How To Build Knowledge Skyscrapers

[Elliot]

How To Build Knowledge Skyscrapers

[alanforr]

When I click this link the page doesn’t load. Nor does criticalfallibilism.com.

[alanforr]

It’s working again now, thanks.

[alanforr]

CR objection: Knowledge is a web not a skyscraper.

Reply: Yeah, but a skyscraper, tree or pyramid is an easier to understand mental model that makes a good, useful approximation for many purposes. CR’s arguments don’t contradict that.

Having a bad link in a web is a bit like having an overloaded bottleneck in a factory. It’s difficult to get any good knowledge out if one of the links in your web is blocking progress.

[Elliot]

I didn’t do anything and don’t know what happened. If it recurs, you could read the copy you received by email. You might also have a copy in an RSS reader.

[doubtingthomas]

Another great article. I can share some of my ideas that make me feel good which might explain why some of the CR objections exist.

I think of creativity as this magical thing which can find connections which everyone else missed or it can create a guess that solves a big problem. So just like Darwin I too could end up having an insight which might solve the problem of how to make AGI. The opposite of that is I have to systematically build up knowledge in increasing order of complexity which tells me that to achieve some kind of breakthrough I have to do disciplined work to improve. This is one of the reason I like the ‘foundations don’t matter’ misinterpretation of CR.

[Fire]

I really like this article. The most counter-intuitive idea to me was the bit about how focusing effort on non-error problems is an error. I only somewhat understand the concept of bottlenecks. I have net yet read Goldratt. I find the pyramid/skyscraper metaphor useful and more intuitive than the web of knowledge idea.

Revisiting older and more foundational ideas is also interesting because of the ability to bring additional context to those fundamentals.

[Eternity]

Read this article last night.

I liked it.

@Elliot

9. There are several converging, abstract reasons why medium errors are rare.

I get through the example why medium errors are rare, I think. Big errors are things that become problems fast. Small errors don’t really become problems. A medium error is something that becomes an issue somewhat far in, but not too fast. What are the abstract reasons they’re rare?

@Elliot

CR objection: Foundations don’t matter.

Reply: No, CR says that there are no ultimate or essential foundations. That is, we can add an unbounded number of basement levels, and there are many different valid skyscraper designs. But what sub-parts you build an idea from does matter, particularly when learning, even if that idea ends up fairly autonomous and could have been built in other ways.

oh

I think I’ve read in passing about the “Foundations don’t matter” stuff and just took it at face value. Ok. So there’s more to it.

[Elliot]

@Eternity

I get through the example why medium errors are rare, I think. Big errors are things that become problems fast. Small errors don’t really become problems. A medium error is something that becomes an issue somewhat far in, but not too fast. What are the abstract reasons they’re rare?

Here’s a relevant thought:

Errors can cause failure or not cause failure. What is a medium error? For quantitative issues (which are less inclined to the failure/not-failure binary than non-quantitative issues), medium is moving a factor in a bad direct to be near near a breakpoint so that random variance can cause problems. Abstractly, most factors are not near breakpoints, breakpoints are sparse, and if you move a factor by a large amount you’re not likely to land near a breakpoint (nor is moving a small amount likely to land near a breakpoint).

[Eternity]

@Elliot

Errors can cause failure or not cause failure. What is a medium error? For quantitative issues (which are less inclined to the failure/not-failure binary than non-quantitative issues), medium is moving a factor in a bad direct to be near near a breakpoint so that random variance can cause problems. Abstractly, most factors are not near breakpoints, breakpoints are sparse, and if you move a factor by a large amount you’re not likely to land near a breakpoint (nor is moving a small amount likely to land near a breakpoint).

Ok. A medium error is moving something in a way to be near a breakpoint? Or a medium error is something near a breakpoint, so, with variance, it can sometimes be serious issue, sometimes not?

Hmm. Something I thought after posting:

Larger errors are bottlenecks. Small errors are non-bottlenecks. It seems like a binary to me. Either it is a bottleneck or it isn’t.

[Elliot]

In a simplified model, some trait currently has a value of 40, and it causes failures for the whole system/goal when less than or equal to 10. It has random variance of +/- 3.

A small error reduces it to 38, causing no failures.

A big error reduces it to -50, causing failure.

A medium error reduces it to 11, which leads to sporadic failures due to random variance.

To get failures sometimes, instead of always or never, the error has to get the trait within 3 of 10. You can see why that’d be rare compared to errors that don’t get near 10 or go way past 10. The target range is small. In general, breakpoints and the areas near them are small and rare compared to all numbers. It’s a lot more common to be far from a (relevant, important) breakpoint than near one.

This is one simplified model. It’s sometimes useful and sometimes doesn’t fit well. It is not a universally correct way to look at things.

Note: People tend to focus on non-randomly selected traits which are nearer to breakpoints than average. Their intuitions often assume being near a breakpoint because they tend to look at things that are and ignore things that aren’t. This is often reasonable or practical, but it’s misleading when understanding abstract theory about how things work. It’s sort of like how a ton of people are used to cities and spend most of their time in cities but most land isn’t in a city and you need to get away from a city-biased perspective if you want to be a good geographer.