Nice way to deflect, are you going to address the elephant in the room or just pretend you didn't read my last post?
Since you didn't want to engage, I usually don't push it.
Since you are now aware that gen-ai is trained via exploitative child labor, i don't think that you are interested in talking about what I am bringing to the discussion.
Ignorance is bliss, right?
However, since you're pushing for it, I will do what I generally don't do, which is underlining the logical fallacy of your post.
First, you present some elements pointing to the existence of countries with much lower social standards. Which is known to absolutely everybody on earth with a modicum of common sense. So yes, the task of captioning will be outsourced to low-wage countries. This is called the mechanical turk, in reference with the scam of the 18th century "mechanical" chess player, and was already mentionned in the thread (at least by me) as the logical alternative to training on a massive amount of data with poor captioning. Experts think that it is an effective way to train model: less data, better caption, so of course the captioning will go to countries where the minimum wage for unskilled labor is 10-dollars a day rather than Luxembourg (250-300 dollars a day country, roughly, as a minimum wage). This is especially true for prisoner's labor, which are often indemnified less for their work, irrespective of country (for example, in the US federal prison system, according to wikipedia, the pay range is ghastly less than a single dollar an hour, so it's in the lower range of the horrible examples of wages given in your blurb). If anything, it was the expected outcome from the copyright holders' backlash against scraping.
Then, you try to mix low wage countries with child labor, presenting reports that several AI-related economical actors (namely, the US Department of Defense, Meta, Microsoft, OpenAI, Google, Amazon) are relying on child labor. Connecting the two is the first logical error, as if outsourcing low-margin services was the same as using child labor.
The second logical fallacy is that you conclude your post by "The global south, impoverished children, displaced refugees in camps, and prisoners are all included in the labor pool that has been exploited to create your shiny gen-ai tools. How you deal with that is your choice, but personally, I wouldn't touch gen-ai tools with a ten foot pole." The argument here is that your saying "An handful of AI actors are culprit of child labor, therefore all AI actors are guilty of using child labor", therefore you shouldn't use generative AI. Or, more exactly, "there I don't use generative AI, and you can do whetever you want" which is implying that doing so would condone child labor.
This is a fallacy because clothing is a sector widely known to be using child-labor, yet nobody in their sane mind is running around stark naked, saying "I wouldn't touch textile technology with a ten foot pole". They would be arrested quickly, especially if going commando around schools. And why does nobody do that? Why aren't jails full of naked people? Are all the clothing adepts supporters of using children as modern S-word? Has humanity stooped so low? No, they don't walk around in the nude because because everyone see that your argument is a fallacy and people usually try to avoid the brands that rely on child labor, at most, and not the whole technology.
"Some AI actors rely on child labor, therefore all AI actors must rely on AI child labor" is a hasty generalization. "Some X are Y, therefore all X are Y" is blatantly unsound logic. Try to explain that with X = Black People and Y = any negative quality and see if your reasoning stands, it should be apparent to all that it is not. Or rather, "the Caravaggio and Cellini were murderers, Picasso committed theft, Egon Schiele exploited child girls... therefore artists are criminals, all of them." You must see that it doesn't work, mustn't you? The logical error is blatant. So blatant that I find difficult to imagine that you're doing this in good faith and not trying to do that deliberately to insinuate that all AI users condone child labor (much like others point at "art theft" in the training of a model to criticize all generative AI, even those clean of "art theft"), which would mark the end of our discussion. But if you're really eschewing clothing out of contempt for the child-exploiting clothing industry, than I apologize for thinking you're using rhetorics here. But in that case, I'd strongly advise against applying the same logic to the Domino pizza case @trappedslider mentionned above and sever your link with the eating technology. Please do continue eating nonetheless and focus your legitimate wrath on a more narrow target, like... the specific stores that were found guilty. Especially those that served ananas-coverd pizza, because that's the true evil.
So, once the fallacies have been addressed, what's left? The accusations that some entities are encouraging exploitation of children and other vulnerable people, and the question about my personnal stance on them.
If you're really interested in my purchasing behaviour, which I doubt anybody is, I have no contractual relationship with the US Department of Defense, Meta, Microsoft, Amazon, Boeing or Google. I prefer Qwant but I will use Google if set up as a default search engine on a public computer. I have an OpenAI subscription as part of my job, where I don't really have any say on it and I have an indirect relationship with Boeing (it might happen that my flights to Asia are on Boeing aircrafts as I don't go out of my way to select Airbus A350s over Boeing 787s despite the added comfort and reduced noise). So basically, I am already not patronizing those entities, which is probably rare, because very few people are able to boycott Microsoft, Google and, especially, the US government, and I can't really do anything more than not giving any of them a single dime. I don't claim any moral ground here, I didn't know they were encouraging child labor before reading from you.
Last edited: