AI users are Fake-Nerds
👁 7
https://social.linux.pizza/ : 👁 1
https://blenderdumbass.org/ : 👁 2
![[avatar]](/pictures/favicon.png)
by Blender Dumbass
Aka: J.Y. Amihud. A Jewish by blood, multifaceted artist with experience in film-making, visual effects, programming, game development, music and more. A philosopher at heart. An activist for freedom and privacy. Anti-Paternalist. A user of Libre Software. Speaking at least 3 human languages. The writer and director of the 2023 film "Moria's Race" and the lead developer of it's game sequel "Dani's Race".
This is just a thought I had the other day thinking about why I dislike AI so much, while also having a problem with the concept of copyright. Strange, right? I both hate copying restrictions on things I want to copy or share. And both hate biggest copyright attacker in the history of copyright attackers: AI. How's that? Well I hope my "Fake-Nerds" analogy will be helpful. ↩ Reply
What is a Program?
I like computers. Computers run programs. Little pieces of information designed to do certain tasks with other pieces of information. In a nutshell a program is a set of instructions. Sometimes those instructions are to wait for a user to do something. And then do something based on what user have done. ↩ Reply
You are probably reading this article through a web-browser. A web-browser is a program with a certain web-browsery set of instructions. First the instructions say to open a window on your computer, draw the url bar on the top of the window, draw various buttons to either side of said url bar. And wait for the user to do something. ↩ Reply
When you type in
blenderdumbass.org into the url bar, to go to this website, the web-browser has a different set of instructions. It is instructed to download the webpage from the internet and then it is instructed to render said webpage in the rest of the browser window. And then it is instructed to wait for your input again. If your input is to scroll, it is instructed to scroll the page. If your input is to click a link, it is instructed to download and then render that page you clicked on. And so on and so forth.
↩ Reply
Obviously I'm simplifying. I didn't talk about DNS, or HTTPS, or JS, or any other ass, in this explanation of mine. But you can imagine, there is a lot of various instructions to handle all those various things. ↩ Reply
A nerd would want to know more. A nerd would want to know about DNS. A nerd would want to know about HTTPS. A nerd would want to know about JS and then would want to forget about JS because it's an ass. ↩ Reply
A nerd would want to know more and will know more. And then a nerd will be a person that knows enough to be able to develop it further. This is how we used to get progress in technology. Somebody would learn about something enough to see where it can be improved. And that somebody would be knowledgeable enough and smart enough to figure out how to improve what needs to be improved. ↩ Reply
On that basis alone in 50 years we got from primitive console machines to systems that are fun, convenient and easy to use, by pretty much any person on the planet. Thanks to the nerds we got to live in a digital world full of interesting things. c:0 ↩ Reply
But maybe nerds made a mistake making computers so fun and so easy to use. I mean, not-nerds got interested in them. ↩ Reply
What is a Machine-Learning algorithm?
A program is a set of instructions. Usually a program is a set of instructions that is trying to solve some problem. ↩ Reply
If your program in question is a web-browser, it is trying to solve a problem of making it possible for people to view web-pages. Obviously the web-browser is not the only program that solves this problem. You have DNS and HTTPS and other ass connected to it, to solve that one problem. So sometimes you have multitudes of programs working together to solve a problem. But you get my point. ↩ Reply
Say you want to multiply two big numbers. It is a problem. You could take a piece of paper and try to solve it the old good way. But there is software for that. Pretty much any computer comes with some sort of calculator that can solve that problem. ↩ Reply
Long time ago nerds figured out how to use electricity to solve this problem. And then, as soon as this problem got solved, they used their solution to build more and more complex programs that solve more and more complex problems. c:2 ↩ Reply
Say you want to play a racing game. That is a much more complicated problem than just multiplying 2 numbers together. Here you have to have a model of the street and a model of a car. And some program that can take those models and render them onto the screen. There also should be a way to make those models. Making them by hand is probably possible, but using a computer program that aids the artist in the creation of the model, would be much more preferable. And so to solve this one thing, you have a 3D modeling software, and a 2D drawing software, and the game logic software, and some sort of rendering software, all working together to solve the problem of you wanting to play a racing game. ↩ Reply
But say there is a problem so beyond an easy solution that to even start to think about how to solve it, would be discouraging enough, that nobody will want to solve it. ↩ Reply
Say you want to a make a program that translates from one language to another. You can imagine making a program that takes every word and replaces it with a word of a similar meaning in the other language. You could make a database of meanings of words ( like a dictionary ) and run your program through that dictionary for every word, to get a translation. ↩ Reply
Let's try it. Let's translate this to English: "И так. Я сейчас пишу статью про тот как мне ненравится искуственный интелект." c:3 ↩ Reply
If you replace every single word it would read something like this: "And so. I now write article about that how to me not liking artificial intelligence." ↩ Reply
It should be actually more like "So. I write an article about how I dislike artificial intelligence." ↩ Reply
As you can see just replacing words does not solve the problem. It gets us close enough, I suppose, to being able to understand the sentence. But not enough to be a good translation. ↩ Reply
A nerd way to do a good translation would be linguistic. A nerd would create a set of very complex instructions based on the actual real linguistical science about the 2 languages in question. It would parse one sentence using this linguistical knowledge, to arrive at some sort of direct meaning. And then it would encode, using the linguistical knowledge of the second language, that meaning, into that second language. ↩ Reply
But imagine how complex it would be to make such a program? I mean it would be amazing. In the source code of those instructions, there will be a very applicable, very real, linguistical science. But it would be a pain in the ass to get all this knowledge in the first place. Let alone to translate all of it into software source code. c:4 ↩ Reply
Nerds would go there. They would do it. I even believe some actually did, or tried to do it. But not all people are nerds. And with technology becoming mainstream, people needed this sort of universal translator more and more. So not-nerdy, executive types at companies like Google, decided to provide a solution. ↩ Reply
Instead of improving the science of linguistics, they decided to cheat the system. They decided to use a much simpler statistical analysis algorithm on a lot of text in multiple languages, to statistically arrive at a plausible model of how sentences in language A, should look in language B. ↩ Reply
This yielded some sort of set of instructions that could be called a program. And that runs like a program. A program that does the task. A program capable of translating a sentence from one language to another. But a program that is useless for the nerds. These instructions work, but they don't mean anything. ↩ Reply
The machine learned how to translate. But the learned set of instructions is a garbled mess of unintelligible nonsense. Is there an equivalent of HTTPS, or DNS, of even JS in there? No-one knows. If it is there, it has no name. Only function. Where is it? Hard it know. There are no labels, no comments and no source code. Just pure function. A black box. You pry it open and you see a mess of data so tangled together, it is impossible to read, impossible to understand, impossible to improve upon, impossible to nerd about. But it works. And for most people it is enough. c:5 ↩ Reply
What is an AI generated program?
And so here we are, years later. And nerds are no longer relevant. Any average Joe can "vibe-code" these days. Open ChatGPT or Claude or who-knows-what and let the "program", which was "trained" on a lot of examples of what programs should look, "figure out" how to "solve" the problem you are asking it to solve. ↩ Reply
Who knows anything anymore? ↩ Reply
When some douche-bag asks ChatGPT to write him a program, or to fix a bug in some existent code-base, there is nothing there. ChatGPT itself, like the translator in the previous chapter, is an unintelligible mess of data generated by a statistical analysis algorithm. That itself generates something that pretends to be intelligible, just so our douche-bag would not need to feel like he is becoming a nerd. c:6 ↩ Reply
Learning is for pussies now. Knowledge is for idiots. The true "alpha" is offloading all of his thinking to a machine that doesn't know nothing. Is the result any good? If it works, nobody fucking cares. ↩ Reply
Okay, at least with the slop vibe-coding generates, the resulting "code" is something a person can read. It is not the completely unintelligible black box, that a machine-learning algorithm is. So a nerd could, in theory, look at the slop and figure it out. And try to improve it. But a nerd could improve something much better and with far less fart-noises if the damn slop-spaghetti box from hell, was avoided in the first place. ↩ Reply
There could be good uses of a machine learning algorithms. Translators or Denoisers are nice to have around. It would be superior if some nerd would actually figure out how the damn things work, and would re-implement them in code that can be understood. In sets of instructions that could be comprehended. It would be nice. It would be amazing. c:7 ↩ Reply
But a machine learning algorithm designed itself as a machine learner? A black box designed to spit out even blacker boxes? A fucking generative Idiocracy-inducing fucktoy from hell? What even is this? ↩ Reply
If you are calling yourself a nerd and use this shit, or even worse, rely on this shit, you are fake nerd, god damn it! c:8 ↩ Reply
Happy Hacking!!! ↩ Reply
Find this post on Mastodon
![[avatar]](/pictures/user_upload/Troler/55NKGQKE7IRKMLHV.png)
Troler
c:0 June 09, 2026 On that basis alone in 50 years we got from primitive console machines to systems that are fun, convenient and easy to use, by pretty much any person on the planet. Thanks to the nerds we got to live in a digital world full of interesting things.
⤴ ViewHardware wise, yes. But software? I wouldn't call LiGNUx by means a modern system, it is for modern hardware, but still very much archaic. Same goes for Windows and even MacOS.
... replies ( 1 )
Blender Dumbass
c:1 June 09, 2026@Troler It is not perfect yet. But it is not getting closer to being perfect if people think ChatGPT is a good idea.
Blender Dumbass
c:1 June 09, 2026... c:0
Troler c:0 June 09, 2026 On that basis alone in 50 years we got from primitive console machines to systems that are fun, convenient and easy to use, by pretty much any person on the planet. Thanks to the nerds we got to live in a digital world full of interesting things.
⤴ ViewHardware wise, yes. But software? I wouldn't call LiGNUx by means a modern system, it is for modern hardware, but still very much archaic. Same goes for Windows and even MacOS.
@Troler It is not perfect yet. But it is not getting closer to being perfect if people think ChatGPT is a good idea.
Troler
c:2 June 09, 2026 Long time ago nerds figured out how to use electricity to solve this problem. And then, as soon as this problem got solved, they used their solution to build more and more complex programs that solve more and more complex problems.
⤴ ViewFirst they used gears. From the Antikythera mechanism, through the differential analyser all the way to Z3. After Z3, computers became more and more dependent on electricity.
Troler
c:3 June 09, 2026 Let's try it. Let's translate this to English: "И так. Я сейчас пишу статью про тот как мне ненравится искуственный интелект."
⤴ ViewAnd so. I'll now write statyo about how I don't like artificial intelligence.
Troler
c:4 June 09, 2026 But imagine how complex it would be to make such a program? I mean it would be amazing. In the source code of those instructions, there will be a very applicable, very real, linguistical science. But it would be a pain in the ass to get all this knowledge in the first place. Let alone to translate all of it into software source code.
⤴ ViewThe system could work, but it would not be flexible. Language is inconsistent and if you want a better solution, you need a more so hands-off approach. When working with messy data, you need something good enough with approximations.
Troler
c:5 June 09, 2026 The machine learned how to translate. But the learned set of instructions is a garbled mess of unintelligible nonsense. Is there an equivalent of HTTPS, or DNS, of even JS in there? No-one knows. If it is there, it has no name. Only function. Where is it? Hard it know. There are no labels, no comments and no source code. Just pure function. A black box. You pry it open and you see a mess of data so tangled together, it is impossible to read, impossible to understand, impossible to improve upon, impossible to nerd about. But it works. And for most people it is enough.
⤴ ViewThe last part is the problem with design. When you create a system that's opaque to its logic, there's a huge problem. Imagine if it was impossible to open the hood of the car and when it broke down, you'd need to buy the entire replacement of the hood. Not good, hackers, not good.
Troler
c:6 June 09, 2026 When some douche-bag asks ChatGPT to write him a program, or to fix a bug in some existent code-base, there is nothing there. ChatGPT itself, like the translator in the previous chapter, is an unintelligible mess of data generated by a statistical analysis algorithm. That itself generates something that pretends to be intelligible, just so our douche-bag would not need to feel like he is becoming a nerd.
⤴ ViewChatGPTs are being marketed as cure for all diseases, while in reality they're just good prediction models. Which is very beneficial for compression, searching through data and other tasks involving finding fuzzy information. Is it bad that there are programs that automatically tag images with tags, depending what's depicted in them? As long as the processing is local, I don't see too much of a problem there. But expecting such a system beat human tagging is just silly. At least for now.
Troler
c:7 June 09, 2026 There could be good uses of a machine learning algorithms. Translators or Denoisers are nice to have around. It would be superior if some nerd would actually figure out how the damn things work, and would re-implement them in code that can be understood. In sets of instructions that could be comprehended. It would be nice. It would be amazing.
⤴ ViewHow do you view programming language translators?
Troler
c:8 June 09, 2026 If you are calling yourself a nerd and use this shit, or even worse, rely on this shit, you are fake nerd, god damn it!
⤴ ViewIs Voice-To-Speech considered also bad?
![[thumbnail]](/pictures/thumbs/fork.png)
![[thumbnail]](/pictures/thumbs/where_you_draw_line.jpg)
![[thumbnail]](/pictures/thumbs/fa.jpg)
![[thumbnail]](https://peer.madiator.cloud/lazy-static/thumbnails/c07756b1-ce79-4c8a-ab26-056704c77915.jpg)
![[thumbnail]](/pictures/thumbs/ai.jpg)
![[thumbnail]](/pictures/thumbs/upbge_ai.png)