Danielle ForéAt the end of the day, AI is a technology for people who like the idea of owning other people. That’s what makes it inherently misogynistic, inherently racist, and inherently fascist. It is an accelerated extraction and exploitation device
Shiv 🐍@danirabbit AI is just another tool like books/ internet/ search engines, by itself it is not bad, depends on the people who use it.
Marianne@hufflepuffWizard @danirabbit theft is bad actually
@noodlemaz @danirabbit all technologies we have invented to share knowledge are some kind of theft. Whole internet also was built based on ideas and knowledge stolen from somewhere else. We just need to de-centralise LLMs and make them open source. No single company or country should control it like the internet, it is a very important technology for humans now.
yoasif@hufflepuffWizard @noodlemaz @danirabbit Important technology is based on theft?
Why should people make things if they are just working for some decentralized borg stealing their works?
@yoasif @noodlemaz @danirabbit we should find ways to correctly compensate people on whose works LLMs are getting trained, but just giving up on AI and completely ban it just because we have not yet found ethical ways to build it is not the solution. AI can increase our productivity 100X and make us do things which were not possible earlier, ultimately helping everyone.
@noodlemaz @yoasif @danirabbit stats pulled from my own experience working in tech. Things which used to take 10 hours by scanning documentation and understanding how to do now takes only few minutes.
@hufflepuffWizard @noodlemaz @danirabbit So you compressed 10 hours of understanding into minutes and you think nothing was lost?
That is hilarious.
@yoasif @noodlemaz @danirabbit 10 hours of learning and memorising the syntax of a language which i don’t need to memorise in the first place. I would rather spend time solving actual problems rather than learning how to write regular expressions to match a string format.
@hufflepuffWizard @noodlemaz @danirabbit You haven't learned it, though. You substituted learning for a plausible result.
@yoasif @noodlemaz @danirabbit that’s the point, I don’t want to learn it, like I don’t want to learn how computers were programmed using punching cards 50 years back. I want to make human-machine interaction as natural as possible, you can’t stop this advancement to AI so we should find ways to make it better and ethical. We are not at that point yet but when wo do reach there, only ideas in the human minds will matter, not the execution.
@hufflepuffWizard @noodlemaz @danirabbit Not the ideas, the output, since you are compressing knowledge into the resulting outputs.
That is essentially a theft of labor, so I don't know that there is an ethical way to do it.
@hufflepuffWizard @noodlemaz @danirabbit Just think about it this way: you don't want to learn regular expressions, but you want the magic computer to do them for you. Great. Instead of someone building a NLP processor to convert your commands to an output, the LLM companies have stolen people's knowledge and actual work to generate outputs that resemble a plausible output.
@hufflepuffWizard @noodlemaz @danirabbit Why should other people learn and create those expressions for you for free so that you don't have to learn? Why do we presuppose that we need the outputs when we haven't figured out how to pay for the labor?
What entitles you and other LLM users to works that you yourself clearly don't want to learn how to create?
@yoasif @noodlemaz @danirabbit we can’t hard-code all the logic in this world into a computer, at some point computers would need to learn themselves like how humans do. Since computers need to learn it at a very large scale, from where will they learn it? Only humans and their work is there to teach them and like i said humans should be correctly compensated for their work on which AI is getting trained. This training will keep going forever, if humans keep creating
@hufflepuffWizard @noodlemaz @danirabbit But the computer isn't learning, it is compressing outputs.
You are advocating for thinking machines, but what we have are plagiarizing machines.
It would be one thing if you were telling us that the machine is reading the manual and learning how to write regular expressions. That isn't what is happening.
Humans are learning regexes, and then the computer copies the resulting regexes output from human knowledge.
@hufflepuffWizard @noodlemaz @danirabbit You are proposing that we simply steal those outputs instead of teaching the computer.
Even if you aren't, that is what is happening, and you are hand-waving the fact that the outputs are simply stolen -- "the outputs are good and we need to have them!" is what you are saying.
Why would people want to learn regular expressions and produce the outputs when YOU are unwilling to?
What gives YOU the right to their labor?
@yoasif @noodlemaz @danirabbit you are right, current AI systems are not learning anything, if I want to move to a system where we don’t need to use RegEx at all, current LLMs are not doing that. At most they can understand me and convert it into a regex internally based on how other humans have done till now and run it. Maybe we need to start building new AI agents which are actually able to think and understand in pure machine language instead of human language.
@yoasif @noodlemaz @danirabbit i just realised LLMs are the dream technology for communists 😂 if i know python and someone else knows java, now we both know both the languages. Nothing was stolen from anyone if they got something else in return ( both got to use one extra language albeit without understanding it) NOTE: this argument has many flaws, please don’t take it seriously
None Hitwonder@hufflepuffWizard @noodlemaz @yoasif @danirabbit
Exactly: you just described your own ass.
"It benefits me personally, so it's good for everyone."