• 0 Posts
  • 91 Comments
Joined 10 months ago
cake
Cake day: August 27th, 2023

help-circle
  • The laws are currently the same for everyone when it comes to what you can use to train an AI with. I, as an individual, can use whatever public facing data I wish to build or fine tune AI models, same as Microsoft.

    If we make copyright laws even stronger, the only one getting locked out of the game are the little guys. Microsoft, google and company can afford to pay ridiculous prices for datasets. What they don’t own mainly comes from aggregators like Reddit, Getty, Instagram and Stack.

    Boosting copyright laws essentially kill all legal forms of open source AI. It would force the open source scene to go underground as a pirate network and lead to the scenario you mentioned.













  • It’s not about collecting data, the fediverse freely shares all its data by nature.

    Threads has a lot more users and Meta would use that to attack smaller instances, a bit like what Lemmy world was doing a few months ago.

    The biggest instances usually have the most communities and if they ban other instances, rival members are more likely to migrate over than to create copy cat communities.

    It’s not terrible at the moment but bringing in threads is like throwing a shark in a shrimp tank.