An update to Google’s privacy policy suggests that the entire public internet is fair game for it’s AI projects. If Google can read your words, assume they belong to the company now, and expect that they’re nesting somewhere in the bowels of a chatbot.
Yet they attempt to prevent any web-scraping of their services. Interesting.
All for me and none for you
The classic motto of a company that’s simply too big and is asking to be broken up.
I expect every AI company is doing that. At least Google is being honest about it.
That’s the ugly truth. There’s plenty of reasons to be upset with Google, but this ain’t it.
𝕱𝖎𝖓𝖊! 𝓛𝓮𝓽’𝓼 🅂🄴🄴 𝕥𝕙𝕖𝕞 ѕ¢яαρ 𝔱𝔥𝔦𝔰!
Oh, hey cool - look guys, more recaptcha fodder!!
This has been discussed elsewhere, and by people smarter than I, but chat bots are going to start learning from other chat bots and it’s going to be less and less reliable over time, no?
Like there is an internet BEFORE ChatGPT, which is about as reliable of data as one could hope to find, and then there is a post day one chatgpt, which the data is already getting polluted by random LLM gibberish. How is google’s webscraping going to know if the data it is getting is legitimate human being thoughts, or just random madeup shit from a LLM?
There was an article recently about this (too lazy to search it). It’s already starting to happen. If most of the content they train on is the internet and more internet content is created by LLMs without being tagged as AI generated content (can’t be guaranteed by all actors), then it’s inevitable. High signal training data is out the window.
likely they would limit training data to only include pre-2020 or earlier to avoid this
Then you run into the problem of having outdated information. As more AI generated content pollutes the internet and more time passes the problem will only become more severe.
There are experiments with feeding LLMs output of other LLMs and the results are awful. Seems for now they can only generate sensible text if fed human output.
Right, but if they are training all new AI on shit they find online, like this comment, wouldn’t that pollute that dataset, considering I generated this comment with AI?
I can’t tell if this is a joke or not lmao
It was… perhaps?
considering I generated this comment with AI?
Woah
Your comment seems too intelligent to be AI generated.
Guilty.
Not sure how they can enforce their terms and conditions on me when I don’t use any of their services?
It’s not a TOC, you don’t have to agree to it. They’re just kind of telling you what they feel like they can get away with. I don’t see anywhere in the new terms where they outright assert that they own it though, but they just kinda say “Yo, if we can see it, we’re going to use it to train AI”.
That’s what Google does with search and advertising, isn’t it? They take everything they can see and use it to make money.
They don’t need to enforce anything, you do. If you post anything onlline, they’ll scrape it. If you have a problem with that, the onus is on you to enforce your copyright (if you have it)
But when the world needed them most the robots(.txt) were no where to be found
It’s kind of you to assume that they respect those, lol.
Oh and they won’t. We had a lengthy verification process where we proved we owned these sites we didn’t want aggregated and they still refused to honor robots.txt.
And yet somehow, they still have the shittiest AI this side of ChatGPT and Claude. (Source: tested PaLM 2 on Poe, and Bard). Checking out the fine tunings of Llama, it doesn’t seem like more training is always better. You’ll hit a wall. And even tuning seems to be an art at the moment.
Dang, remember when Google’s motto used to be “Don’t Be Evil”?
They stopped to not being evil, that is…
There has always been a symbiotic relationship between search engines and content site owners. The deal being “l (site owner) will let you index my site to make it easy for people to find my content. In exchange, you (Google) can make money by building user profiles and selling targeted advertising.” Conceptually this is no different, except that Google is now using the data to build new applications and businesses - AI rather than ads.
I believe that Google does respect robots.txt (though these need to be well specified and located), so it’s relatively easy for site owners to opt out of being indexed. Whether being indexed should be on an opt-out basis (as opposed to opt-in basis) in the first place is perhaps the key question, one I’d argue should have been discussed 20, 30 years ago.
or they could maliciously opt-in by including lots of hidden garbage text to poison the scrapers
Cool. Cool, cool, cool
I’m applying the same principle to every corporately owned IP out there, too, then. If I can see it, it’s mine.
What a surprise
I don’t think they belong to the company any more than the words you read belong to you
Exactly, this reads like hysteria. If you’ve placed your words on a public website, it’s a shocked Pikachu moment when someone (or in the case of an AI-in-training something) reads those words. It’s basic fair use.
If someone put up a billboard with some text on it and then got angry whenever someone else read it I would question their sanity. Even if that “someone” was the Google street view car.
Yeah, I don’t really see the fuss about people’s content being used to train AIs. It’s not really any different from a human reading your content and using their brain to make something similar.
There’s a surprising number of people who seem to think LLMs contain a database of everything it’s trained with, and that it just spits out snippets from there. There are also lots of very vocal artists against image generation models who claim that these 5-10 GB models contain all their copyrighted art, claiming that the models just create collages from existing images.
People simply don’t understand how these things work.
good, I’d like it if AI thinks a bit more like me and a bit less like the rest of the internet
Does this include emails as well? If so, I guess I’ll have to migrate over to another email service.
Gmail has scanned your emails, and those sent to you, since the day the service went live. It is part of the Ts and Cs.