If Reddit kills 3rd party apps it can absorb (or at least hope to do so) users of those apps and have complete control over how they access Reddit. Reddit can then feed them more ads, trackers and whatnot, all of which would translate into more revenue for Reddit, which is a net positive for shareholders.
There’s also the fact that companies training LLMs would be interested in paying those exorbitant fees to get training data as they likely can afford those fees.
So in short, Reddit likely wants to become a content farm for LLMs. As for the users, Reddit doesn’t care given their recent statements. So if some c*cks stay on Reddit, spez will just inundate them with more ads because why not, free money is free money, until everyone leaves.
Maybe, the problem with old reddit is that it’s as much about the comments as it is the content. Comments are hard to advertise on. Where card style scrolling content is great for advertising. Scroll a couple cards with a 20-60 second engagement, look at an ad, scroll a little more. . …
However, I don’t see this as being a great way to farm content for language models, the content engagement tends to drop significantly with the endless scrolling so my guess is that it’s a short term play to prove they can sell ads before the IPO.
Please correct me if I’m mistaken but isn’t the reddit dataset used to train LLMs from before Chat GPT became widely known? I was under the impression data from that point onwards was poisoned and not useful for training purposes
I can’t seem to find it now but I remember there being a ~90gb .zip megadb upload that got passed around a lot on machine learning reddit subs that was a snapshot of reddit before x date
Third party app users only made up ~3% of total reddit traffic. The revenue potential there is miniscule. It was never about the money, but about control.
I don’t trust reddit reporting that value accurately. Not saying it is a huge percentage, but they likely doing some things to minimize the numerator and maximize the denominator.
But yeah, definitely about control. They could have monetized users on third party apps with more reasonable API pricing. People would probably have barely noticed and that’d be new revenue for reddit that didn’t exist before.
If they charged 1/3 of what they are for third party clients to access the API and required Reddit Premium for API access…. They could have made a mint!
Plus they could have reported huge upticks in subscribers to their service. THAT would have looked good to Wall Street for an IPO.
Yeah, that would have been smart. But Huffman has publicly (and unwisely) admitted that he admires Elon, which means he is currently not at home to Mr. Smart.
If Reddit kills 3rd party apps it can absorb (or at least hope to do so) users of those apps and have complete control over how they access Reddit. Reddit can then feed them more ads, trackers and whatnot, all of which would translate into more revenue for Reddit, which is a net positive for shareholders.
There’s also the fact that companies training LLMs would be interested in paying those exorbitant fees to get training data as they likely can afford those fees.
So in short, Reddit likely wants to become a content farm for LLMs. As for the users, Reddit doesn’t care given their recent statements. So if some c*cks stay on Reddit, spez will just inundate them with more ads because why not, free money is free money, until everyone leaves.
Maybe, the problem with old reddit is that it’s as much about the comments as it is the content. Comments are hard to advertise on. Where card style scrolling content is great for advertising. Scroll a couple cards with a 20-60 second engagement, look at an ad, scroll a little more. . …
However, I don’t see this as being a great way to farm content for language models, the content engagement tends to drop significantly with the endless scrolling so my guess is that it’s a short term play to prove they can sell ads before the IPO.
Please correct me if I’m mistaken but isn’t the reddit dataset used to train LLMs from before Chat GPT became widely known? I was under the impression data from that point onwards was poisoned and not useful for training purposes
I can’t seem to find it now but I remember there being a ~90gb .zip megadb upload that got passed around a lot on machine learning reddit subs that was a snapshot of reddit before x date
Third party app users only made up ~3% of total reddit traffic. The revenue potential there is miniscule. It was never about the money, but about control.
I don’t trust reddit reporting that value accurately. Not saying it is a huge percentage, but they likely doing some things to minimize the numerator and maximize the denominator.
But yeah, definitely about control. They could have monetized users on third party apps with more reasonable API pricing. People would probably have barely noticed and that’d be new revenue for reddit that didn’t exist before.
If they charged 1/3 of what they are for third party clients to access the API and required Reddit Premium for API access…. They could have made a mint! Plus they could have reported huge upticks in subscribers to their service. THAT would have looked good to Wall Street for an IPO.
Instead we get this fuster cluck.
Yeah, that would have been smart. But Huffman has publicly (and unwisely) admitted that he admires Elon, which means he is currently not at home to Mr. Smart.