- Lucretius in "De rerum natura", probably
Not necessarily related but this would’ve been a big distraction:
https://blog.catbox.moe/post/813932072453455872/happy-11th-b...
Expires in 2d: https://streamable.com/dkyvu8
Edit: bad actors spamming Catbox
https://blog.catbox.moe/post/813932072453455872/happy-11th-b...
@Anthropic get your Claude under control:
“Catbox has been running for 11 years now, and for 9 of those years, growth was pretty linear. Traffic goes up, storage used goes up, support goes up. This is the “organic” nature of Catbox. For the last 2 years, amplifying in the last 6 months, both storage used has gone up significantly compared to traffic and support. I first investigated this last year around May, as it was starting to put pressure on the storage space available to Catbox. I was able to find that most of the storage used was from a handful (35 or so) IP addresses that were uploading over 500 GB of content to Catbox in very short spans of time anonymously. After purging those uploads and banning those IP addresses, things seemed to be fine, however later last year, around September, disk consumption began to increase exponentially again compared to traffic. Doing what I could to mitigate it at that time involved Project Lain, as well as light monitoring of high usage IP addresses, like before. However this time there was no “super users” that were eating up storage. I let it be for a bit while purging a couple that I could. This problem increased even more in the last 60 days, to the point where I was burning through around 200-300 GB per day. Review of upload data shows hundreds of datacenter and proxy service IP addresses uploading 10-20 GB each for a few days, then dropping off. Looking at the files, it’s various “slop” content, including:
- Low resolution AI generated porn
- Tiktok Videos from the Middle East and SEA
- Clearly scraped LinkedIn/publicly available photos
- blob files containing junk data
Clearly this is not the “organic” traffic I mentioned earlier, and since the IP addresses are so varied, it’s clear something is happening here. I was alerted by someone that Claude will use Catbox in its coding projects as a “dumping ground” of sorts for when it needs to redirect content. This is clearly an abuse of the service, and stops today. …”If there's a hint of sepia in the original image and the training data contains a lot of sepia images, it will certainly get reinforced in this process. And the original distracted boyfriend meme certainly has some strong sepia tones in the background. Same way that Dwayne Johnson's face looks a tad cartoonish. And in the intermediate steps they both flow towards some averaged human representation that seems pretty accurate if you consider the real world's ethnic distribution.