3 ms·
How is this policy enforceable? There aren’t any reliable ways to detect LLM-generated text, are there?
by felideon 2y ago
How is this policy enforceable? There aren’t any reliable ways to detect LLM-generated text, are there?
- rany_ 2y agoUsing common sense, probably. You don't need an LLM detector to tell that something is AI generated.
- paxys 2y agoYou don't need an LLM detector to tell that something is visibly AI generated. With some minor prompting and editing it can be indistinguishable from any other writing.
- TuringNYC 2y ago>> Using common sense, probably. You don't need an LLM detector to tell that something is AI generated. This could quickly turn into a witch hunt
- barfbagginus 2y agoMany people now believe things are AI generated the moment the writing says something they disagree with. That's fun!
- sebastiennight 2y agoThere's money to make with a Chrome extension that flags comments contrary to your beliefs and tags them as "likely AI-generated misinformation" while highlighting the "report" button.
- barfbagginus 2y agoMaybe there is money to be made. crypto lock the PCs of anyone petty enough to use such an extension. They deserve it. And boom, revenue. Disclaimer: I'm joking - don't do this. It could land you in the big house.
- Verdex 2y agoIt looks like you're getting downvoted, but my opinion is that your statement is correct. Yes, technically, common sense doesn't work because some things written by AI won't be obviously written by AI and some things written by people can appear to be written by AI when it was not. However, I suspect that medium doesn't actually care if something is written by AI or not. They want to stop the person writing hundreds of garbage articles a day. The person writing a nonstop deluge of poor quality articles sans AI is going to be erroneously banned. The person writing high quality articles 100% powered by AI won't be banned. AI will simply be the justification for culling what they do not want.
- matwood 2y agoExactly. People are getting too caught up on AI or not. If an article is good, it's good, regardless of AI.
- overstay8930 2y agothis works until an entire generation of people are used to reading LLM generated text and they write like an AI
- speedylight 2y agoThat strategy works for lazy publishers; with a bit editorializing and prompt refinement you can make a AI generated text indistinguishable from written text. Believe me it’s easier than it has any business being.
- EA-3167 2y agoRead a TOS some time, it's enforceable because they can choose to do absolutely whatever the hell they want. If they're wrong, well... tough nuts.
- SkyPuncher 2y agoBecause the actual policy is “we can remove anything”. Just the positioning is they’ll only use it on AI.
- iainctduncan 2y agoThey have years of data on how fast a regular account can generate content. If you are producing 100 times more content.... or content 100 times faster....
- dandesim 2y agoThe policy is enforceable through moderation. They're probably looking for multiple indicators then doing a manual review. The only way to monetize Medium is through the partner program, which is reliant on high volume of posts or a post to go viral. It's unlikely an AI generated post will go viral, but it is very likely people will spam the system and start posting 3-5x a day to get a few dollars. I don't think they're as concerned with someone using AI as an assistant to write a few posts a month, but rather AI slowly draining their payouts from real people.