3 ms·

Note that this works within a single sequence of tokens. It might be consistent with "episodic memory" metaphor if we consider a particular transformer run as i

Note that this works within a single sequence of tokens. It might be consistent with "episodic memory" metaphor if we consider a particular transformer run as its experience. But this might be very different from what people expect from "memory" - i.e. ability to learn vast amounts of information and retrieve it as necessary. This is more like a refinement of transformer attention: instead of running attention over all tokens (which is very expensive as it's quadratic), it selects a subset of token spans and runs fine-grained attention only on those. So it essentially breaks transformer attention into two parts - coarse-grained (k-NN over token spans) and fine-grained (normal). It might be a great thing for long-context situations. But it doesn't make sense when you want millions of different facts to be considered - making them into long context is rather inefficient.

yorwba 1y ago
It would be inefficient if you had to do it from scratch for every query, but if you can do it once as a preprocessing step and reuse the prepared context for many queries, it might start to become more efficient than a shorter context that includes only some documents but has to be reprocessed because it's different every time.
- killerstorm 1y ago
  Yes, I think it might be a good solution where you have a context up to 10M of tokens and you do a lot of requests with that context. It might be relevant for agentic stuff which tends to produce long chat logs - especially with some gadgets on top, e.g. some 'episodes' might be completely removed as obsolete. But I don't think it's a good solution for bigger amounts of data - as in that case it's more beneficial if that can be formed into independent memories.