Loading video...

Video Failed to Load

Go Home

Cognitive scientist Elan Barenholtz says memory isn't retrieval. It's generation. When you remember something, you're not accessing a stored file. You're prompting your mind, like an AI model, to synthesize a response. There is no image of your mother. He's exploring how we might improve our own prompt design,...

116,852 views • 1 year ago •via X (Twitter)

9 Comments

vitrupo's profile picture
vitrupo1 year ago

Source:

Pale Blue Mediocrity's profile picture
Pale Blue Mediocrity1 year ago

It's called "lossy compression", and that's what any modern media file does. So yeah, it's exactly like pulling information from a file. "Confabulation" is the word for his own "memory".

Shawn's profile picture
Shawn1 year ago

I think it's like everything else in the brain it's stored as a pattern connected to other patterns like emotion, experience and so on.

David Zhang (▲)'s profile picture
David Zhang (▲)1 year ago

Our own hallucinations don’t need to be accurate. They just need to be useful enough to survive in the savanna

Nishanth PS's profile picture
Nishanth PS1 year ago

That is how the human brain works. Consciousness is a controlled hallucination using our priors.

Tyler's profile picture
Tyler1 year ago

Michael Levin is studying this as well.

Julien's profile picture
Julien1 year ago

If this is true… then sure the “memory” in the imagined and intuitive form might be generated, but those inputs required to perform the generation are still retrieved aren’t they? So I guess memories are just the output of our wet ware performing RAG .

LIFE 2030 and Beyond's profile picture
LIFE 2030 and Beyond1 year ago

Talking about memory, Elan Barenholtz says each human has an engine that is able in the moment to generate the image of the past on command with the appropriate input. That image is not in your brain, but generated on demand. Current research supports his claim that the image is not stored in the brain. But human brain might still have to store some related info about the image for later retrieval. That is where hippocampal indexing theory comes in. The theory proposes that the hippocampus acts as a "memory index" or a bookmark, not storing memories directly, but rather forming and retaining an index of neocortical areas activated during an experience. How does it work? 1) When you remember something, by using the "memory index" your brain reactivates the same network of brain cells that fired when you first lived the event. 2) But the reactivation does not give you a perfect image or scene, but more like a sketch or outline. 3) Then the hippocampus triggers the rest of the brain (including the neocortex) to generate or rebuild the detail of the image/scene/smells/words on demand. 4) So, through this way, the memory feels both familiar and flexible: Familiar: The same rough wiring pattern is switched on, so the memory stays anchored to reality—you don’t confuse your mother’s face with a stranger’s. Flexible: The gaps are padded with whatever makes sense now, so small details can drift: her shirt colour, the background, even her expression. That’s why each recall can vary: mood, current goals, or new information can tweak what gets filled in. The generated memory can be altered by mood? Yes, this is supported by the current research (see the 2nd and the 3rd screenshots). So, the generated memory in the human brain may also be distorted and susceptible to misinformation.

Banjo Guy's profile picture
Banjo Guy1 year ago

I wish scientists would stop saying, "...like..." to make things sound similar. If I say that cats bringing dead mice are like humans sacrificing virgins to the sun, it doesn't mean they work the same way. It definitely doesn't mean they should have the same rights as humans.

Related Videos

Most people treat AI like Google: ask a question, get an answer. But what if AI could think *like/with you?* I reverse-engineer "Theory of Mind" to test if the model can form a "theory of my mind". Using AI as a mirror to understand myself by giving it personal context and seeing how well it can embody/reflect my expertise back to me helps me to evaluate the model's thinking. This is something like a single-user, multi-model human-AI synergy benchmark where the task is "strategic alignment with your own expertise." Here's my hypothesis: Does pre-loading context create emergent synergy and enhance collective intelligence? If the LLM has: - Your projects (what you've built) - Your career details (how you think about growth) - Your self-description (your values/philosophy) 1. Does this response sound like something I'd say? 2. Does it reveal blind spots I hadn't considered? 3. Does it reduce the "explanation tax" I pay in every conversation? Different models excel at different aspects of "me." Before you think I'm building a digital mirror to talk to myself: this isn't about AI companionship or AI psychosis. It's actually bout me understanding the model's reasoning so I can deploy the right model for the right task. I'm my own test subject because I have ground truth: I know what I actually think, what I'd actually decide, what I'd actually prioritize. Most benchmarks lack this. When a model says "Muratcan would choose X," I can immediately verify: "No, I'd choose Y because of Z." I think everyone should do this with their own domain expertise. - If you're a manager, load your case history and test which model best applies your reasoning. - If you're a creative, load your portfolio and test which model understands your aesthetic principles. - If you're a founder, load your strategy docs and test which model identifies blind spots in your go-to-market plan. AI should serve as an extension of your strategic thinking, a collaborative partner.

Muratcan Koylan

24,128 views • 8 months ago