One big problem is that Google also lies about dates when you do a 'before' search, so AI spam still shows up even if you filter it out.
Scraping a new dataset without contaminated content is impossible now. There's no authoritative source on the age of any given website. (You can try places like the Internet Archive, but most such archivers aren't interested in handing their data over to AI firms because of the bad PR.)
The only way to ensure you don't have AI content is to use a dataset created before ~2021-2022. This too is hard because all the big ones are illegal to posses as they contain CSAM.
The real answer is that AI developers played themselves. Humanity was creating exponentially more data each year, and by rushing these AI tools out the door, all that data is useless now. By the end of this year there'll be more new "AI-polluted" data on the web made after 2021 than the total amount of "clean" data made up to 2021.
By the time the actual content comes back up, it won’t matter.
It's the same here, we have many trusted sources that we don't believe would lie about when content is from - Internet Archive, newspaper archives, widely used web crawls, Wikipedia exports, etc. Obviously you don't use an AI (or a black box search engine) to get non-AI content, you get it from people and systems you trust.
We've been yelling about Web of Trust for decades by now, and never built it and now it's too late.
I think you're being realistic about the underlying adversarial dynamics. I.e., it's an arms race regarding recognizing AI authorship.
>*Google Books is indexing low quality, AI-generated books that will turn up in search results*
This was never about articles, it is about books, though. You can't retroactively publish a book in 2021 if we're not in 2021.
It's bewildering to see how people here are so confident in their tone while arguing about things they think are said in the linked post, but don't know for sure because they never even opened it to read it.
You make a valid point though, which is actually more scary now that I think about it... it is becoming less and less possible for non-google (/other huge) companies to train models on verifiably human-generated content, as only Google (/other huge companies) have enough accurately timestamped data to train on. Hmm.
It's not like this is exactly a new phenomenon - Plato's famous "Allegory of the Cave" was a thought experiment in how the media we consume can create what we perceive as reality - it's just that we're discussing an uncomfortably recent intersection of philosophy with computer science. I said it a few times in the threads that blew up over Google's diverse Nazi bungle; I didn't get into computers to discuss anthropology and how humanity constructs meaning; I actually did it to avoid that and do work I considered purely practical after washing out of a liberal arts degree. That said, the whole thing ironically felt like very good art in the sense that it got people to discuss the nature of these generative AI systems. For me especially, the image of the Black Pope was very thought provoking as a (lapsed) Catholic; why is it that there has never been a black pope?
There were a few popes of African descent, though; don't know if their skin color happened to be black
Still, I don't get why it would matter to any Catholic (I'm not Catholic), except to (some) Catholics in the US of A maybe, since race seems to reign supreme over there, even more than God Himself to (some, supposed) believers
I can get into my personal feelings if it matters; I don't think they matter much to anyone except me. The quick summary is that I dug the grave of a great-great Aunt of mine by hand at a small cemetery attached to the Catholic church in the tiny, forgettable town she was born in. One quirk of this rural, very white (I assume the entire population is a fairly close cousin of one type or another to me), Catholic town is that the priest is an African immigrant. My aunt was quite racist and so I thought it was a sort of ironic that the last group of people who saw her out of this world were twenty-five percent black.
So an AI-generated image of a black pope made me stop and think for a minute about what Catholic identity means; a concept that as someone who was raised in that tradition but is unable to myself believe in the divine has always been problematic. Basically, it was good art, though I don't think that was the intent of the corporation that created it.