Scientists monitored brains replaying memories in real time | Hacker News Reader

Scientists monitored brains replaying memories in real time

nih.gov

191 pointshhs51 comments

"NIH researchers found that our brains may store memories in neuronal firing patterns that are replayed fractions of a second before remembering."

So it's like a phonograph record. First the track is inscribed from the source sound (storing a memory). Then the needle is placed in a groove (by some triggering experience). Then the needle traverses the inscription, playing back the recording (remembering). The brain just does it on a different chemical medium than vinyl.

To me the most magical part is however the brain so deftly places the needle in the right groove as a response to the triggering experience. That seems to go beyond the phonograph metaphor.

To me, it much more closely resembles a key-value retrieval system. I wonder if we could learn anything about that by training an artificial neural net to perform key-value storage?

Yeah, my brain can prolly be 80% modeled as a Redis fork...

I think it's more like a graph with edges between associated bits of information. It's easy to remember things in order, but random access is hard unless you've established links for it. For example, if you memorized how to play a song, each part links to the next but harder to ask someone what is the n-th note in a song, even if they can play the song. It would explain why making associations is a useful memory technique. Traversing this graph when you aren't near what you're looking for can be hard ("it's on the tip of my tongue").

Not only that, but sometimes as happened to me recently with my own song, you completely forget how to play it until you hear it and start playing it, then suddenly you remember every little note and nuance you do did, but only while you're playing it and good luck explaining to someone what you're doing even as you're doing it.

Content addressable memory more like.

Machine Learning phd student here! I am currently working on a branch of ML called Continual Learning that deals with this theme exactly. It turns out Artificial Neural Networks have a hard time recalling things they they see in the past [1].

One of the most effective strategies to counter this issue is storing and periodically replaying old training items (i.e. “memories”) for the network [2]. There are also approaches that set out to mimic what we understand of the brain and adopt cue-based approaches [3], but they don’t seem to have reached meaningful outbreaks just yet...

[1] https://en.wikipedia.org/wiki/Catastrophic_interference [2] https://arxiv.org/abs/1902.10486 [3] https://arxiv.org/abs/1711.10563

I’ve been saying for years that it’s like a hashtable, where the key itself is hashed in a predictable quickly retrievable pattern or namespace

This article does not invalidate the idea

It's odd, because it very much resembles relational memory for me. I can summon no memory just by it's name, only from it's association. The space for sole base values is very limited. The plus side is that i can generally remember every caveat and detail there is to doing a thing once I've seen it once.

Yes, this idea is called sparse distributed memory https://en.m.wikipedia.org/wiki/Sparse_distributed_memory

That said, unlike an old gramophone which has a pretty direct linkage the final sound-waves, I believe the hints/associations that are retrieved are then subject to an astonishing amount of inference and extrapolation.

We don't remember things so much as imagine them.

The much more magical part to me is how firing these patterns of electrical activity gives rise to an experience of remembering...

Also, how exactly do they know the neuronal firing patterns happen before remembering? If you have to signal you remembered something (as in the article, a word pair), isn't there always a delay between the experience of remembering and signalling your recall? Isn't it more accurate to say the electrical signals are fired fractions of a second before communicating what you recalled? How to know if what you remember doesn't correspond 1-1 with the neuronal firing patterns taking place (or even before the electrical activity).

I wonder if you could measure someone remembering an incident, and then measure them remembering the same incident but with different imagined outcomes to find a sort of style gan embedding space for qualities of the memory?

I’ve always wondered if there is a way to stimulate a brain so that you could replay a memory so vividly that it practically feels like you are there reliving it again, like the way you might experience a dream but with more clarity, almost like a form of time travel.

I remember reading about some experiment in a psychology class where an old woman’s brain was probed and she was able to hear her highschool graduation sonata clearly as if it was being played in the room.

There must be a way, there’s so many memories I’d like to revisit firsthand, and people I’d like to see again that are no longer with us.

https://www.ncbi.nlm.nih.gov/pubmed/22282432

You can achieve this clarity in a lucid dream... how accurate the memories relived would be, is another matter

It's still very sci-fi now, but I would bet a fiver that in my life time-ish or 2 we will have bidirectional brain computer interface.

Take some ketamine :) ... I'm stating this as someone with first hand experience, so it may seem hand wavy and "made up": memories are visual and those visuals directly trigger your nervous system which then responds and often impacts the visual memory replay (often cutting it off or moving to something else). Your nervous system is also constantly telling your visual system what to "lookup". Ketamine dissociates the nervous system from visual memory so you can sort of play the memory through like a spectator. I think this is why it is so effective at treating things like PTSD -- the memories are often so frightening or triggering to the nervous system that they cannot be replayed, understood and integrated... but with Ketamine cutting off the nervous system they can be understood without a panic attack or heart rate shooting through the roof.

How do I start?

Something always worth remembering: The brain is not even as well understood as we'd like to. For what it's worth, they might just be measuring something that's analogous to a data bus. E.g. in PC terms: I can probe the data bus' copper traces on a PCB, but that doesn't really tell me how the actual flash memory silicone works.

(This doesn't make the research pointless though).

There is a theory on how memories are formed on neural networks through cycles of firings by an artificial intelligence pioneer. I’ve searched for it again to no avail.

Is there such a theory by Minsky?

I think a memory is not just repetition, but collision of different paths that lead to permanent memories

One of the most fascinating things this article claims is that apparently these scientists were able to detect the firing of individual neurons. I have briefly looked up some of the technologies involved in brain scanning before and if this is true it would be greatly exceeding typical resolutions.

Well, it's not a scan, it's a recording from a micro-electrode array (MEA) implanted directly into patient's brains who's skull was already opened for (I guess) surgery.

Direct link to the article (with paywall): https://science.sciencemag.org/content/367/6482/1131

DOI: 10.1126/science.aba0672

What a timely article, as I JUST watched Brainstorm for the first time tonight. The story is exactly this.

Probably not a coincidence

Well, considering it's been in my Netflix DVD queue for years, and it just happened to be next in line. Pretty big coincidence!

It's called the Baader-Meinhof phenomenon (https://en.wikipedia.org/wiki/List_of_cognitive_biases#Frequ...), and now that you know that, you're gonna be seeing examples of it everywhere.

Movies like "Brainstorm": "The Final Cut" "Strange Days" "Total Recall" "Altered States" "Lawnmower Man" "Ex Machina" "The Thirteenth Floor" "The Truman Show" "Inception" "Selfless" "Flatliners"

also wim wenders' "Until The End of the World"

This reminds me of a tangentially related episode of mind field, in which a Japanese researcher is working on tech that one day could be able to read people's dreams: https://www.youtube.com/watch?v=AgbeGFYluEA&vl=en

They basically ask people to look at different faces and measure their brain waves while they're doing so and then later ask them to think of different faces from the ones they saw and try to guess which one they were thinking of at that time. Pretty cool.

I think this is good place to remind everyone that your brain doesn't actually remember events. It only remembers the last time you remembered it. Every time you access a memory it's a lossy rewrite influenced by the emotions you are currently feeling and other memories you associate with it.

Reminds me a little bit of the book "Recursion", which I can highly recommend (4.2/5 on Goodreads).[1]

[1]: https://www.goodreads.com/book/show/42046112-recursion

Just to be clear, because the headline is a bit awkward - the scientists didn't replay the memories.

There's no device to implant false memories or anything in this article.

Luckily we can already implant false memories with advertising.

https://www.wired.com/2011/05/ads-implant-false-memories/

How long until we see the headline “Summons allows court to monitor brains replaying memories in real time”

It will never be easy. This kind of tech will rely on the subjects thinking of that memory. And you can always, well, choose not to. Or even corrupt the memory deliberately by imagining things.

Isn't it very hard to "not think about it" when required to not think about it ?

It's a skill that you can practice through meditation. But false-memories and fabricated memories will shoot this idea down long before that stuff comes into play. Memory is no where near reliable enough to be used as evidence like that.

Witness statements are used right now, and they're memory filtered through people's perceptions and desires; potentially way worse.

Correct - and they're now gradually having less and less reliability assigned to them, which is good.

You don't choose anything. I'm finding it hard that anyone can think otherwise with research such as the article indicating function before awareness.

In the U.S constitutional complaint about right against self-incrimination. But if it worked well it would be too useful to let that right stand, so I'm sure the courts would decide it was not covered by the right.

I'm sure they'd apply the same thing to it as they did with polygraphs.

If the memory monitor was shown to be 95% accurate in some study I'm not so sure you would be allowed to get out of it.

Probably at least long enough to get the same result without the currently necessary surgery.

Feels like a waste of time. Just make a memory dump and analyze it offline.

Anyone else reminded of Black Mirror's The Entire History of You [https://www.imdb.com/title/tt2089050/]?

Finally, we may need those much talked about tin foil hats as we walk down the street.

This begs the question of why they used animals for 20 years.

They will keep using animals because we know fuck all about how actually brains work and you cannot slice human brains and probe them to see electrical properties all the time since they are supply-constrained. This study is using a information channel of very very high-level, top-down approach.

Reply on news.ycombinator.com