I know a lot of FOSS people are hostile to AI in general, and this is an immediate problem, but I feel like a better solution for everyone would be for there to be some sort of central repo of this information that AI companies can pull from without externalizing their costs like this.
FOSS infrastructure is under attack by AI companies
131–140 of 631 posts
Re: FOSS infrastructure is under attack by AI companies
#132Earlier quoted context omitted.
> Copyright the content and sue those who use it for AI training. This costs money, time, and ongoing commitment. FOSS isn't typically known for being overflowing with cash.
Good thing law firms are overflowing with the desire to make large amounts of cash from contingency based cases.
It's unclear if you'd win this case right now.
Re: FOSS infrastructure is under attack by AI companies
#133We need a project in the spirit of Spamhaus to actively maintain a list of perpetrating IPs. If they're cycling through IPs and IP blocks I don't know how sustainable a CAPTCHA-like solution is.
Why only in the "spirit of Spamhaus"? Spamhaus still exists. Add Google and Microsoft AS to the DROP/NOROUTE list, that would be hilarious.
Re: FOSS infrastructure is under attack by AI companies
#134I am self-hosted (email + web), I did quit the DNS (which registrars are now mostly hostile to noscript/basic (x)html browsers anyway), and I thought it would give me some relief...
Nope.
You don't have only the AI crawlers, you have also scans and hack attempts (which look alike script-kiddy stuff), all the time. Some smell of AI strapped to javascript web engines (or click farms with real humans???).
Smart: IP ranges from all over the world, and "clouds" make that even worse since the pown systems or bad actors (the guys who scan the whole ipv4 internet for its own good AND MANY SELL THE F* SCAN DATA: onyphe, stretchoid.com, etc) are "moving", in other words clouds are protecting those guys and are weaponizing hackers with their massive network resources, wrecking small hosting. No cloud is spared: aws, microsoft, google, ovh, ucloud.cn, etc.
I send good vibes to the brave open source software small hosting (until they are noscript/basic (x)html compatible ofc).
Many fixed-IPv4 pown systems have been referenced by security communities, often for months, sometimes years, and the people with the right leverage, don't seem to do a damn thing about it.
Currently, I wonder if I should not block all digital ocean IP ranges... and I was about to do the same with ucloud.cn IP ranges.
The second you host anything on the net, it WILL take a significant amount of your time. Do presume you will be pown, that's why security communities are referencing each other too.
Then I am thinking going towards 2 types of "hosting": private IPv6+port("randomized" for each client, may be transient in time depending of the service) thanks to those /64 prefixes (maybe /92 prefixes are a thing, for mobile internet?). Yes this complicated and convoluted. Second type, a 'standard' permanent IP, but with services which are implemented in an _HARDCORE_ simple way, if possible near 100% static. I am thinking going even further: assembly on bare metal, custom kernel based on hand compilation of linux code (RISC-V hardware ofc, FPGA for bigger hosting?)
I don't think anything will improve unless carrier scale network operators start to show their teeth.
Re: FOSS infrastructure is under attack by AI companies
#135Copyright the content and sue those who use it for AI training. I believe there is a lot of low hanging fruit for lawyers here. I would be surprised if they weren't preparing to hit Open AI and alikes. Very badly. Google get away with deep linking issues as publishers after all had some interest in being linked from the search engine, here publishers see zero value.
> Copyright the content and sue those who use it for AI training. This costs money, time, and ongoing commitment. FOSS isn't typically known for being overflowing with cash.
The FOSS community can transfer their copyrights to an organisation in order to take legal action against AI operators.
Re: FOSS infrastructure is under attack by AI companies
#136Insane, I wonder if we eventually end up with a non-search-engine indexed version of the web that's more like browsing in the 90s where websites just had to link to oneanother to get noticed . . . .
I love that the solution to LLM scraping is to serve the browser a proof of work, before they allow access - I wonder if things like new sites start to do this . . . It would mean they won't be indexed by search engines, but it would help to protect the IP
Re: FOSS infrastructure is under attack by AI companies
#137At this rate, it's more than FOSS infrastructure -- although that's a canary in the coalmine I especially sympathize with -- it's anonymous Internet access altogether. Because you can put your site behind an auth wall, but these new bots can solve the captchas and imitate real users like never before. Particularly if they're hitting you from residential IPs and with fake user agents like the ones in the article -- or…
We're half way there already. It always hits me whenever I am doing some mapping for OpenStreetMap and I'm looking up local businesses without their own internet presence. They use Facebook, Instagram, X, etc. for their digital calling card. I normally don't use Facebook (or Instagram, and gave up on X) and have no account there, and every time I follow one of those links, you get some info, and then you get a dialogue screen telling you to make an account or get lost, or you just get some obscure error.
I don't mind registering an account for private communities, but for stuff which people put up thinking it is just going to be publicly visible it's really annoying.
Re: FOSS infrastructure is under attack by AI companies
#138[flagged]
Re: FOSS infrastructure is under attack by AI companies
#139Earlier quoted context omitted.
And Usenet, and IRC with a registered user prereq to join. Also, set AI tarpits as fake links with recursive calls. Make then mad with non-curated bullshit made from Markov chain generators until their cache begins to rot forever.
Usenet, as far as I remember, used to be a fucking hell to maintain right. With each server having to basically mirror everything, it was a hog on bandwidth and storage, and most server software at its heyday was a hog on filesystems of its day (you had to make sure you have plenty of inodes to spare). The other day, I logged into Usenet using eternalseptember, and found out that it consisted in 95% of zombies sendin…
Usenet wasn't that bad if you didn't take the binary groups.
> and found out that it consisted in 95% of zombies sending spam you could recognize from the millenium start
I like to imagine a forgotten server, running since the mid-90s, its owners long since imprisoned for tax fraud, still pumping out its daily quota of penis enlagement spam.
Re: FOSS infrastructure is under attack by AI companies
#140Does anyone maintain a list of ai products developed without scraping?
I don't think it's possible to develop a frontier model without mass scraping. The economics simply don't add up. You need at least 10 trillion tokens to make an 8 billion parameter model. 10 trillion tokens is something like 40 terabytes. You simply can't get 40 terabytes of text without mass scraping.
> You need at least 10 trillion tokens to make an 8 billion parameter model
Are you sure it is not just very inefficient?