Definitely not just Medium. It's everywhere. I have noticed that someone immediately calls out pretty much every bit of content as AI generated, whether it obviously is or not. I guess this is a best case scenario. I do wonder what will spring up next to allow high quality content to be found.
AI Slop Is Flooding Medium
71–76 of 76 posts
Re: AI Slop Is Flooding Medium
#72Definitely not just Medium. It's everywhere. I have noticed that someone immediately calls out pretty much every bit of content as AI generated, whether it obviously is or not. I guess this is a best case scenario. I do wonder what will spring up next to allow high quality content to be found.
Human curation will become more and more valuable, perhaps precisely because it does not scale. You know which travel advice I prefer? The one from a small, local blogger that loves their city. You know where I find my music? At house parties. More and more, I also crawl the profiles of individual users for content that I might like. The cosy web is my refuge from scale and its consequences.
And that lack of scale means it has little mass monetization potential. It's actively hostile to weirdos just trying to make a buck. I don't get randos from who-knows-where trying to hawk their wares in my private group chats, so those recommendations actually carry weight.
Re: AI Slop Is Flooding Medium
#73Re: AI Slop Is Flooding Medium
#74> CEO Tony Stubblebine says it “doesn’t matter” as long as nobody reads it. It seems like the leaders of all of these online platforms are going to just bury their heads in the sand and pretend that this massive problem isn't one. This AI trend is absolutely going to destroy many (most?) of these companies. For better or worse, the new age of the internet is already here.
is there anything they can realistically do?
Re: AI Slop Is Flooding Medium
#75Earlier quoted context omitted.
What gives you confidence in this AI detection model?
We benchmark on pre-2023 datasets of O(10M) documents not in our training set. Other detectors seem to have between 1-3% false positive rate and ours is around 1 in 10,000 as of our latest model update. We do a lot of active learning + core set selection to keep FPR low and improve recall on larger LLMs. Our white paper with some methodology is here: https://arxiv.org/abs/2402.14873
I noticed something, you achieved near 100% accuracy on most domains in every domain but scientific which made me wonder, how much is that could be due to how "strict" and "profrssional" these papers could be, or maybe how a slightly disproportionate number of the training data for these LLM could be from science based articles and papers, as they are generally viewed to be "high quality"
Interesting read either way, best of luck on your project (:
Re: AI Slop Is Flooding Medium
#76> The 12-year-old publishing platform
basically