If they just worked with the users I'm pretty sure a lot of us would be absolutely proud of reddit for making it big and raking in the dough. All this time they'd left open old.reddit.com as a nod to us who've been here through all these years, and the new users were content with the redesign that was designed to get reddit paid.
Could you imagine how it would go down, if reddit had quietly negotiated with developers to allow for user API access and limiting the rates to that which a normal user would consume? (e.g. 400 calls a day or whatever). This would both allow for 3p apps to continue hooking us up with the oldschool forum board we like, WHILE preventing LLM models from scraping 1 Gigabyte of text every day.
Literally everyone would have been happy. Users would have the slight inconvenience of maybe having to re-log into their app, LLM and Researchers would pay for the privilege of getting to scrape 1 Gigabyte of text a day.
The only loophole here is malicious apps scraping the user's data as they browse and selling that black market. But is that really such a big hurdle as to completely discard the approach?
When I used to live in a small room, one day I put my guitar kind of half-way into my closet by the corner of the room. A few mins later I yelled something (I forget what) and it was as though my guitar strummed the A string on its own. It freaked me out for a sec but then I started belting different notes, and as soon as I got back to a certain note the guitar would strum itself again!
Though that might just be the case for most things, I don't think it would be fair to say that for all things. Two traits could develop in conjunction with another, and trait 1 gives an advantage, while trait 2 is a disadvantage. The most in-your-face example I think is the stereotype of nerds and glasses. In short, a good enough trait could get you laid and fed even though you have a crappy trait that came with it.
edit: alternatively, a bad enough trait could completely derail the development of a good train. Maybe dolphins born with 200iq brains develop some other kind of bad trait as a result of the same mutation that doesn't allow for the 200iq trait to permeate.
Could you imagine how it would go down, if reddit had quietly negotiated with developers to allow for user API access and limiting the rates to that which a normal user would consume? (e.g. 400 calls a day or whatever). This would both allow for 3p apps to continue hooking us up with the oldschool forum board we like, WHILE preventing LLM models from scraping 1 Gigabyte of text every day.
Literally everyone would have been happy. Users would have the slight inconvenience of maybe having to re-log into their app, LLM and Researchers would pay for the privilege of getting to scrape 1 Gigabyte of text a day.
The only loophole here is malicious apps scraping the user's data as they browse and selling that black market. But is that really such a big hurdle as to completely discard the approach?