This is a growing problem in many fields, IMHO. I've been wondering for awhile if it's an inherent flaw in knowledge in that if knowledge can't supplant older knowledge in a high compressed reduced form as things progress, we're just building so much information/knowledge for any given field that at some point, it may be quicker to simply rediscover the process than to search the knowledge for the prior work.

Even with optimal compression, there is still a finite minimum size.

This is where longevity research a la Harold Katcher comes in and allows for super centenarian geniuses to shape the future of humanity.

But what if there are necessary prerequisites for making significant advances in longevity, which themselves are so complex that they require more than a human lifespan to master? We could be at a dead end unless the computers start doing the work. And even then, perhaps the same impediment applies to the development of an AI that can bootstrap other AIs into getting the job done.

Harold Katcher: “The Illusion of Knowledge”.

If you could live two or three lifetimes, would you really want to spend them all on the same thing?

Feels like attention would become the limiting factor.

Here is a short story that covers that idea: https://slatestarcodex.com/2017/11/09/ars-longa-vita-brevis/

Thanks for posting it!

I've read it, lost it, and have been searching for it for a few years now.

I think we're kind of becoming multi a cell organism.

There's limit to human knowledge, but we are getting more and more efficient at communication, both with other humans and with machines.

Even without something like Neuralink we are creating abstractions and interfaces allowing us to quickly connect our work with others. E.g. especially with ML you can lazy load all necessary context.

Think of it like Rust eventually supplanting C++ once the ecosystem and libraries are complete enough for people to make the change.

The cognitive burden reduces via the new foundations not requiring deep understanding of the old ones to be able to make new steps forward.

This is not a good example, as Rust does not build on C++, it simply replaces it as the foundation. Of course, the development of Rust is built upon lessons learned from C++.

But there is no universal law that says that the foundations of a field must be simple enough for one human to understand them in 70 years. It may well be that the simplest possible statement of a field of knowledge is still too complex for a single human to understand it in a normal life-span (not to mention that our capacity for storing information is limited - you can't actually continually learn new things for 70 years without forgetting much of what you learned initially).

Even today, you could spend literally your entire life trying to learn everything we know about the human body and you would almost certainly die before having learned everything. Now, fortunately, there is plenty of real work, both as a doctor and as a researcher, that can be done by focusing on just one aspect of the body (say, the circulatory system) and having only relatively shallow knowledge about the rest. Still, there is the possibility already that a mind that could build on all of the deep knowledge we have could come up with new ideas in medicine and biotech that we are unable to because of this silo-ing.

> This is not a good example, as Rust does not build on C++, it simply replaces it as the foundation.

No, that's exactly the point. You can thereby drop the cognitive load of C++ and build greenfield on the Rust foundations.