrum last night because she saw an apple on the tv, decided it meant she wanted an apple, and couldn't understand why she could not have an apple despite seeing one on the tv just then! A toddler is still trying to understand how reality itself works.
A 4 year old knows that jumping off of the stairs onto tile is going to hurt. A 4 year old understands the apple on the tv is an apple on the tv and is not a physical apple in the house.
Obviously a 4 year old is much more together than a 2 year old. But we're talking about a fundamental difference so great that no memories can be preserved. That's a high bar.
> The machinery constructing query vectors has updated its structure enough that its encoding of those query vectors is sufficiently dissimilar from the encoding of the stored memory vectors (which use the encoding from the last recall).
Wouldn't that result in very bizarre memories instead of no memories?
The brain is sufficiently complex that I'd expect gross distortions will get swept under the rug. You'd get either lightly distorted memories (dad is 18ft tall, mom's face is wrong, favorite toy lived in this spot instead of that one) or nothing at all. If a memory is totally corrupted, your brain won't give it to you because it doesn't pass your perceptive filters.
Children believe a lot of silly things that they "grow out" of thinking.
How sure are you that your childhood memories are accurate? How sure are you that you aren't simply conditioned to ignore distorted childhood memories?
Your brain is constantly changing, and much moreso when you're young. Imagine that you're trying to load files from a previous version of a piece of software into the current version - when the format is just a binary dump and nobody has been working on backwards compatibility.
i could never remember a lot far back until an incident caused some more recent years to disapear from memory. now i have some memories of when i was a baby but not from my teenagw years. i think its all in there, paging it in is just practically impossible to do on demand. maybe some people can do it though..
I had a similar experience when deeply intoxicated, I saw a shadow that reminded me of the texture of the wallpaper in the first home I grew up in -- it unlocked a cascade of memories from my early childhood (I must have been 2) that I never would have been able to recall otherwise. But I can't remember what I had for breakfast yesterday.
What you’re describing is dissociation, if you’re looking for a word! Its effects are not as well known and it’s under-appreciated how common it is. Plus having 3 different societal definitions really makes the concept difficult to talk about.
I’ve done a lot of work on this myself and one can connect with one’s parts, if that model seems accurate to you.
Another fun one on this example is somebody after coming out of a coma being effectively fluent in a foreign language. Obviously the person had to have at least some familiarity with it before, but there are countless cases where somebody goes from knowing a few words of a language (presumably having forgotten everything they learned in school or whatever) to suddenly speaking it, more or less, fluently. Similar incidents can also completely change a person's accent and have other such interesting effects.
There's almost certainly some way to tap into this without (what is usually/always) brain damage, which has interesting implications for the long-term future of education.
Could be that we don't really replay memories as toddlers and memory sticks by spaced repetition. When I have weeks where a ton of stuff is happening I don't remember a lot of the details even now at 30 because there's too much for me to repeat
I can't remember the vast majority of stuff before I was 2-3 but I remember seeing my younger sibling when they were bought home, there's less than a year between us. And I can remember a few other things earlier, they they are definitely fuzzier.
I also believe I have several distinct memories from around 1 or 2, but it's not clear that they are real and not just confabulations. I probably have bigger gaps in primary school, things faded away.
Frustratingly, a lot of what we think are memories are in fact: us remembering remembering.
Human memory is increasingly fallible with time because of this.
It's like playing chinese whispers with our own recollection, everything is a copy of a copy of a copy.
It makes sense that you would remember distinct events, but the trouble here is that nearly all memory as you so eloquently described, is "just confabulations"- which feels wrong instinctively, until it's challenged scientifically.
Some fact or visual memory is as that, but our memories are also made up of sensory sensations and emotional states, and combinations abound. In fact, a big part of what makes PTSD is a split off different aspects of a remembered event because it’s so difficult to process the whole experience. This takes place after a traumatic event, during attempts at memory consolidation.
Also traumatic memory seems to be stored differently than day to day memory, which helps explain that triggers and flashbacks are valid and real (for traumatized ppl), while the confabulation effect can also exist. Though I’d note that confabulation can happen but that doesn’t mean it “overwrites” the old memory, or that confabulation is in any way more common than just remembering a part or a whole event.
I have some pretty intense trauma in my life that I can recall extremely vividly even though i have little memory of the weeks surrounding the event. Often when I don't even want to. Certain memories are true memories.
I have some from "shorter than a table" age (I can no longer rember even which year), which later turned out to have been dreams from before I knew the difference between them and reality.
I remember a handful of insanely specific things from this age that I've since verified with my parents - I generally don't seem to misremember or fabricate memories. That being said there are a _lot_ of things that I require prompting via photo or video to remember and probably 90% of my life from that time is unrecoverable. I vividly recall random moments from family functions and some intense nightmares but the family dog, for instance, is a complete black hole. Weird stuff.
Primary school is still knocking around but I think I (and many others) suppress it to retain their sanity in adulthood :-P
I had an illness as a child which meant I was in and out of hospital for a while. I have a few very fuzzy memories from this time. The earliest, I think, was when my grandparents had come to the hospital. All I can remember is the feeling of the situation. They were all worried. I'm not even sure if I could talk at the time. I'm certain this memory is real because nobody has ever talked about this; it's not really a notable event for anyone else involved.
I also remember getting transported to Great Ormond Street in the night. I remember going in the back of a strange car (taxi? Hospital transport?) with a tartan blanket. I also remember being given a suppository, something that I didn't believe was actually real until about 20 years later (people don't really talk about them).
But I can't remember at all whether my siblings were there or not. I'm 2-3 years older than my brothers but in my memory they were just always there.
Erm... OK, but I've actually, you know, talked to my parents about this during the decades since. Nobody remembers this particular event because it was one night out of weeks of things getting bad, then worse, before finally getting better. This was almost certainly not the only time the grandparents visited, but it's the only one I remember.
Adults' memories don't just play out like a recording either. They remember many other details that wouldn't have been shared with me even if I could have understood it.
Maybe the issue is that at that age we haven’t yet learned compressions for our memories. Maybe at that age our best option would be storing something closer to raw experience because we don’t know enough about the world. Eg we might remember a chair as “brown, lots of orthogonal lines, about my size” rather than using our experience of other chairs.
Yeah, I think the first chair we observe is the template for recognizing other chairs later. Surely that 'seed' chair will undergo substantial revision in time as we become aware that it wasn't the ideal basis on which to base our modeling of all chairs to come, and we unconsciously 'normalize' the model for chair in our head (and adjust that memory too).
It stands to reason that this happens to all of the baseline models for the referents we learn (nouns, verbs, adjectives, etc) that subsequent learning surely reshapes.
This is likely comparable to the revision of 'concept' clusters that goes on in all unsupervised learning NN models, especially in humans, those that arise pre-linguistically (since words surely provide landmarks to memories enabling future retrieval). Perhaps we rewrite our early word2vec tables, making early memories less retrievable.
270 comments
[ 2.7 ms ] story [ 464 ms ] threadWell played, sir, well played! :)
> So what is the point?
What Rust helps with is memory safety, which is stuff like dangling pointers and use-after-free. Memory leaks are not a safety issue, they're an efficiency issue.
> Why are people so crazy about Rust?
The whole memory safety thing isn't actually the reason so many people love Rust. In the Rust 2024 community survey, 82% of users agreed that "Rust allows us to build relatively correct and bug free software." Memory safety is part of that, but there's a lot more to it.
So, for example, while Rust does not guarantee the lack of memory leaks, it does make them relatively hard to produce by accident in the first place. This is due to the intersection of a few different parts of Rust's overall design.
Personally I don't trust most memories I have from before around 4 because my parents had around 20 reels of home movies from that time and I know that I saw several of those movies a few years later.
I have a very clear delineation of before and after the age of 5, and I have hundreds of memories from before the move. Most are rather intense events, but I even remember the layout of my house, construction sites, my daycare, neighbors, friends houses, holidays, first bike ride, first lizard I caught, my pets, even some dreams I had, etc...
I didn't have a good idea of time yet, but I can retroactively tell when things occurred based on facts I later learned, like when certain neighbors moved, even my sister was born, etc...
I can’t guarantee that the memory is genuine, but it is very specific and I had to reconstruct my age from discussing the memory with my mom.
You can be continue to be frustrated being incorrect, or you can accept that you’re lying to yourself and move on.
It’s that easy. Human memory just doesn’t work like that.
This isn't true either. Memories themselves are non-falsifiable. So no matter what we either side says we'll literally never know the truth.
Another says, "You cannot remember it, you must have reconstructed it from being told or from pictures."
The memory itself can certainly be falsified: if you remember your mom using an iPhone in 1998, then that memory is false. If you remember talking to your uncle but he died before you were born, the memory is false.
How do we know if the origin of the "memory" is the person's actual experience, or their imagination of it based on descriptions or pictures?
Well if it's an event that had pictures taken of it, if it's the kind of event which grown-ups talk about, then the simplest explanation is that it was reconstructed.
But my son, almost 5, is always coming out with random things that happened when he was 1 or 2, which are absolutely of no importance to us, and which we would have forgotten long ago if he didn't keep remembering them. Once, for example, my wife put a pair of his shoes on top of the car while she put him in the car seat, forgot they were there, and drove off -- obviously at some point they fell of and were completely lost. Every six months or so my son talks about that incident completely unprompted -- something of absolutely no significance to us, but obviously something that struck him. And about the time I lost my temper and sprayed him in the face with water -- definitely not a memory I'm eager to revisit and bring up.
Someone dogmatic person may still say, "You must be bringing that up yourself somehow." That is certainly an unfalsifiable assertion -- there's no way, other than my assertion and probability, to prove that I'm haven't talking about those shoes on a regular basis. But I think any person with an open mind is likely to agree that "he's remembering them" is a more likely scenario than "the grownups are talking about those lost shoes all the time".
Probably these "revisitations" are his brain's way of refreshing the memory as his brain grows, which probably means the memory as they exist in his brain will have been shifted over the years. But that's still an original memory -- that's how adult memories work as well.
If at any point he misses a "refresh", they'll probably be gone; but the memories he manages to do this for will probably stay with him into adulthood. If at any point he "misses" a refresh, they'll probably be gone forever.
The flip side of this, of course, is that many of the memories we think we have as adults are heavily edited too. A few years ago someone recounted to me an emotional conversation that they'd had with someone on their deathbed (as an adult, only a few years prior to that). As it happens, I was in the other room when this conversation happened, and my memory of that conversation was significantly different. Obviously at least one of us is "remembering" something untrue. As I had little "skin in the game" about the content of the conversation, and the memory this person has seems to me very consistent with this person's narrative about their life, I'm inclined to think my memory is more accurate. But who can tell.
And the scars.
I will say that my strong early memories are of things that were either very positive or very negative.
Any scientist studying memory who doesn't have early memories just has a bad memory.
I had memories as a toddler that I confirmed after meeting my mother for the first time in decades a couple of years ago. Specific things about where I lived, events that had happened and how someone else in our family had died. I had another conversation with an aunt that confirmed other things, including a word from another language that I knew from when I was little but didn't realize wasn't a made up word.
Memory is fallible and we have an imperfect understanding of it, but I know for a fact from personal experience that people can remember events from very early in life if the emotional impact is deep enough.
However, the time that a giant plate glass mirror fall off a wall at a department store and crashed through me when I was about two years old? No idea. My parents had to tell me about that one later in life.
The matter came up because some time when I was in my twenties I said, apro pos of something or other, "[grandma] moved to [office] in year X", and they said "No, it was year X-1", and I said "well how come I remember [mundane sense-memory detail], if I was only two?" And they said, "yeah, [detail] is correct, but it was definitely year X-1", rinse-repeat, until they set out to prove me wrong about the year, which it turned out I was, which was a win for them, but then both I and my aunt with PhD in child-development were forced to conclude that I had a memory from when I was two.
Yeah. It blew my mind, too. I didn't think human memory worked like that, either.
Something-something only a Sith deals in absolutes. Alternatively, don't be such a jerk to that other guy. It's not really within the HN ethos.
5 year later show the same two balls and say you will give the kid an ice cream if it picks the same color. Obviously give the ice cream no matter the pick ...
Look for correlation.
My earliest memory was President Kennedy's assassination, but of course I had no idea of what a president was, life, death, etc. I was 16 months old. What I do remember was a bunch of strange people in and out of the house with the TV on and everyone very concerned, but what I remember most of that day was that it was the day I learned how to shake my crib and move it around the room while I was inside. I remember my parents not liking the fact that I was able to do that. I've never discussed that memory with them, and they're both gone now, so I never will.
I've got memories from when I barely knew how to talk, and didn't understand what the people around me were saying. My memory isn't good enough to decode what was said after I learned the language. I can clearly remember sitting in my high chair eating Alphabet cereal, and trying to make words with the letters I had not yet learned. I get sad when I think about how my parents would brush me off, probably thinking that I was too young to learn much. (My parents were both school teachers!)
However, on rare occasion, people send me pictures where I am awake at 3, but I have no memories.
Sorry folks, I see myself out, that was inappropriate, I hope nobody will remember.
We moved a couple months before I turned three and I have a bunch of memories that predate that. I remember being potty trained and lying to my parents about it so I could get a treat. I remember my dad coming home from the family farm and having injured his hand on some machinery. I remember stealing a pack of orange gum from the grocery store and chewing it all up underneath the piano in our basement. I could keep going and going.
My earliest memory was before I was two. Somewhere around 19-20 months. I know this because I had an ear infection and that's when it occurred. I'm sitting in my crib, crying due to the pain in my ear, and screaming "mom" at the door in the room. The room is mostly dark, but the door has been left open a crack to let in a little bit of light. From my vantage, looking through the slats in the crib, the door is open on the left. My mom comes in once, finally, and then leaves.
I also feel like I have an earlier one, being carried in a room which may have been my nursery, looking down on a changing table. I'm pretty sure I'm being held by my father, but I can't say for certain that this is real.
As you know we generally experience timescales logarithmically with age, i.e. your incremental experience of time is always compared in reference to dividing by the total passage of time experienced, which is why children tend to get bored much more quickly than adults, because waiting for ten minutes constitutes a much larger percentage of their current experienced life compared to that of an adult's proportion.
Since a baby/toddler has only experienced a tiny amount of time passage thus far, their tiny reference "yardstick" would result in their memory being snapshotted at an untenable timining resolution thanks to division-by-almost-zero. So perhaps the brain does some form of filtering to prevent the entirety of experienced memory being dominated by the super-early ages, or perhaps there is some equivalent of an overflow in their internal counter.
Keep in mind that the above are purely metaphorical as a functional description, and not to be treated as a literal hypothesis on the mechanistic operation of memory.
My guess has always been there is some optimal approach to learning that works by developing a really basic schema (what does a person look like, which ones are mine, roughly how does this body thing work, etc) then flushing all the training data and starting again. I vaguely expect the machine learning people develop some sort of similar process where they get a lot of value lightly conditioning a model before sinking compute into doing a full train.
Basically, my wild guess is there is a lot to learn in a baby's first few experiences but the risk of mis-encoding the lessons is so high that the brain uses the data to bootstrap but then throws it out as too unreliable and starts again once it orients.
I think the brain compresses experiences.
If you had a routine of k activities every day you won't remember every instance of that routine but you may remember what the routine was.
I think boredom is the brain's hunger more than its watch. It's about the needs of a developing brain, rather than plain time.
Right after the bulk of neurogenesis the brain goes into synaptic pruning where it strengthens some synapses and eliminates others. Developmental pruning is experience dependent and the scale of the process in childhood needs a lot of experiences to fuel it.
Those experiences are literal "food for thought". Building the body takes more food, building the brain takes more experiences.
Second, I've often thought I can correctly remember a scene from a movie or TV show I saw decades ago, but when I see it again I realize I got a tremendous number of details wrong, indicating that particular stored memory is very general in nature and not accurate at all.
I agree with your second paragraph too. I often think of a certain tone of voice or pacing with which a line was delivered, and it ends up being completely different when I rewatch it. And yet there are still Mandela Effect people out there insisting their memories are 100% infallible and it's more likely there's a global time travel conspiracy to change a dot on a KitKat package than that they misremembered something small.
If I ask a difficult recall question (usually some very niche medical knowledge) it will get it wrong or just hallucinate. But if I ask it in the form of a multiple choice question, it often gets it right. If memory works better when prompted with the information this way - I wonder how similar LLMs and humans actually encode memory?
That opposition doesn't quite make sense to me. If you are examining the linguistic view of it, then what we're talking about is that memory requires symbolic understanding. Not just language (a fairly sophisticated symbolic tool), but also just the ability to schematize your various sensory impressions into conceptualized objects. My earliest memory is sitting in a high chair playing with Duplos (large Legos for young children) being frustrated that I couldn't connect two pieces. That requires a few concepts such as a high chair, the tray sitting in front of me, the pieces I was playing with, and emotions like anger. Symbolic understanding isn't necessary for any of that to happen, but it does seem necessary to store it for recall decades later.
The problem with comparing this to rat memory is that I'm not convinced that rats have anywhere near the same kind of symbolic understanding, so how can they be compared to humans in this capacity? Rats can remember things, but babies can also be taught sign language starting at around 6 months, which seems much closer to rat memory (the direct linking of a perception with an emotion, unmediated by a symbolic transformation).
Just last week, my two-year-old spied the freezer pops in storage. She pointed out back and said, "Eat on deck!" Clearly, she remembered eating freezer pops on the deck, but the last time she did that was last summer (northern hemisphere) when she didn't know how to talk at all, let alone say "deck".
The same way one rides a bicycle without telling themselves to balance and operate the brakes.
I took lots of pictures of my kids. Those pictures trigger memories, but it's unclear if they would have remembered those things without the pictures. And there's no real way to tell if those memories are real, or just narratives attached to the trigger.
It's interesting to think about.
As an adult I have very good visual and audio memory, as well as perfect pitch. They're not as useful as they sound.
Make of that what you will.
Let's say neural memories are encoded in some high-dimensional vector space.
And so memory recall is an associative process that entails constructing a query vector and issuing it across the neural memory space.
And the brain is constantly learning, and that learning entails some changes in the structure of the high-dimensional memory space.
And let's say that re-encoding of a neural memory happens upon recall, and only upon recall.
Then it could be that all experience is in fact stored, but because of changes due to learning, those memories become inaccessible. The machinery constructing query vectors has updated its structure enough that its encoding of those query vectors is sufficiently dissimilar from the encoding of the stored memory vectors (which use the encoding from the last recall).
https://www.jneurosci.org/content/32/35/12144
Oops, someone else posted this further down too. Great minds think alike!
https://en.wikipedia.org/wiki/Catastrophic_interference
Age 2: Can point to their own body parts; hold something in one hand while doing something with the other hand
Age 4: Changes behavior based on where you are; can draw a person with more than 3 distinct body parts
There's a huuuge amount of learning that happens through this period. Your brain is learning things like 3-dimensional space, temperatures exist and I don't like some of them, I-have-two-arms, things fall when dropped, I must engage my big toe to stay upright while walking, other people appear to have feelings, other people appear to believe that I appear to have feelings.
And in any case, the difference between 2 and 4 is only relevant to the question of whether a 4 year old can remember being 2, not what this article is about, which is adults not remembering being <4.
Many of those things are completely innate. Walking for example, while people use the word "learn" in casual speech, is something that is innate. I just don't think the original comment is well-grounded in what we know about infant's cognition. And in any case, a 2 year old definitely understands 3D space.
One only needs to see a foal walking less than an hour after birth to be convinced of this.
Part of the problem is that humans are born so premature that people confuse natural maturation with learning. Just as we don't learn puberty, we don't learn how to walk.
Did you read the paper you linked? It describes all the immense amount of learning that actually happens: https://art.torvergata.it/retrieve/e291c0d4-b584-cddb-e053-3...
> One only needs to see a foal walking less than an hour after birth to be convinced of this.
Have you actually seen a foal walking? They are very visibly learning how to do it!
The ones I mentioned in my comment.
>Did you read the paper you linked? It describes all the immense amount of learning that actually happens.
Again, as I said, there is a degree of fine-tuning but the core mechanisms are innate.
Some examples:
> In particular, the core premotor components of locomotor circuitry mainly derive from a set of embryonic interneurons that are remarkably conserved across different species
>Detailed EMG recordings in chick embryos during the final week of incubation showed that the profiles of EMG activity during repetitive limb movements resemble those of locomotion at hatching
> In addition, human fetuses exhibit a rich repertoire of leg movements that includes single leg kicks, symmetrical double legs kicks, and symmetrical inter-limb alternation with variable phase.
I don't think you read the article, or else you think that "development" means learning.
>Have you actually seen a foal walking? They are very visibly learning how to do it!
They can walk right away, but they get better at it. It's innate, but you can fine-tune it. Like I said.
Another (impressive) example of an animal innately walking, this time to avoid predators immediately after hatching https://www.youtube.com/watch?v=B3OjfK0t1XM
Again, I conceded that you have to "fine-tune" to get good at walking. But the contrast that with say, playing golf. That's something that categorically has to be learnt, we don't see fetus practicing their drive in utero.
Your analogy was puberty, which in fact happens with development regardless of trial and error (i.e. learning).
The two developmental processes are clearly distinct. The distinction is that one is a process of learning and the other is not.
My 2 year old went on a mental breakdown of a temper tantrum last night because she saw an apple on the tv, decided it meant she wanted an apple, and couldn't understand why she could not have an apple despite seeing one on the tv just then! A toddler is still trying to understand how reality itself works.
A 4 year old knows that jumping off of the stairs onto tile is going to hurt. A 4 year old understands the apple on the tv is an apple on the tv and is not a physical apple in the house.
They are so far apart.
Wouldn't that result in very bizarre memories instead of no memories?
Children believe a lot of silly things that they "grow out" of thinking.
How sure are you that your childhood memories are accurate? How sure are you that you aren't simply conditioned to ignore distorted childhood memories?
I’ve done a lot of work on this myself and one can connect with one’s parts, if that model seems accurate to you.
There's almost certainly some way to tap into this without (what is usually/always) brain damage, which has interesting implications for the long-term future of education.
Human memory is increasingly fallible with time because of this.
It's like playing chinese whispers with our own recollection, everything is a copy of a copy of a copy.
It makes sense that you would remember distinct events, but the trouble here is that nearly all memory as you so eloquently described, is "just confabulations"- which feels wrong instinctively, until it's challenged scientifically.
Also traumatic memory seems to be stored differently than day to day memory, which helps explain that triggers and flashbacks are valid and real (for traumatized ppl), while the confabulation effect can also exist. Though I’d note that confabulation can happen but that doesn’t mean it “overwrites” the old memory, or that confabulation is in any way more common than just remembering a part or a whole event.
Primary school is still knocking around but I think I (and many others) suppress it to retain their sanity in adulthood :-P
I also remember getting transported to Great Ormond Street in the night. I remember going in the back of a strange car (taxi? Hospital transport?) with a tartan blanket. I also remember being given a suppository, something that I didn't believe was actually real until about 20 years later (people don't really talk about them).
But I can't remember at all whether my siblings were there or not. I'm 2-3 years older than my brothers but in my memory they were just always there.
I’m sorry but, what? You think your extended family visiting a child in a hospital due to illness was not a notable event for them?
> They were all worried.
As a parent I can assure you it was a notable event, or no one would have been there, and they would not have been worried.
Adults' memories don't just play out like a recording either. They remember many other details that wouldn't have been shared with me even if I could have understood it.
It stands to reason that this happens to all of the baseline models for the referents we learn (nouns, verbs, adjectives, etc) that subsequent learning surely reshapes.
This is likely comparable to the revision of 'concept' clusters that goes on in all unsupervised learning NN models, especially in humans, those that arise pre-linguistically (since words surely provide landmarks to memories enabling future retrieval). Perhaps we rewrite our early word2vec tables, making early memories less retrievable.