The Best Toy In The Park

by jefftkjefftk1 min read25th Aug 202015 comments

22

Fun TheoryWorld Optimization
Frontpage

One of the playgrounds near our house has an angled ring shaped merry-go-round:

There are lots of different toys at the park, but this one is different. It has depth.

It's a freely spinning ring, with a 10deg tilt. As you walk up, it spins you back down again. It has enough momentum, however, that you can get yourself to higher positions on the ring by working it back and forth. Because of the tilt, your speed changes as you move around the ring, and you have to adjust your angle relative to the ring as well. First is the challenge of staying upright as you walk at the bottom. Then the challenge of getting yourself to the top. Then the challenge of getting around it at different speeds, changing orientation, riding with other people, rolling on it, jumping, etc.

There are shallow games, like Tic-tac-toe, that are fun for a bit, and deeper games, like Go, that can be enjoyed for years. While I haven't spent years on any of these toys, this angled spinning ring has the most depth of any I've seen.

Comment via: facebook

15 comments, sorted by Highlighting new comments since Today at 7:14 AM
New Comment

(mod note)

I generally am not sure how to think about the Frontpage status of Jefftk posts (and other similar types of posts), which are sort of a frequently-posted-slice-of-life blogposts. Most of them are technically meeting the frontpage norms, but the site'd feel a bit weird if too many of them were frontpaged. I don't have a principled opinion on the matter. I tend to mostly leave them on personal blog because they in fact feel like "personal blogposts."

This one seemed like it was getting at something that I found fairly conceptually interesting, and I sort of Frontpaged it on a whim. It looks like the voting is a bit controversial, which maybe is a sign I should have left it on Personal Blog.

I'm posting this mostly so people have some insight into what it's like what the mods make judgment calls they aren't sure about.

Heh, I am happy I am not in a position to have to defend this on some general principles. I really like this post because it is short, interesting, and unusual. (Having many posts like this would remove the "unusual" part; this is a problem with having rules.) And a part of the reason is that I find it quite funny to have this on the frontpage of Less Wrong. :P

Technically this is a part of World Optimization. At kindergarten level, but one has to start somewhere.

Thank you for the recognition that global optimization requires global knowledge. There are no rules that can be applied to independent posts which will result in the desired mix of frontpage posts, especially if those rules are public and Goodhart-able.

Having a policy "we promote a few lighter-weight posts that may interest a smaller portion of our readership each week, even when they don't meet the formal criteria" crosses boundaries, but leads to the desired results.

That's a pretty moderator-judgement-heavy statement, which may or may not fit the definition of "a rule". But it's required for the actual desired mix of levity and seriousness.

What are the technical frontpage norms? What kind of norms does this post fulfill that a post like https://www.lesswrong.com/posts/c5aycbSsSc38XWPEc/taking-vitamin-d3-with-k2-in-the-morning doesn't?

The vitamin D post looks straightforwardly frontpage. I'm guessing that at the time, it wasn't frontpaged simply because we didn't have good infrastructure yet for making sure we evaluated each post.

I'm still surprised how this post qualifies. Is the current standard document https://www.lesswrong.com/posts/5conQhfa4rgb4SaWx/site-guide-personal-blogposts-vs-frontpage-posts ?

It doesn't define what it means with Useful, novel, and relevant. An older document said:

1.1. Usefulness, novelty, and fun. The frontpage of this site is for serious intellectual engagement with interesting ideas, with a focus on ideas that are important but challenging to evaluate. Topics that lack inherent importance are OK if the discussion quality is high enough, and particularly if the discussion is useful for other purposes, like building skills; but the best topics will usually be consequential and neglected ones.

The topic of which toys in the park are the best doesn't look to me like on of inherent importance. 

The discussion quality also doesn't seem to me very high. It just asserts that the criteria for deciding what the best toy happens to be usuable in different ways. Alternative criteria just as which toy offers the activity with the highest fun or injury rates with the toy that you might use to decide which toy is the best one aren't considered. 

Neither the topic nor the way it's discussed in this post seems to me related to why LessWrong exists.

Looking through the last months it seems that among posts with karma >10 those that aren't frontpaged are mostly about COVID-19. Presumably that's because they aren't timeless. https://www.lesswrong.com/posts/ZpLTTKs9QQ9CZXfFz/isn-t-tesla-stock-highly-undervalued seems to me another post that isn't timeless but is frontpaged. 

In the same time I would see the following posts frontpage material:

https://www.lesswrong.com/posts/oQYaRMYRjsTzNQkzu/how-to-learn-from-a-stronger-rationalist-in-daily-life

https://www.lesswrong.com/posts/Dpt7MGDixG2hiFm4n/how-do-you-study-a-math-textbook

I think the topic of "which toys have deep fun" is fairly tightly coupled in my mind with "how do human values even work? How do we make a future that is good?" 

(Note that I don't think this means this post should automatically be frontpaged, I do still consider it a weird edge case)

I think the other two posts you list are also plausibly frontpage, and I don't currently have a strong opinion on whether it was a mistake to not frontpage them. (There is still some amount of 'stuff falls through the cracks sometimes')

I do think that you can have a general discussion about which toys produce deep fun that can illuminate aspects about how human values work that's on-topic but I don't think the depth in which this post looks at the question leads there. 

My take is something similar to Dagon's: I think I want nonzero posts like this in the LessWrong feed – the fact that it's short and playful and just gets me thinking about stuff without requiring me to boot up a "deep thinking mode" is in fact a key part of why I like it.

I agree there is also a "what sort of toys produce deep fun" bigger blogpost that goes into more depth that should maybe exist some day. But meanwhile I actively do want to have blogposts with a range of depth, and for some to be more like "here's one concrete example" than "here's a bunch of examples and an overarching theory."

I don't currently have a principled stance of "exactly how often and when should posts like this be on frontpage", like Dagon noted elsethread it's sort of important that that be vague and non-gameable.

the fact that it's short and playful and just gets me thinking about stuff without requiring me to boot up a "deep thinking mode" is in fact a key part of why I like it.

If the LessWrong frontpage isn't about booting up "deep thinking mode" what is it's purpose?

Ideally, I would expect that reading things that require "deep thinking mode" is why people go to the frontpage. 

I'm not saying that's not a key purpose, but I don't think it can or should be the only purpose. 

Note that the current LW frontpage post is on average much longer than the original sequences – those contained many instances of Eliezer spelling out one idea with one example very clearly and concisely. And I think this was good. I think this was both good pedagogically for readers, and good for Eliezer's own writing/work ethic.

I think there is potentially some argument that posts like this should be shortform rather than top-level posts, but I currently lean against it. But if so, I'd want shortform better integrated than it currently is, such that reading shortform is a more natural thing to build LW habits around. 

But I think you get much more intellectual progress if you enable small bite sized posts like this than if you don't. I don't think Jeff was at all likely to build up a theory of what makes toys fun and write an extensive post on them. But I do think other people are more likely now to take this example and have it mulling in the back of their head, and have it feed into more comprehensive ideas.

(One of) the points of frontpage is to give newcomers a sense of what to expect from LessWrong overall, that is representative of where the overall site is trying to go. This includes ideas at different stages of the pipeline. Again, this is not a post I'd want to have here all the time – I frontpage maybe 1 in 20 of Jeff's personal-blog-style posts, and current there aren't many other personal-blog-style posts that get written. All I'm arguing here is the number should probably be non-zero.

On the readership side – if every LessWrong post required booting up deep thinking mode, then people would only come to LessWrong when they felt able to Deep Think, which is actually pretty rare. A key point of LW in my mind is to leverage a lot of untapped intellectual capital via the power of "being fun and feeling low effort." We're not paying people to be here.

I think allowing discussion to go to much to being low effort was one of the ways LW1.0 failed over time. People were spending less energy on effortful posts and then more on effortless posts. 

On the readership side – if every LessWrong post required booting up deep thinking mode, then people would only come to LessWrong when they felt able to Deep Think, which is actually pretty rare.

The fact that LW1.0 drew a readership with an average IQ of 140 (in the census) does suggest that it drew readers that have an ability to think deeply that's quite rare. That's no bug.

You seem (to me) to be making an extremely strong claim that the number of posts like this should be zero (as opposed to like 1-3 per year).

I'm still pretty uncertain about whether this post should be frontpaged, but that just seems like an extremely strong claim to me.

I assume that the point this post is making is that the cerebral logic exercised by the spinning wheel must be much richer, as you're performing an optimization in continuous physical space, which requires coordination of many muscles in very subtle ways, which depend quite specifically on your orientation and position on the wheel; whereas (supposedly) Go is very discrete and thus can't necessitate the same degree of cognitive complexity/multidimensionality. Managing to stand and walk on uneven terrain is something that our cerebral machine has been trained for as long as things have legs (i.e. around 350 million years) and prepared before, and it has remained a crucial element of evolutionary fitness, so must have been honed for all this time to great precision. I don't doubt that there is more "total depth" in processes involving physical interaction.

However, Go is a process which we perform consciously, as opposed to walking and staying upright, which is overwhelmingly unconscious. There is little conscious depth, i.e. depth that we can introspect, experience or enjoy. We don't think much about which specific centimeter we'll place our foot at, we just feel the correct motion and perform it. We could probably reduce the problem to a mathematical model, but that would forgo the depth we've assigned it above (the resulting complexity would be comparable to Go). In Go, we are able to analyze the step and introspect its decision process, to—if desired—absolute precision, while still observing the heuristic that produced the step (unconsciously). That's the depth I find stimulating and interesting. I'm aware of no person that has been enjoying spinning wheels for years, hours a day...

(Also, FTR, Go and even Tic-tac-toe typically have humans sitting opposite you, interaction with which is just as deep and more fulfilling than running endlessly on your lonely spinning wheel!)

Edit: After actually following the vendor link in the post, I see that there is potential for interactive multiplayer fun on these. The above reflects purely on solo player experience (or for Go, playing online/against an engine). Not sure if anyone would want to play on these for years with other people either, though.

There is little conscious depth, i.e. depth that we can introspect, experience or enjoy. We don't think much about which specific centimeter we'll place our foot at, we just feel the correct motion and perform it.

[Epistemic status: personal observation of mental states which are difficult to describe well]

This doesn't quite match my experience (though I haven't had much of this experience for a while, so take this with some extra salt). What I remember is being able to have deep conscious interaction with an ongoing complex motor process like that, but in a less synchronous way. Activities like playing board games involve conscious manipulation in the same subjective timeline as the main flow of action: you consciously think about what move to make, then you reach out to make it, then you consciously observe what your opponents are doing, then repeat (depending on the game of course). Activities like playing music or running, by contrast, involve primarily unconscious cycles as the “main” flow of action, but the conscious mind can still watch it happen and then reach out and touch it in parallel, placing constraints and nudges and altering parameters. What it doesn't get is waited on for a say in every microdecision, because those are happening too fast—but consciously remembering a finer-grained history lets you try to extrapolate what nudges to give to create the pattern you want next time, which is how I would realize the loop of deliberate practice in motor skills, which I just now notice does make the “(consciously) think, then act” pattern again, but one level of temporal chunking up. And it's possible to have a conscious say in an upcoming microdecision if the conscious mind predicts them far enough in advance and the unconscious mind has enough spare processing power that the information can be integrated in time.