Third-Gen Copies
AI, bootleg anime, and deep understanding
If you read the subtitle, you may wonder - “What the heck?”
Way back in the mid-80’s, anime wasn’t a “thing” yet. Akira hadn’t been released, much less Spirited Away or Cowboy Bebop. Sure, there were oddball derivative shows like “Battle of the Planets” that had been massively redone from the original, Star Blazers, Speed Racer, and a few other mecha shows I recall from my summers in New England, and then Robotech based on Macross, Southern Cross, and Mospeada, but Viz hadn’t even released it’s first manga adaptations as standard comic books yet (Mai the Psychic Girl, Area 88, and Xenon, Heavy Metal Warrior).
But, if you were into anime, various conventions had tables and booths, with merch, illustrations, and so on, and they and would show movies like Lensman and Project A-Ko. And you could pick up books from other cons with playlists as well as synopses.
And you could get hooked into a ring of people swapping copies of movies and shows, mostly movies and “OVA’s” (shorter animated pieces roughly an hour long). Someone, somewhere, would buy laserdisks - not cheap - and copy them to VHS. You could “buy in” by supplying a few blanks, and get sent a copy of something, and once you had a few, start trading single copies. In this manner I saw Honneamise, Crusher Joe, Lensman, Grey, Iczer-1, Bubblegum Crisis, Vampire Hunter D, Arion, Urusei Yatsura, Megazone 23, and a number of other movies and OVAs, which because they were cartoons and I was blessed and cursed with the usual X-er latchkey upbringing, meant I got exposed to not only some amazing artwork and storytelling, but a level of tits and extreme violence that would shock anyone who thinks “cartoons are for kids.” Things like eyeballs exploding out as a metallic tentacled thing rips through a starship bridge , said tentacles plunging through people and out of skulls. Yeah, that was Megazone 23 part II.
One thing that you looked for though, was generations of copies.
VHS tapes wore out. The read off of the original tape could be flaky if the heads were out of alignment, or any of the analog circuitry was “off”. Even without that read loss, you had signal loss patching into a TV or another VCR, and of course, recording quality.
In short, every copy was worse than the original. Every further copy of that copy was even worse still. If you wanted a copy that was watchable, and you wanted to be able to further trade it, you wanted to make sure that the generational count of the movie you were swapping for was as low as possible. And you wanted them made with decent quality VCRs.
Entropy is a bitch.
The more distant you are from the source, the more you copy a copy, the more derivative and degraded it gets. It’s not just the physical too - but inhabits the realm of ideas as well. On a literary level, it’s why so much fiction, especially genre, is slop. Star Wars was an homage to Flash Gordon and the like. To something vital from the pulp era. Vance, Zelazny, Cook, Merritt, and others borrowed directly from myth and legend and folk t tales, from ancient literature. Modern writers all too often borrow from these, already second-order sources, and are shallower for it.
How many Star Wars knockoffs and zombie tales do we need?
It also doesn’t help that in knowledge, much like copying media and conenction/equipment quality, the quality of your engagement matters as well.
In my time as a nuke, it helped I’m smart - though I thankfully had most of my overweening pride in the primacy of said smarts knocked out of me - but what got me through my advancement exams was a pattern of: read the book and manuals, and highlight anything of note. Read through it again, slower, and write out every important point in my own notes. review the notes. It didn’t surprise me much later to find out about a series of studies that showed students who wrote out their notes instead of typing them had consistently better measured retention, backing up decades of theory that the more ways you engage yourself, including physically with what you’re trying to learn, the better you will learn it.
It’s also why being fit is important - the body carries the mind. A healthy body keeps the mind healthy. It’s great to build a model of an engineroom in your head, but sketching it out carves it that much deeper in your memory. The better you can move, the better you can engage all of yourself and bring it into focus.
And it is also why, for real understanding, you don’t just read a summary of “what it is”, but try to go to primary sources, or at minimum, take it apart and see what makes sense, and what doesn’t. If it doesn’t make sense, is it because you don’t understand something? If it does make sense- how sure are you about your assumptions?
If you just take it in as something drifting by, you don’t learn it as well. Cliff’s notes may give you enough to sound like you understand the point of a book, but it isn’t the same as reading it. It is someone else’s understanding, and perforce, simplified, meaning that your understanding of the book is by necessity even more simple. You have a second-generation copy of the story and the ideas in your head, not the first generation copy based on reading it.
It’s what I’ve discussed before as “tribal knowledge” from my Navy days. There are a lot of nuances to doing maintenance, or starting up a plant, that just do not exist in manuals. In most cases, following the manual or developed procedures blindly as long as you have some mechanical understanding will still work out, because the manuals aren’t usually wrong, they’re incomplete1. They presume at minimum enough knowledge about how the systems involved work to not shoot yourself in the foot2, but how they’re integrated can change how you do things in ways that differ from the standalone manual.
My issues with LLMs or “AI” is less that it will “take our jerbs” - in all too many cases it will because bottom line and stock prices, and the result will be slop - but instead revolves around the question of what makes us human. If we are thinking animals, but we offload the thinking, are we still human? If the machines tell us to do something and we no longer have the personal experience and expertise to go “this is not a good idea,” where does that leave us? If all we’re doing is “recognizing the correct answer when told” even as experts, how long before we stop really taking a close look and understanding the result?
It’s not just that study with the programmers, where everyone expected, and felt like they were getting a 25% speed boost, but instead were programming slower, I’m seeing this more and more in people around me.
I’ve tried using LLMs for scaffolding and code stubs. The results were mixed. I especially loathed getting deprecated methods in the results because they were more common, especially when you prompted to use the current method.
On the one hand, one of the people I work with has written up powershell scripts that he otherwise never could have because he never really tried programming. On the other, basic things like sanitizing user inputs, validating within a dialog box, or generalizing calls as functions or iterating over arrays with sets of input values instead of repeating code several times for each case were utterly beyond ChatGPT or Copilot, and he of course didn’t have the experience to realize it. In another example, an experienced web dev I know integrating new features couldn’t really explain to me how it worked when I asked for guidance on it (I rarely work on web programming projects, so I was knocking some rust off) because he was just letting ChatGPT handle it.
And I’m getting tired of “chatgpt says”, or responses to questions that were obviously copypasted from a LLM prompt. The tone and style - I refuse to call it a voice - are distinctive, roundabout, overly verbose, and grating, yet only minimally informative.
I want to know what you think. If I wanted a summary of a series of reviews, message threads, and opinions, I want it to be at least something that you read, processed, and cross-checked. Not a LLM summary, or your take on a summary, or even a set of summaries.
Of course, as I’m finally hitting publish, I’m late to the party. I can especially recommend “AI Bugmen” because it does an excellent job of generalizing what I was discussing in part above - how recycling our inputs, or limiting them to a “safe” subset, results in the inability to deal with the world around us.
Meet Our Friends
The Last Redoubt is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Please also join the Pilum Press Discord and/or the Autarch Discord. Pilum is the publisher of several books and short story collections including Shagduk and Thune’s Vision. Autarch is the home of the Adventurer Conquerer King RPG.
That said, I’ve lost track of the number of times that the manual for the individual pump or motor states to do a procedure in a certain order, but the procedures developed for the system it’s integrated into tell you to do something different…
The less of that knowledge you have, the closer you get to the tech-priests of the warhammer 40k universe, repeating processes and chants who’s meaning is lost to time.
The VHS trading network sounds fascinatin, especially the whole concept of tracking copy generations. Its wild how that physica degradation mirrors exactly what happens with knowledge when we just read summaries or let ChatGPT do our thinking. Your point about writing notes by hand vs typing really hits home too. There's something about that physical engagment that makes things stick better.
I used to love Speed Racer as a kid and remember Battle of the Planets well. Sadly, saw on YT that America completely changed the episodes of Battle of the Planets.