4 ms·

Reddit is trying to flip loss to profit. They have a wealth of data for AI models and AI is going to be big. They are trying to cash in on it. What they should

Reddit is trying to flip loss to profit. They have a wealth of data for AI models and AI is going to be big. They are trying to cash in on it. What they should have done was make a free model for general API use like most companies to develop against. And after that tier the model accordingly. Data is expensive, even though it is getting cheaper per byte we are producing more of it per volume.

zarzavat 3y ago
I don’t get the AI angle. The data already exists in dumps. How does restricting access to the API stop that? Sure, you don’t get access to new data but I doubt anybody cares about that given that there’s 15 years of existing data already. Especially if they kill all the interesting subs and turn Reddit into cat picture central.
- Sparkyte 3y ago
  True, but old data is not what companies want since it is readily available and probably already in models. Reddit produces more data than any other social media website in the text and picture format. I think Tiktok is relatively close but with videos.
  - ryanbrunner 3y ago
    Reddit is still also open to indexing, so the data is still currently available with no authentication required to anyone interested in crawling it. The training data for AI theory doesn't hold water.
    - Sparkyte 3y ago
      The other thing to also consider is that Reddit funding is partially attached to investments. Investments for Reddit and many other investment based companies dried up over the government pulling back interest rates to deal with inflation.
dyno12345 3y ago
reddit has been around for nearly 20 years, the AI thing just seems like a hype train they're trying to jump on to.
- Sparkyte 3y ago
  Hypetrain like crypto. But AI has potential unlike crypto. Not part of the hypetrain as you really need to get it behave properly. The biggest loss in AI right now is the data needed to for an AI model to work. Not only does it need source data but it needs to store its model, those models can be MASSIVE. Which means expensive, to make it accessible you need to store it in memory... to make it better it needs to be faster... faster larger memory is $$$.

zarzavat 3y ago

I don’t get the AI angle. The data already exists in dumps. How does restricting access to the API stop that? Sure, you don’t get access to new data but I doubt anybody cares about that given that there’s 15 years of existing data already. Especially if they kill all the interesting subs and turn Reddit into cat picture central.

Sparkyte 3y ago
True, but old data is not what companies want since it is readily available and probably already in models. Reddit produces more data than any other social media website in the text and picture format. I think Tiktok is relatively close but with videos.
- ryanbrunner 3y ago
  Reddit is still also open to indexing, so the data is still currently available with no authentication required to anyone interested in crawling it. The training data for AI theory doesn't hold water.
  - Sparkyte 3y ago
    The other thing to also consider is that Reddit funding is partially attached to investments. Investments for Reddit and many other investment based companies dried up over the government pulling back interest rates to deal with inflation.

dyno12345 3y ago

reddit has been around for nearly 20 years, the AI thing just seems like a hype train they're trying to jump on to.

Sparkyte 3y ago
Hypetrain like crypto. But AI has potential unlike crypto. Not part of the hypetrain as you really need to get it behave properly. The biggest loss in AI right now is the data needed to for an AI model to work. Not only does it need source data but it needs to store its model, those models can be MASSIVE. Which means expensive, to make it accessible you need to store it in memory... to make it better it needs to be faster... faster larger memory is $$$.