4 ms·

I don't think Reddit was telling the truth either. But AI developers/companies seem to almost universally believe they have fair-use rights to train their mode

by jprete 3y ago

I don't think Reddit was telling the truth either. But AI developers/companies seem to almost universally believe they have fair-use rights to train their models on any data they can get their hands on, and a sufficiently expensive API at least forces them to do engineering work to get all the data. So at the time I believed Reddit's reasoning.

tetris11 3y ago
Pushshift was the best unofficial reddit API for years, and did so without any funding problems
- jprete 3y ago
  I'm talking about deterring ML data harvesting, or at least getting a better price for it.
  - Semaphor 3y ago
    If it were about ML, why did they go heavily after third-party apps, even making up lies about one of the developers?
    - jprete 3y ago
      Because it wasn't about ML. As I said, I believe Reddit was lying. But the ML story was at least plausible at the time.
johnnyanmac 3y ago
>But AI developers/companies seem to almost universally believe they have fair-use rights to train their models on any data they can get their hands on Google and many other companies for almost 2 decades have spent their time scraping petabytes of data from the web. A lot of that with no expectation of payment. Some companies became billionaires off of that ability to freely access mass bulks of data. Data scraping has always been a grey area, but I find it strange how it's suddenly taken a turn for some people whenever modern AI comes up. We can't really be drawing lines based on what we feel is good/evil, because we will never agree as a whole on what is good/evil.