Large language models are fed on what data can be found crawling the Internet. The more data you feed ChatGPT, Gemini, or Claude, the higher the quality of their outputs. But that can work in rever…
And some AI companies like Perplexity just shout “YOLO” and try to stealth-crawl it anyway. [404 Media; Wired]
it’s fucking wild that automated scraping without permission used to be something you did as a last resort under strict restrictions and secrecy, cause whoever had the data you needed wasn’t exposing a usable API. but not in the AI industry, there it’s the fucking foundation of the entire company
Yeah. There probably was a fair bit of stealth-crawling up to this point, but the perps knew they needed to keep it on the down-low.
The AI bubble, on the other hand, lacks the ability to keep it subtle, making it plainly obvious people’s shit was getting stolen and showcasing AI bros/techbros’ utter disregard for anyone but themselves (e.g. by ignoring robots.txt).
(Off-the-cuff prediction: anti-AI scraping measures will likely start feeding false info to AI scrapers they detect - beyond simply throwing a wrench into those models, it’d also make it less likely AI scrapers will realise “hey, our shit’s getting blocked”)
Which is terrible since scraping is an extremely important tool for normal people. Like, if YouTube gets good at blocking scraping there will be literally no way to watch their videos anymore.
it’s fucking wild that automated scraping without permission used to be something you did as a last resort under strict restrictions and secrecy, cause whoever had the data you needed wasn’t exposing a usable API. but not in the AI industry, there it’s the fucking foundation of the entire company
Uh, it has been done on a massive scale for years…just not regurgitated so readily until now
Yeah. There probably was a fair bit of stealth-crawling up to this point, but the perps knew they needed to keep it on the down-low.
The AI bubble, on the other hand, lacks the ability to keep it subtle, making it plainly obvious people’s shit was getting stolen and showcasing AI bros/techbros’ utter disregard for anyone but themselves (e.g. by ignoring robots.txt).
Personally, I expect this will lead to much stronger scraping protections being developed to combat shit like this - Cloudflare’s already offering to block AI scrapers for its users and Kudurru’s offering a similar service, I can easily see a new market opening up here.
(Off-the-cuff prediction: anti-AI scraping measures will likely start feeding false info to AI scrapers they detect - beyond simply throwing a wrench into those models, it’d also make it less likely AI scrapers will realise “hey, our shit’s getting blocked”)
Which is terrible since scraping is an extremely important tool for normal people. Like, if YouTube gets good at blocking scraping there will be literally no way to watch their videos anymore.