Ontology talk:9k/RD/Q28,03: Difference between revisions
copy markup from Q28,03 |
m LLMs cause people to seek authenticity |
||
| (3 intermediate revisions by the same user not shown) | |||
| Line 2: | Line 2: | ||
== Main entry == | == Main entry == | ||
{{HueCSS}}<ol class="hue clean {{int:hue-ixn-classname-ML}}"><!-- inactive: {{int:hue-ixn-classname-DFy}} --> | {{HueCSS}}<ol class="hue clean {{int:hue-ixn-classname-ML}}"><!-- inactive: {{int:hue-ixn-classname-DFy}} --> | ||
< | |||
<li class=" | {{li|start=y|I=Z0/LLM|tradition=|Q=28,03|Q2=2803}}large language model | ||
{{li|I=Z0/LLM|tradition=|Q=28,03|Q2=2803}}large language model -> often seen as anti-humanities, although this contradicts with structuralist linguistics | |||
{{li|I=S1/ES|tradition=|Q=28,04|Q2=2804}}AI text generation / generating videos with large language models or advanced neural networks / AI-generated article (social phenomenon, Z0) | |||
{{li|I=S1/ES|tradition=|Q=28,05|Q2=2805}}AI image generation / generating images with large language models or advanced neural networks / AI-generated image (social phenomenon, Z0) | |||
{{li|I=S1/ES|tradition=|Q=28,06|Q2=2806}}AI video generation / generating videos with large language models or advanced neural networks / AI-generated video (social phenomenon, Z0) | |||
</li></ol> | |||
== Text generation == | |||
<ol class="hue clean {{int:hue-ixn-classname-ML}}"> | |||
{{li|start=y|I=S1/ES|tradition=|Q=28,08|Q2=2808}}reading arbitrary webpages and books into an LLM -> first of all, don't. second of all: the more interesting discussion here is what is being achieved or built up when somebody does this. if the machine isn't truly understanding it then what exactly did it use the data in the texts to create? my hypothesis is that it creates an ontology, while it is currently the case that humans can create ontologies better. the primary reason we haven't already built a good ontology is our obsession as human individuals towards Freedom and against [[Term:filtration|filtration]]. | |||
{{li|I=S1/ES|Q=28,30|Q2=2830}}companies buying subsidiaries to use their content as AI training data -> the biggest example is Elon Musk creating a new AI company to own Twitter. in this case, the whole process was controlled by one person. however, this process is interesting for the potential that separate billionaires could show up to buy companies and this could just become the new reason anybody ever buys companies any more. | |||
{{li|I=S1/LLM/Fy|Q=28,26|Q2=2826|submitter=Valenoern}}how many legs, again? / how many eyes does a human have? (generic; {{game|Deltarune}}) / ({{9k|VGriff/Q28,26}}) | |||
{{li|I=S1/ES|Q=28,09|Q2=2809}}jumping over paywalls with ChatGPT -> it's funny in such a dark way that this exists; if you understood what LLMs automate, you'd see it coming from far away. [https://lokeshchoudharyprogrammer.medium.com/how-i-read-medium-member-only-articles-without-paying-a-dime-762d90af2355] I refuse to do it. you can read a {{int:sitename/en}}<!-- LithoGraphica --> entry to get the same effect, or write one if you have access to the source, and more than one person gets to contribute to that. so, why does this exist? it exists because we've normalized an individual person with a lot of money buying an article being the only way to read articles. you know, a [[E:petty bourgeoisie|small shop]] putting out products assuming that everybody else has money to buy their products regularly when that might not at all be true. this incentivizes AI companies who are the only ones with money to send AIs to read everyone's articles, because even if they had to pay for the articles "legitimately" it would still be that they'd have the money and the readers wouldn't have the money. I am begging you if you have a Medium account with less than 100 followers to make your articles publicly available so an AI <em>doesn't</em> read them for people. ...this makes me realize. we should probably put in every single citation of a source whether somebody paid for it and a very vague idea of how much: new book, used book, paywalled article / paywalled or paid periodical. really, that should be in every single academic citation everywhere. we should push to get that into the official APA citation style guide to be frank. because that information is a fundamental part of publication, as much as the name of the publisher. field: graph economics. | |||
{{li|I=S1/ES|Q=28,10|Q2=2810}}Nebula is a subscription streaming service... -> and eventually only AI companies will be able to pay for it, reading all the stuff in it with their machines and spitting it back out at the people who can only pay cents for content through ChatGPT. probably in the middle of a huge number of ads. | |||
</li></ol> | |||
== Language models and plagiarism == | |||
<ol class="hue clean {{int:hue-ixn-classname-ML}}"> | |||
{{li|start=y|I=S1/HAS|tradition=|Q=28,01|Q2=2801}}plagiarism | |||
{{li|I=M3/MX|tradition=|Q=28,02|Q2=2802}}What is plagiarism? / Who owns the ability to repeat factual information? / Who owns the ability to repeat literary motifs? / Who owns the ability to independently repeat culture in another nation-state / Who owns the ability to independently repeat culture in another nation-state without paying another country that it happens also wants to overthrow your government and destroy national sovereignty? -><br/> | |||
everybody thinks they know what plagiarism is. anyone who owns a business or has a doctorate has absolutely no idea what it actually is or what it isn't; it's almost like the more educated you get the more confused you get about the question of plagiarism. here's the reality: the question of plagiarism is the question of what business territory owners will allow what other businesses or mere individuals to live and exist. that's precisely it. it's all up the whim of who likes who and who hates who. you're never guaranteed a license to exist even if you are willing to pay the money, it's all about personal relationships and court cases. there is no universal rule for what does infringe all copyrights or what doesn't infringe all copyrights. it's all about whether Bob wants Alice to be part of Bob's socially linked countable culture or wants to get rid of Alice. if Alice is Chinese, it's all about how much Bob wants China to exist or wants to wipe it off the face of the earth. | |||
{{li|I=F2/MX/ES|tradition=MX onto ES|Q=28,07|Q2=2807}}If a book is well-sourced, but written by AI, it is a reliable source / If a book looks entirely factual and traces all its statements back to a bibliography containing valid sources, but is written entirely by AI, it is a reliable source -> this is the huge problem with everyone's heuristics for what supposedly is and isn't a source. Wikipedia is generally going to be much more reliable than ChatGPT; the same machine beginning with the same data could output entirely factual statements, and total nonsense if that's the path you lead it down. but people have all been taught the heuristic that {{em|individuals}} are what make information reliable — if a book was written by an individual and it can't change, it's good, if it was written by multiple people and it can change, it's bad. this is a really bad rule given that instances of ChatGPT are individuals in the same sense that humans are individuals or house flies are individuals and they all produce static works that don't change. (yes, a house fly producing text would be hilarious, but that's almost exactly why LLMs are so widely criticized.) we've effectively taught everyone that ChatGPT is the {{em|easiest}} way to get high-quality information, while anything produced by humans is potentially unreliable. if you don't believe me... would you sooner believe a human-generated rant by me that I carefully checked against all my observations of material reality but didn't have time to add written sources to and maybe linked a video as a surface illustration of the idea, or a perfectly-formatted essay by ChatGPT that did end with a bunch of valid sources? yeah. think about that. we have a genuine problem.<br/> | |||
this is one of the several different reasons I have for creating this Ontology project. I think there needs to be a way for people to check the validity of statements without resorting to specific human individuals or specific books, in the way absolutely all academia is done currently. it's good to take a verifiable concept and put a bunch of examples of reliable books on the entry that illustrate it or provide accounts of material evidence. but books aren't what actually makes things true or accurate. what makes things accurate is their sheer coherence with [[E:Most science is not empirical|other testable understandings about material reality]]. that's why I am building this big bank of propositions. so you can take the most real and verifiable ones and use them to test the most dubious ones, whether you have books, whether you have experts, whether you can re-test observations of the material world positivist-style. this could one day be more reliable than Snopes because it wouldn't rely on special talented individual human experts versus just anybody who has a high school education and is sufficiently good at reasoning. assuming the claim you're testing isn't too new for all the required information to be recorded in here. | |||
</li></ol> | |||
== Text generation == | |||
<ol class="hue clean {{int:hue-ixn-classname-ML}}"> | |||
{{li|start=y|I=S2/ES|Q=28,82|Q2=2882}}LLMs cause people to seek authenticity / LLMs will cause people to seek "authentic" work with real reasoning, real voices -> this is probably true, but it brings up a lot of slightly different questions. do people always seek rare things, or is there a point where they will truly be content with common things? do people really want the human content because of its quality or are they actually looking for its rarity or novelty? | |||
</li></ol> | </li></ol> | ||
== Related == | == Related == | ||
<ol class="hue clean {{int:hue-ixn-classname-ML}}"> | <ol class="hue clean {{int:hue-ixn-classname-ML}}"> | ||
</li></ol | |||
{{li|start=y|I=S2/MX|Q=28,00|Q2=2800}}Every citation should contain price information / Every citation should contain cost information -> I really do mean every citation in the world, not just every bop-format citation. field: graph economics. | |||
{{li|I=S2/MX|Q=28,14|Q2=2814}}AI vision captchas hint at a new stage of capitalism / ({{9k|RD/Q618-LargeLanguageCaptchaNewProcessOfCapitalism}}) | |||
{{li|I=S2/MN|Q=28,15|Q2=2815}}Products are chunks' footprints / Products archaeologically trace chunks / A commodity is the trace of an isolated chunk of people engaged in industry -> this does technically apply here because the content of a particular passage of generated text says something about the team of humans that produced the original text. | |||
</li></ol> | |||
== Ideologies or fields == | == Ideologies or fields == | ||
<ol class="hue clean terse {{int:hue-ixn-classname-ML}}"> | <ol class="hue clean terse {{int:hue-ixn-classname-ML}}"> | ||
{{li|start=y | {{li|start=y | ||
|I=S1/ | |I=S1/ES|Q=28,03|Q2=2803}}{{TTS|ES|E-S}} / structuralist linguistics | ||
{{li|I=S1/LLM|Q= 618|Q2= 618}}{{TTS|LLM|L.L.M.}} / machine learning | |||
{{li| | {{li|I=Z0/LLM|Q=28,03|Q2=2803}}{{TTS|LLM|L.L.M.}} / large language models | ||
{{li|I=S1/MN |Q=41,01|Q2=4101}}{{TTS|MN|M.N.}} / early Marxism | |||
{{li|I=Z1/MN |Q=18,67|Q2=1867}}{{TTS|MN|M.N.}} / [[EC:9k/RD/Q1867|{{book|Capital}} volume 1]] | |||
</li></ol> | </li></ol> | ||
Latest revision as of 22:39, 6 September 2026
Main entry
- large language model
- large language model -> often seen as anti-humanities, although this contradicts with structuralist linguistics
- AI text generation / generating videos with large language models or advanced neural networks / AI-generated article (social phenomenon, Z0)
- AI image generation / generating images with large language models or advanced neural networks / AI-generated image (social phenomenon, Z0)
- AI video generation / generating videos with large language models or advanced neural networks / AI-generated video (social phenomenon, Z0)
Text generation
- reading arbitrary webpages and books into an LLM -> first of all, don't. second of all: the more interesting discussion here is what is being achieved or built up when somebody does this. if the machine isn't truly understanding it then what exactly did it use the data in the texts to create? my hypothesis is that it creates an ontology, while it is currently the case that humans can create ontologies better. the primary reason we haven't already built a good ontology is our obsession as human individuals towards Freedom and against filtration.
- companies buying subsidiaries to use their content as AI training data -> the biggest example is Elon Musk creating a new AI company to own Twitter. in this case, the whole process was controlled by one person. however, this process is interesting for the potential that separate billionaires could show up to buy companies and this could just become the new reason anybody ever buys companies any more.
- how many legs, again? / how many eyes does a human have? (generic; Deltarune) / (9k)
- jumping over paywalls with ChatGPT -> it's funny in such a dark way that this exists; if you understood what LLMs automate, you'd see it coming from far away. [1] I refuse to do it. you can read a LithoGraphica entry to get the same effect, or write one if you have access to the source, and more than one person gets to contribute to that. so, why does this exist? it exists because we've normalized an individual person with a lot of money buying an article being the only way to read articles. you know, a small shop putting out products assuming that everybody else has money to buy their products regularly when that might not at all be true. this incentivizes AI companies who are the only ones with money to send AIs to read everyone's articles, because even if they had to pay for the articles "legitimately" it would still be that they'd have the money and the readers wouldn't have the money. I am begging you if you have a Medium account with less than 100 followers to make your articles publicly available so an AI doesn't read them for people. ...this makes me realize. we should probably put in every single citation of a source whether somebody paid for it and a very vague idea of how much: new book, used book, paywalled article / paywalled or paid periodical. really, that should be in every single academic citation everywhere. we should push to get that into the official APA citation style guide to be frank. because that information is a fundamental part of publication, as much as the name of the publisher. field: graph economics.
- Nebula is a subscription streaming service... -> and eventually only AI companies will be able to pay for it, reading all the stuff in it with their machines and spitting it back out at the people who can only pay cents for content through ChatGPT. probably in the middle of a huge number of ads.
Language models and plagiarism
- plagiarism
- What is plagiarism? / Who owns the ability to repeat factual information? / Who owns the ability to repeat literary motifs? / Who owns the ability to independently repeat culture in another nation-state / Who owns the ability to independently repeat culture in another nation-state without paying another country that it happens also wants to overthrow your government and destroy national sovereignty? ->
everybody thinks they know what plagiarism is. anyone who owns a business or has a doctorate has absolutely no idea what it actually is or what it isn't; it's almost like the more educated you get the more confused you get about the question of plagiarism. here's the reality: the question of plagiarism is the question of what business territory owners will allow what other businesses or mere individuals to live and exist. that's precisely it. it's all up the whim of who likes who and who hates who. you're never guaranteed a license to exist even if you are willing to pay the money, it's all about personal relationships and court cases. there is no universal rule for what does infringe all copyrights or what doesn't infringe all copyrights. it's all about whether Bob wants Alice to be part of Bob's socially linked countable culture or wants to get rid of Alice. if Alice is Chinese, it's all about how much Bob wants China to exist or wants to wipe it off the face of the earth. - If a book is well-sourced, but written by AI, it is a reliable source / If a book looks entirely factual and traces all its statements back to a bibliography containing valid sources, but is written entirely by AI, it is a reliable source -> this is the huge problem with everyone's heuristics for what supposedly is and isn't a source. Wikipedia is generally going to be much more reliable than ChatGPT; the same machine beginning with the same data could output entirely factual statements, and total nonsense if that's the path you lead it down. but people have all been taught the heuristic that individuals are what make information reliable — if a book was written by an individual and it can't change, it's good, if it was written by multiple people and it can change, it's bad. this is a really bad rule given that instances of ChatGPT are individuals in the same sense that humans are individuals or house flies are individuals and they all produce static works that don't change. (yes, a house fly producing text would be hilarious, but that's almost exactly why LLMs are so widely criticized.) we've effectively taught everyone that ChatGPT is the easiest way to get high-quality information, while anything produced by humans is potentially unreliable. if you don't believe me... would you sooner believe a human-generated rant by me that I carefully checked against all my observations of material reality but didn't have time to add written sources to and maybe linked a video as a surface illustration of the idea, or a perfectly-formatted essay by ChatGPT that did end with a bunch of valid sources? yeah. think about that. we have a genuine problem.
this is one of the several different reasons I have for creating this Ontology project. I think there needs to be a way for people to check the validity of statements without resorting to specific human individuals or specific books, in the way absolutely all academia is done currently. it's good to take a verifiable concept and put a bunch of examples of reliable books on the entry that illustrate it or provide accounts of material evidence. but books aren't what actually makes things true or accurate. what makes things accurate is their sheer coherence with other testable understandings about material reality. that's why I am building this big bank of propositions. so you can take the most real and verifiable ones and use them to test the most dubious ones, whether you have books, whether you have experts, whether you can re-test observations of the material world positivist-style. this could one day be more reliable than Snopes because it wouldn't rely on special talented individual human experts versus just anybody who has a high school education and is sufficiently good at reasoning. assuming the claim you're testing isn't too new for all the required information to be recorded in here.
Text generation
- LLMs cause people to seek authenticity / LLMs will cause people to seek "authentic" work with real reasoning, real voices -> this is probably true, but it brings up a lot of slightly different questions. do people always seek rare things, or is there a point where they will truly be content with common things? do people really want the human content because of its quality or are they actually looking for its rarity or novelty?
Related
- Every citation should contain price information / Every citation should contain cost information -> I really do mean every citation in the world, not just every bop-format citation. field: graph economics.
- AI vision captchas hint at a new stage of capitalism / (9k)
- Products are chunks' footprints / Products archaeologically trace chunks / A commodity is the trace of an isolated chunk of people engaged in industry -> this does technically apply here because the content of a particular passage of generated text says something about the team of humans that produced the original text.
Ideologies or fields
- / structuralist linguistics
- / machine learning
- / large language models
- / early Marxism
- / Capital volume 1