Even claude with “memory” enabled isn’t really “remembering” anything. It just injects it into the context and you hope it happens to find it relevant in its attention mechanisms, and then remembers to actually act on it. Anthropic’s own documentation states claude can and will ignore/truncate these. It’s a context trick, nothing approaching actual “memory,” and in fact, arguing with it will make a bunch of memory files, sometimes contradictory, and clutter up the context and act even worse.

I'm not a big fan of arguments like "it's not the real [human quality], it's [mechanistic explanation]." They lack a part: "because the [human quality] allows us to do X, Y, Z, which is impossible with [this mechanism]."

I agree that the relevance of retrieved pieces and the management of long-term storage could be improved, though.

It just acts fundamentally different than someone who would remember.

If someone only remembered vague scraps of what you'd expect them to remember, you might say the person can't remember.

It's much closer to notetaking and reviewing before responding than it is memory.

The issue with anthropomorphizing like this is that "memory" comes with baggage of expectations for it to do certain things, and it breaks them.

Just like "thinking" implies chain of thought, but you'll frequently get those reasoning traces and then a 180 in the final message.

> Just like "thinking" implies chain of thought, but you'll frequently get those reasoning traces and then a 180 in the final message.

Be me, think extensively about an exam question, do 180 because the most probable option is just too obvious to be true (no, this part wasn't verbalized, it's how I describe what I felt about making the decision to do 180, or maybe it's a rationalization and it was a natural analog of an unfortunate sample from a probability distribution).

Yeah, I think I've come to the conclusion that the biggest breakthrough we need before we can replace human thought is going to be some mechanism for live update of weights. "Learning" by injecting into context just isn't good enough.

But billions of dollars are going towards research to find these breakthroughs, so we'll get there eventually.

The AI doesn't actually go to sleep at night, it's [mechanistic explanation]

I just don’t find it really relevant to the argument presented I guess. I disable auto memory and have my own mechanisms and infrastructure with how my agentic system “knows” and “remembers” things which is roughly an automated, sometimes self-correcting working index on the file system. It behaves much better than claude’s automated “memory” system, so I use that, but digging into how that worked and making something of my own just makes me really dismissive of comparing it to something like actual memory, so I apologize if it came off dismissive.

I guess I will expand on what I meant why I react to claude memory acting mechanically or logically anything like human memory, is because it isn’t how memory in the brain works, they’re not comparable.

The layman’s understanding I have of memory, as someone that has dealt with memory issues much of my life, is that memory formation is heavily tied to emotions. emotions are triggered by input which sends a complex set of signals throughout the brain - you’re not just finding where in your head to store this, your brain is deciding how important it is, and what else to correlate it with - so it can tie them to other related memories. then on top of all this, much of the sensory experience you intake is subconsciously compared against high priority memory impressions and deciding what to pay attention to.

you could, argue that the sensory input is the simple md files and the emotional mechanism is the same effect as to how attention mechanisms work in llm’s. Ok, I can almost buy that, but these tools lack a fundamental ability to decide how important things are.

an analogy. you tell a person “if you pick a daisy in the next five years, an assassin will come to kill you” and they hold a knife to your throat while they say it, your brain whether you like it or not is going to say “THIS IS AN IMPORTANT MEMORY I NEVER MUST FORGET” and you’ll see something that looks like a daisy and have a panic attack 3 years later. that memory is never fo tell me claude or other tool harnesses using memory harnesses can prioritize memories the way that human would, instead they forget even when reminded, because the human brain is just so much better at it

I’m normally not a big fan, but in this particular case it matters a lot. I could come up with some functional argument, but really I care from a model welfare perspective, whether the model understands its reasoning traces to be a part of itself or it’s simply predicting what a character who wrote the current intermediate tokens would output next.

Reading this is like watching someone call cars moving a “trick” because it uses gasoline as fuel while humans don’t move with gasoline.

Bro the LLM is a token machine, it is reasonable to have its short term memory be represented as tokens in context because the LLM is a token machine. Call it a trick if you want but it does fit the actual definition of what memory in actuality is.

Why argue with it? Why not just edit the memory yourself?