Miniblog

A small update. As of now, there are no indications that anyone has noticed my post on the situation. Unfortunate, but unsurprising: it's unlikely that people will randomly stumble upon my website as of now, and the methods of contacting Neocities directly seem distinctly unmonitered. Nothing on the github front, either, but again, not unexpected, since it's on a closed issue.

I'm a little more worried than I was yesterday. After all, the intent of the change was to improve Neocities' front-facing quality, so, if that is truly being valued above all else, than it would be an objective mistake to revert the change, except perhaps in the very long term. And it would be very easy to decide that 'yeah, keeping it as is is fine, actually' after months of very little reaction to it.

And the lack of attention doesn't seem set to change, either. That's fine for now, since I still believe that there's a greater than even chance that this situation will resolve amicably, but if it doesn't, or if the radio silence goes on for too long (say, into December), I will probably start to post about this on other peoples' websites to try and get the word out.

You may be asking: Why bother? Way back when, websites didn't need to show up on a platform to get attention, so why is it so important now? Well, for one, I think it's a good thing that Neocities was a good way to get people onto your website/find new websites. And the current state of things encourages annoying strategies like advertising on other peoples' sites and abusing Neocities' social media features. And, quite frankly, how Neocities responds to this situation will be a good indicator as to whether they have misaligned incentives. If they do, there's a good chance they end up doing worse in the future.

In the interest of keeping attention on this issue, I won't be updating this site's front page until it is resolved.

As you may have noticed, I'm back! As for why I stopped, I got distracted by other hobbies, then this site just didn't cross my mind frequently enough to catch me in the mood to write something.

Then I saw someone say something about the Dark Forest Hypothesis, got annoyed, then decided to actually write an idea I had on the backburner.

This one was frustrating. Of my writeups, it wasn't take the longest in terms of time seriously invested into it (that was the AI one by far), I only started writing it a day before publication, but it sure felt like it. The issue wasn't coming up with counterarguments (if anything, it was too easy: I kept getting interrupted by coming up with more), but rather structuring them when writing. Every specific scenario failed in its own special way, making arguments hard to group together, and I quickly had way too many to keep organized in my head.

Ah well, at least a new Ancient Americas video released just as I finished writing it. A much-needed reward, if I do say so myself. This one is about a native Floridian monarchy that subsisted mostly on fish.

Oh, I said I was going to address something I said in point 7 in the preamble, so I guess I'll do that now: since the universe would exist for an infinite time, and every state is possible, it would therefore be possible for the universe to have gone from a fully settled state to a fully untouched state before we arose.

You may recall that, a year or three ago, that there was significant coverage of a major fusion breakthrough. If you don't, the tl;dr is that the National Ignition Facility managed to get a fusion reaction to produce more energy than was required to initiate it*. This was done by, basically, shooting a fuel pellet with powerful lasers so that the outer layer heats up and expands so forcefully that it causes fusion in the interior of the pellet.

You may also remember, in the months following, there was a smattering of criticisms by 'skeptics' (scarequotes will be explained in a moment) writing that this wasn't a very important development after all. Their arguement was that, though the above claim is true, that there was over an order of magnitude energy used to power the system than was produced by the fusion reaction, so it clearly doesn't mean anything in regards to the viability of fusion.

Though this is true—over 400MJ input vs ~3MJ output—things become interesting when you look at the energy usage breakdown. In particular, the vast majority of that energy was being used to drive the laser, which had ~2MJ of energy. That seems like a really low energy efficiency for a laser, does it not? As far as I can tell, the inefficiency in the NIF's lasers is a result of a combination of having been designed and built decades ago and the need to vary the laser's properties for experimentation.

In contrast, modern laser designs can have an efficiency of over 50%, and 30% efficiency would be perfectly reasonable to acheive even accounting for constraints needed for a commercial fusion reactor. That would be 6MJ input to 3MJ output: still not breakeven, but not as absurdly far away as certain people say.

And it isn't like this avenue of fusion research is exactly exhausted, either. Indeed, there are several ways to increase output energy more than input. One is to increase the laser's energy: since, as of now, most of the fuel in a given pellet is unspent, increasing input energy into the pellet increases the proportion of the pellet undergoing fusion, rapidly increasing the yield (actually, the NIF seems to think that the increase in energy would be more than proportional; 3MJ laser to 30MJ fusion reaction). Another way, of course, is to increase pellet size.

*Actually, they did this in 2014. The NIF uses an indirect drive method, which means that the laser is pumped into a small chamber that captures and reradiates the laser light into the pellet. The amount of energy that actually reaches the pellet is on the order of 10s of kJ. My understanding is that a direct drive method (literally shining the laser directly onto the target) wasn't feasible when the project was concieved. Of course, it should be noted that a commercial reactor would very likely be using a direct drive method, which represents another massive efficiency gain.

A hypothesis for how viruses evolved:

At some point, there was a cellular organism that had evolved a unique form of parasitism. It would latch onto another cell, inject a copy of its DNA (and probably some other things to help support the process) into the target, then move on. The target, having had its genes replaced, would morph into the same species as the parasite.

As time goes on, innovations are made to this strategy. Chief among them is the creation of 'spores' that could carry out this process away from its progenitor cell. Since now only spores needed to be created for genetic propogation, rather than a whole cell, genetic propogation became decoupled from species propogation.

From here, propogation accelerates. An evolutionary 'race to the bottom' in terms of genome size occurs, as well as a massive adaptive radiation to occupy a whole world of previously unheld niches. Since these new species are so much more efficent as genetic parasites, the strategy used by their progenitors is no longer viable, hence why we don't see any such strategy in use today (alternatively, it could be that such a strategy inevitably undergoes such a race to the bottom very soon after emerging).

Of course, actually testing this hypothesis is basically impossible, since viruses mutate so quickly that phylogenies are difficult to establish at the best of times.

I notice that a lot of people reject the idea of major technological progress happening in the future. Specifically, they tend to reject disruptive technological progress; these same people won't object to the idea of high end telescopes getting better or transistors getting smaller, for example (up until the point that the improvement inherently implies a disruptive technology, of course).

Addressing some of the more common themes argued from this perspective:

• Pointing to limitations in current iterations of a technology and extrapolating to the future, even though it isn't at all obvious that these limitations will remain. Especially egregious if there heve been regular improvements to those limitations recently.
• Pointing to past failure/lack of development and extrapolating to the future, i.e the 'always 20 years away' arguement. Literally every technological innovation refutes this: it doesn't work until it does.
• Pointing to the limits of naturally occurring equivalents to the technology and assuming that those are hard limitations. Counterexamples abound, one that immediately springs to mind is light harvesting, which natural systems do with 6% efficiency maximum, several times lower than even decent photovoltaic cells. Not to be confused with hard physical limits, such as the speed of light.

As for why they are like this? Well, it is usually ideological in nature, with a side of general resistance to change. After all, it's a lot easier to reject something (that hasn't even already been done, mind you) as impossible than to accept it and reanalze your worldview in light of such.

This is a demonstration of how an apparently complex object can actually have a low Kolmogorov complexity.

For those unaware, an object's Kolmogorov complexity is the size of the smallest program needed to generate that object, and is thus the most compression that object can undergo before losing information.

Take a cellular automaton that exhibits explosive growth, let's say it's rule B1357/S2357 for this scenario, take a small starting pattern in an unbounded universe, and run it for, say, 10 trillion generations. The resulting pattern would require a significant fraction of the world's total digital storage to be stored, even when subjected to contemporary compression algorithms.

This object has a very high apparent complexity: it doesn't look completely random, but it's close (note: randomness is a sign of high complexity precisely because there are fewer underlying, simpler patterns). And yet its Kolmogorov complexity is much lower than its apparent complexity; it's just the size of the smallest program that runs a certain pattern in B1357/S2357 for 10 trillion generations, which is probably smaller than 10^24 bits (a rough estimate of the size of this object uncompressed).

It would be very easy for me to go from here and write about how computation is able to produce unbounded complexity, and what that and the previous text imply about the nature of emergence, but I won't do so for now.

Probabilities aren't very useful if you don't have a familiarity with the domain the probabilty is describing. To illustrate what I mean here, I'll give a few examples.

Consider a mechanism that is an integral part of some larger machine. Say that it is expected to last 50,000 cycles, and that you want to display the probability of it failing.
One way of doing such would be to give the probability of it failing on any given cycle. In this scenario you will notice that the majority of the probability space (0.01 to 1) describes the part failing, on average, within 100 cycles; outcomes within even an order of magnitude of what's expected are concentrated on 1 extremum of the probability space.
Another way could be to give the probability of failure in 5 billion cycles. Same problem, just on the other extremum. The obvious solution is to give the failure probability in 50k cycles, but that misses the point: for any given probability space, the extrema will be more 'concentrated' than the bulk of the space.

And it can get much worse if the given probability doesn't directly match with the quantity information is desired about, as this opens the opportunity for a range of probabilities to correlate with outcomes in unexpected ways. Here's an example to help build intuition.

When browsing the internet, I will often open a new tab if I see something interesting while I'm busy to look at later, and closing them when I do. If it just so happens that I don't get around to doing so? Well, the tab stays open; I am a tab hoarder. This, of course, is the superior way to browse the internet: you can open up multiple lines of research, investigation, etc without abandoning all of the rest, you learn to prioritize or else face insurmoutable clutter, you build up a record of your interests over time... you get the point.

Nonetheless, over time, the tabs tend to build up. I've been building up this current tab backlog for 4-5 years now. Needless to say, it would be a good idea to go through and clear some of those out, which I have been doing for the past day or so. I just finished my first pass, where I closed tabs that I had lost interest in, I had already browsed at one point elsewhere, or were subject to link rot. A few observations:

• Tab groups are very useful for resolving clutter.
• Youtube was, by far, the most susceptible to link rot out of everything. Over 90% of the link rot in at most half of the tabs I had open. The rest of the rotten links were from twitter.
• Neocities websites (and forums to a lesser extent) were hard hit by Discord's recent changes to image hotlinking.
• I lost interest in a lot fewer of the tabs than I had expected. Unfortunate, as that means that I have a backlog of a dozen hours of video and several million words of text to go through.

In regards to the Bouba/Kiki effect (for those unfamiliar) , I think it has more to do with how the brain processes information in general than anything specific to language processing.

Long, pointy objects tend to make sharp, higher-pitched sounds (think of two knives sliding together) and round, blobby objects tend to make soft, lower pitched sounds (think of a pillow hitting the floor). These match how 'kiki' and 'bouba' respectively sound. At least in English, there is another sensory correspondance: the word 'kiki' has more spokes and less rounded edges, while 'bouba' has less spokes and more rounded edges.

To me, it seems like this effect is a natural consequence of how the brain works: it draws on past experiences to guess at the properties of the drawings, notices similarities between certain aspects of those experiences and the words, and matches based on that.

Also, my best minesweeper time is now 92.

Something terrible has happened: I've started to play minesweeper. This is bad because I got carpel tunnel the last time I did that (also, it's eating into my already limited pool of productivity). My best time is 110 btw

I had a list of things I wanted to write about here that I put together during my impromptu hiatus, but I accidentally deleted it. This isn't so bad, since I still have a few ideas and can throw another one together in about an hour, but it's nonetheless a setback.

And, to close this post off, a point of frustration: a lot of websites today automatically replace emoticons with emojis. If I'd wanted to use an emoji, I'd've used an emoji; please don't force it on me. Fortunately, it can be bypassed by inserting a zero-width (somewhat easily typed by pressing Alt + + 2 0 0 b, input using your keyboard's numpad when possible) space in the middle of the emoticon.

Hello! I'm not dead!

Sorry for not updating for a while, but I have been doing things. I'll detail those below.

First, and the biggest one: I've decided to start learning Japanese. I've low-key wanted to do this for a while, but hadn't seriously considered actually doing so until recently. At this point, I can read basic sentences with common words unassisted (slowly) and more complex sentences with software assistance (a lot more slowly). I will be doing a writeup on my learning experience once I become more proficient.

And... that's about it. Yeah, learning a language has been a time investment, but not so much of one that it's taken up my free time (though I'd probably be learning a lot faster if I were spending that sort of time on it); I'm only spending 1.5 to 2 hours per day on it (honestly closer to half an hour these days since I haven't replaced the old excercises I have finished yet). I vaguely remember resolving to improve my typing speed at around the same time, but that evidently fell through quickly. I was doing stuff with Multistate INT CA, too, but that was very much an on-and-off sort of thing, and I haven't done anything with it since December anyways.

I have been doing a bit more more recently, though. I finally resolved to start documenting a worldbuilding project I've been tossing around in my head for the past few years (don't expect anything to be published here on this for a while). I was also inspired by a certain video to make a base-2(ish) notation system that I actually finished, which I will be producing a writeup for so long as I can be bothered to make the supporting visualizations to help explain why what I did for it is so neat.

I do think I will be updating a bit more frequently from now on, so you can look forward to that (don't get your hopes up).

Sorry for not updating recently. I haven't been active because my life has become a bit more busy recently and I haven't thought of much to write about.

I decided to do this update now because I had a dream last night about updating here. When I checked in today, I also noticed that the site has had almost 100 visits, so it's a good time to post regardless. Consider this post a commemoration of such.

I also noticed that I acquired my first two followers in the interim. If they're reading this: Hello! Thank you for being interested enough in what I have to say to follow. I haven't checked your pages yet, but I'll definitely do so soon.

I just noticed that a twitter account that I really enjoy was recently reinstated after months.

They post about several things, but I am following him mainly for his in-depth discussions on Mesoamerican history. He has posted quite a bit on the subject over the years; I'd highly recommend scrolling through his account for that. He also contributes to the Ancient Americas youtube channel.

I had already been falling off of twitter by that point (albeit I never had used it overly much in the first place), but his account being banned was the nail in the coffin for me in that regard. Even though he is back now, I probably still won't be using twitter as much as I had been.

I came up with an experiment for Machine Learning a few years ago, but never bothered putting it anywhere. Since I can't think of anythink else to quickly write about at the moment, I'll document it here.

The experiment goes as such:
Start with 2 neural nets. The first one takes an image and outputs n characters of text. The second one takes that text and a large set of images, one of which will be the same image as was given to the first one, and outputs an index. The goal is that the index output by the 2nd neural net always match the image provided to the first.

Repeat this for several values of n. When this is done, analyze the text output by the 1st neural net for each one and try to find any structure.

Since I have neither the experience nor the resources to conduct this experiment myself, it will be up to someone else to conduct this experiment.

This got longer than I expected, so I made it into its own writeup.

I saw this page when surfing neocities recently and was hit by a blast of nostalgia.

You see, I had encountered the flamingchickens LTE (the first of its kind, and by far the most popular, though that's a relative term in this case) in a random google search (something along the lines of 'the world's longest webpage') in the early 2010s (zero chance you could ever stumble upon something like it today with google) using the 3DS web browser. Having started using the internet not too many years before, some time after the death of the personal webpage in terms of influence, I hadn't seen anything quite like it before, so it stuck in my mind.

As for what it was: An LTE was a long-winded, stream-of-thought type of blog done by the flamingchickens webmaster and others in her friend group. And that's all it would've been, but apparently others had encountered it rather recently and decided to do their own take on the idea.

"Yeah yeah," you may be saying, "But is it, in fact, the Longest Text Ever™ (which we will define here as the largest amount of coherent text on a single webpage)?" Well, no. The whole thing comes in at just over 38k words, which is a lot, but you will occasionally run into longer chapters (which are almost universally displayed on a single webpage) in fanfics and webnovels, at the very least.

ミツバチ

I recently found a video and was reminded just how much I hate it when people try to specify a pronounciation without using the IPA. Please, please use it if you don't want to make people (specifically me) die inside reading it.

My own experience learning the IPA was as follows:

• Read the Wikipedia page.
• Used this website to learn what each symbol sounded like.
• Start making those sounds to myself, especially when I saw the corresponding symbol.
• Spent like a week trying to pronounce the sounds I couldn't (/r/) in the shower until I got them.
• Subconsiously muttered random consonants under my breath for years afterword.

Go do this. Seriously, you don't know what you've been missing out on.

I made a new page! It's been about a month since I last updated this website (I was busy with a project for one of my other hobbies; here for those curious), so I figured I may as well make up for the lack of progress a bit.

This page is for entries that are more casual and/or too small for their own writup.

In the interim, I found out that a certain element on the main page displays slightly differently on a different OS. Somewhat concerning, but I have no clue how to go about fixing that, so ¯\_(ツ)_/¯

I've also found that different fonts are rendered slightly differently between different browsers. For example, I tried using Times New Roman on this page: it looked really nice on Firefox, but not so much on Chrome. I should probably investigate how to detect the browser a visitor is using...