How much is death a limit on knowledge accumulation?

post by Gordon Seidoh Worley (gworley) · 2023-02-14T03:54:16.070Z · LW · GW · 7 comments

This is a question post.

Contents

  Answers
    5 Dave Orr
None
7 comments

The older I get, the more it feels like death is a major bottleneck to the accumulation of certain kinds of knowledge and insights.

It's not much of a blocker to shallow things that are easy to explain precisely. For example:

It does seem to be a blocker for deep knowledge that's hard to easily explain and teach. For example:

The former category of things benefit from being easy to put into writing and be transmitted accurately every time. This allows generations to easily build cumulative mountains of knowledge. Yes, each generation a person has to learn more before they can contribute to the mountain and, as of maybe 200 years ago, there's too much accumulation for one person to know everything known, but we don't seem especially close to the limit, and very recent advances like LLMs may offer a way to scale our learning further by outsourcing the remembering of many details.

The latter category suffers because each of them is hard to transmit and it's easy to be confused and transmit the wrong thing. Thus, to some extent, each person needs to rediscover the knowledge of past generations for themselves and only after the fact may recognize that previously generations had some of these same details figured out. Try as we might with institutions that teach young people the best of our wisdom of the past in these areas, it's imperfect and you have many people who will reject the knowledge because they don't understand it and will continue to reject it until they work it out for themselves. Thus even if somewhere there is an unbroken line from Socrates to the present day of ever advancing philosophical knowledge manifested in a single person, almost nobody would believe whatever they have to tell us because the inferential gap would be too large.

But is this story right? Can we quantify this? Is there some way we can measure how much knowledge we miss out on because we die and each generation has to spend time rediscovering things before they can advance it further? And how much additional knowledge would we accumulate, and how quickly, if death was not an obstacle?

I don't actually know, which is why I'm asking. This seems solvable, but I've also not looked into the question. Feels like a thing maybe progress studies folks would have a grasp on, but I don't recall seeing this specific question addressed, especially regarding hard to transmit knowledge where it's somewhat hard for us to know how much progress it's even possible to make.

Answers

answer by Dave Orr · 2023-02-14T06:08:10.816Z · LW(p) · GW(p)

I think in practice roughly the opposite is true.

As people age, they become less flexible in their beliefs and more set in their ways. If they are highly influential, then it's difficult to make progress when they are still alive.

Science advances one funeral at a time: https://en.m.wikipedia.org/wiki/Planck's_principle

It's true that as you age you accumulate experience, and this stops when you die. For you, death is or course a hard limit on knowledge.

For the world, it's much less clear.

↑ comment by Andy_McKenzie · 2023-02-14T12:33:34.755Z · LW(p) · GW(p)

"Science advances one funeral at a time" -> this seems to be both generally not true as well as being a harmful meme (because it is a common argument used to argue against life extension research).

https://www.lesswrong.com/posts/fsSoAMsntpsmrEC6a/does-blind-review-slow-down-science [LW · GW] 

7 comments

Comments sorted by top scores.

comment by Vladimir_Nesov · 2023-02-14T15:55:51.147Z · LW(p) · GW(p)

Part of variation in capability is learning speed disparity. Harder things could take 10 years to learn for the best mathematicians, but 50 or 500 years for less capable people, and so that never happens. There are probably things that take 500 years to learn for the best mathematicians.

comment by romeostevensit · 2023-02-14T05:06:22.765Z · LW(p) · GW(p)

I feel like the latter have poorer feedback loops and that's more relevant than 80 year lifetimes. I don't think there's actually all that much total rigorous content in those fields relative to the former categories.

comment by Algon · 2023-02-14T12:55:35.708Z · LW(p) · GW(p)

Relevant SSC: Ars Longa, Vita Brevis.

comment by SomeoneYouOnceKnew · 2023-02-14T08:26:18.897Z · LW(p) · GW(p)

Something possibly missing from the list, is breadth of first-hand experience amidst other cultures. Getting older and meeting people and really getting to know them in such a short lifespan is really, really hard!

And I don't just mean meeting people in the places we already live. Getting out of our towns and countries and living in their worlds? Yeah you can't really do that. Sure, you might be able to move to <Spain> or <the Phillippines> for a couple years, but then you come home.

It's not just death here, but the breadth of experiences we can even have is limited, so our understanding of others, the problems they face, and thus the solutions we can come up with, often wind up with terrible failures.

Left as comment, rather than answer because it feels tangential.

comment by Dagon · 2023-02-14T13:35:42.662Z · LW(p) · GW(p)

I feel like it is confusing to call this “learning” and “knowledge”. The concern seems to be not about propositional information, but some kind of internal modeling, intuitional ease, and habits of cognition and behavior. I guess “wisdom” is the closest word that comes to mind.

I don’t see it as monotonously increasing for most individuals or groups, it seems more like a noisy bathtub curve - rapid improvement until adulthood, decline later in life, so it would surprise me if lifespan is the easiest improvement.

Just to stir the discussion pot, I note a correlation between suffering and wisdom. Perhaps we could train our wisdom faster by increasing life difficulty.

comment by Gunnar_Zarncke · 2023-02-14T10:42:15.611Z · LW(p) · GW(p)

If you view society as a single learning agent, then the population of individuals is a set of sub-models and the learning strategy is to train the most predictive sub-models off each other. You might not even want to keep sub-models that have converged on a stable state around for too long. Instead, you might want to train new models with a combination of the best models from the previous generation.