A slightly more serious blog than a big rock

The Enigma of Amigara Fault

This short story by Junji Ito is about a fault that appears in Amigara mountain after an earthquake. The earthquake exposes countless human-shaped holes in the mountain which seem to have been made about a thousand years ago. People, intrigued by these  silhouettes, gather at the site and that’s when things get creepy.

“Tbh I’d probably just”

the last time I reblogged this post right before I got a great job, in a permanent work-from-home position, with benefits, retirement, and a salary literally 3x what I was making before, doing something I really like. 

So you know. 

Anyway, the next night we watched a thing about some pig man who is supposed to live in Vermont. People said it had claws and a pig nose but walked upright like a man. Now, I happen to know that sideshows used to shave bears and present them as pig men. So every piece of evidence they gave of this monster sounds to me like a bear with mange.

So now the running joke in our house is that everything is bears. Aliens? Bears. Loch Ness monster? Bear. Every cryptozoological mystery is just a very crafty bear.

Act 1: JJ Abrams does his mystery box crap to avoid actually writing a story

Act 2: Rian Johnson sees what he is doing and smashes his mystery box with a hammer.

Act 3: JJ Abrams gets mad and pretends he had something inside of the box the entire time (he didn’t)

The British government was unwilling to do anything for two reasons. First was the laissez-faire capitalistic ideology, that put the rights of property owners to make profits above human lives. Rent freeze was unthinkable and they even were unwilling to do proper relief efforts as free food would lower the cost of food. The second reason was disdain for the Irish, and the thought that they were “breeding too much” and the famine was a natural way to trim down the population, aka genocidal reasoning.

This is why it’s important to stress it was not a potato famine. The potato blight was all over Europe but only in Ireland there was a famine. The reasons behind it had nothing to do with potatoes and everything to do with the British.

Apparently what made Choctaw want to offer relief to Irish was the news about the Doolough Tragedy. Hundreds of starving people were gathered for inspection to verify they were entitled to recieve relief. The officials would for *some reason* not do that and instead left to a hunting lodge 19 kilometers away to spend the night and said to the starving people they would have to walk there by morning to be inspected. The weather conditions were terrible and many of them died completely needlessly during the walk thorough day and night.

This apparently reminded the Choctaw of their own very recent (and much more explicit and bigger scale) experiences of ethnic cleansing, where they were forcibly relocated. It was basically a death march and thousands of Choctaw died from the terrible conditions also completely needlessly.

In 2015 a memorial named Kindred Spirits was installed in city of Midleton in Ireland to commemorate the Choctaw donation.

And here is an actual source because y’all never have them and that’s why you spread so many myths: https://www.wesleyjohnston.com/users/ireland/past/famine/summer_1847.html

silksong telling us that hornet really likes fluffy bugs and wants to pet them is the best thing to come out of this game

You may ask: why not just sculpt a horsie! You can sculpt! And that is not the point for the people who do this. The point is to WARP and BOTHER the material I think.

Many, many people read this post addition and wanted to correct it with: "well actually, it's actually much easier to resculpt a horsie than to make one from scratch. And also, Breyers have an active community, and it's nice to be in a fandom."

Many of these corrections were well-intentioned, aimed at general readers, not aimed at me personally, and were intended to share knowledge. I am not singling anyone out and it's great to see how many people have passion for a hobby.

When people kindly messaged me to explain, in detail, the artistic intentions of the model horse customisation community, I said merely, "Thank you for your kindness in telling me." However, to me, this is a really educational example of a post and how we can use the information in posts. There is no information about what is going on in the original video or post.

To the post, I added the following information:

identified the brand of model horse as "Breyer," from less than 2 seconds of video showing the original horse.

identified the hobby being shown as "Breyer customisation."

identified the material used to construct Breyer horses as "resin."

noted that the general public may well ask why people do this hobby, instead of sculpting their own horse.

said that "sculpting their own horse" is not the point for people who do this hobby.

added jokingly that "bothering" the original horse is a source of satisfaction for people who do this hobby.

noted that custom breyer competitions exist, indicating, at this point, not only knowledge of a niche hobby, but knowledge of its community, and FURTHER, what its elite practitioners within the community do at the pinnacle of taking it seriously.

used a strikethrough for this text, indicating a certain sheepishness about knowing this at all.

Now, let's look at the information imparted in the seven sentences in my original post addition, and what they mean. We can see how each of the seven sentences contains information, and point 8 of the meanings has meta-information (here, a possible interpretation is that the person is embarrassed, or joking about being embarrassed, at publicly admitting the depth of their knowledge). When we think about what these sentences MEAN, it looks like the person writing them has some information about this hobby.

Now we can reflect on a few questions.

From the post, do you think that Elodieunderglass has ever heard of model horse customisation?

From the post, what level of knowledge do you assume Elodie might have about it?

Do you think that "why not just sculpt a horsie" is Elodie's question? Do you think that they are genuinely framing this as a request for information?

Based on the post, do you think that Elodie provides an answer to the question in the next few sentences? Do you think that Elodie needs an answer?

What level and depth of information do you think Elodie needs to receive in order to truly understand the hobby of model horse customisation? Beginner/intermediate/advanced?

What tone indicators can you see in the post? Do you think it is serious and literal, or joking and playful? What text is present? What is the subtext?

If, reading through, you start feel hot and defensive and say, "Well, I'm sorry, but just excited to share about my hobby," here is another way to frame that.

This person has indicated that they probably share your hobby. It's entirely possible that you could start from a different point, and end up with an entire friend who shares your hobby, instead of mildly irritating a peer.

Instead of using your knowledge to hit someone with a sledgehammer of a lecture, you could use your knowledge to reach out your hand to ask.

"I don't always read tone well, but I feel like you have some knowledge about Breyer customs! Do you do them? How serious are you about it? Would you like to see any of mine?"

I am extremely sympathetic to the excitement and possessiveness of a special interest or hobby. Here are three further pieces of advice.

Having a special interest or passionately held hobby does not mean that you hold more knowledge than others. And that's fine - the people with equal knowledge are your community.

People may occasionally use a joking tone about their own interests and hobbies. People may also be frequently sarcastic or critical of their own hobbies - often because people who hold a lot of knowledge about something are the most aware of the problems/criticisms of the Thing. Other times, affectionate joking is a way to express affection for the Thing.

If you are normally "bad at reading tone indicators," that's fine! However, because other people frequently use tone to communicate, being "bad at tone indicators" should be a mental save point - a clue to read especially thoroughly for tone before reacting. Skimming and reacting based on what you THINK the tone was, or what you THINK the text said (based on the tone you interpreted), is a recipe for really irritating people.

Okay, not to worry, back to everyone's regularly scheduled horfposting. Just.... yeah.

here is a more in-depth source from the Washington Post; here it is without a paywall

The Washington Post article's graphic states 'A 100-word email generated by an AI chatbot using GPT-4 [...] Once requires 519 milliliters of water, a little more than 1 bottle [...] Once weekly for a year requires 27 liters, about 1.43 water cooler jugs [...] Once weekly for a year by 1 out of 10 working Americans (roughly 16 million people) requires 435,235,476 liters, equal to the water consumed by all Rhode Island households for 1.5 days."

Methodology of the study: Water and electricity costs were calculated by Ren for ChatGPT-4 at an average American data center. A full methodology can be found in their paper “Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models.” D.C. electricity consumption figures calculated from the U.S. Energy Information Administration’s 2023 state energy reports. Rhode Island daily water use from the National Environmental Education Foundation. Beef and rice per capita consumption numbers from U.S. Department of Agriculture food availability data and the OECD’s agricultural outlook. Water footprint estimates sourced from: Livestock Science’s “Assessing water resource use in livestock production: A review of methods” and Science’s “Reducing food’s environmental impacts through producers and consumers.”

In addition, source "C" from above states "In a paper due to be published later this year, Ren’s team estimates ChatGPT gulps up 500 milliliters of water (close to what’s in a 16-ounce water bottle) every time you ask it a series of between 5 to 50 prompts or questions."

Lastly, I would argue that the water and electricity wastes of AI usage are drastic and intense. It's not just that using generative AI to produce AI-generated text wastes water and electricity (see the Washington Post article for electricity data), but training large language models (LLMs) also uses a vast amount of water and electricity.

For example, "In their study, Ren and his team – based at the University of California, Riverside, and the University of Texas at Arlington – estimate that training GPT-3 could have consumed as much water as required to produce 370 BMW cars or 320 Tesla electric vehicles." This study's findings are also reported here and here.

On top of all of this, AI models are often trained using underpaid workers (see more here, here, here, and here.) And the water used when you use AI to generate a 100-word email is unnecessarily wasted. You could have just written that email yourself, whereas you genuinely need to use water to do things like taking a shower or watering your plants. It's a bit of a false equivalence.

Regarding training, it is a significant cost, but it’s important to note that it’s an investment that is amortized as it’s used. Developing *anything* has an overhead cost, but the most people use the thing, the less the cost is per use. Training shouldn’t be compared to producing producing a model of a car, but *inventing* a new car model. Honestly I’m surprised it only took as much water as producing 370 BMWs to train GTP-3, I thought it was much higher, but maybe I’m just underestimating how much water it takes to make a car. When comparing the amortized costs, that many cars can only service like a few thousand people tops, while GTP-3 easily had millions of users. I do think there are issues with the pace at which new models are being trained and the plurality of redundant models being created out of market competition, but the usage of LLMs in and of itself is not unsustainable.

Not going to comment on data anotators being underpaid since underpayment is itself such a loaded concept - people don’t agree on how to determine what fair pay is. Conversely, one can instead argue these created opportunities for people in need, with low pay being better than no pay. I recall discovering while working on my Master’s thesis that data annotation jobs for training AI significantly improved gainful employment for autistic adults.

Whether it’s unnecessary wasted is a matter of opinion. If there wasn’t some sort of personal benefit to using LLMs, people wouldn’t use them. Yeah, someone could write an email themselves, but generating one saves time and head ache - I don’t use LLMs for the task, but I found the autocomplete Google has included since before LLMs were a thing to be extremely useful for dealing with routine emails. Also, don’t forget that everything has a cost. Saving time not only gives more time to do other things, but if you save enough of it it actually ends up reducing energy consumption. I’ve seen claims that generating AI images can actually save energy costs relatively to drawing something in Photoshop (though of course that potential gain would also need to be weighed against the fact it’s soulless slop that likely imitates existing work without permission). I think generative AI is of questionable utility myself, but in terms of personal impact it’s small enough that people should be able to decide for themselves, and the scapegoating of AI serves to create division, demonize other legitimately useful applications of AI in the process, and distract from more pressing issues.

My AC keeps me - living in a tropical country with temperatures up to 40 centigrade - comfortable and ALIVE. Therefore I can work and make a living for myself without hurting anyone else.

My AC costs 700 watt-per-hour for my entire house. That's 16kw per day. ChatGPT expenses 0,34 watt-hours PER QUERY according to it's founder. 10 queries = 3.4 wH. I'm not counting that, because somebody else had. Oh look, it's a shitload. Per DAY.

Let me clarify that.

2. Your calculations are completely incoherent because you’re not comparing the same units. Kw/hour is not a unit any one uses, watts themselves are joules per second and joules is the ultimate measure of energy consumption. Most AC units for a home in the US are around 3000 watts, which if ran all day goes to 259,200,000 joules. An upper bound estimate for query cost (which is the one you’re referring to with three watt-hours) meanwhile is 10,800 joules. That means someone would need to makes 24,000 ChatGTP queries in one day to have the same energy cost as this person running their AC 24/7. That’s an absolutely ludicrous number of ChatGTP prompts, their AC habits clearly has a much larger energy burden than anyone’s ChatGTP usage.