Comment by RugnirViking

17 hours ago

I agree - I think one of the biggest reasons memory systems fail in LLMs is that they have poor theory of mind - they're terrible at considering how others will react. Both humans yes, but also future versions of itself. When asked to give advice to itself, it pontificates at length about trivial stuff it already knows and fails to emphasize the stuff that was new or interesting

Alfred North Whitehead talks about the notion of "Importance" as fundamental to the human (and all other living things) way of being. Living creatures first and foremost select information that is important to them from the broader environment, and then make decisions and take actions. (Of course at a physiological level it's much more complicated than this, but it's a sound philosophical description of how living things work.) LLMs lack this entirely. They have no selective filter because they weren't designed to have one (interesting question if you could even do that) and they're not evolved beings with a survival imperative. When they enter a self-conscious or other-conscious mode like you're describing, they just emit text that looks like the thoughts of a self-or-other-conscious person. They can't direct a stream of attention or hold a concept in the forefront relative to other concepts or (to your main point) think about what matters to the other person/being because they don't experience "matters." All they can do is emulate the verbal output of beings that actually experience these things, and given that I don't find it surprising they get trapped in loops over trivial things.

  • It will be interesting to see if LLMs “evolve” importance as they run out of ram and storage to think.

    For humans with limited space, “Importance” is an output of a first pass “of the available infinite amount of information, what do I need to consider to solve this problem”. And it’s not necessarily a good algorithm. People misidentify “importance” all the time.