Splinelinus

3 ms·

1. 1.▲
   by Splinelinus 7mo ago

   This is also my thinking. A(ffero)GPL does something similar by saying a user of an API to AGPL code is bound by the AGPL license. You can always choose not to use the code, and not to use the license. For the parent comment on discoverabil
2. 2.▲
   by Splinelinus 7mo ago

   I'm waiting for AGPL to become AIGPL: If you train a model with some or all of the licensed work, you agree that the weights of that model constitute a derivative work, and further for the weights, as well as any inference output produ
3. 3.▲
   by Splinelinus 8mo ago

   I think you're seeing a magic trick that's been around forever: https://www.youtube.com/watch?v=o-KDuOvMAfQ&t=80s
4. 4.▲
   by Splinelinus 2y ago

   This article is kind of vague on that tbf: To conclude, we observed no credible evidence for a beneficial effect of L-dopa (vs. Haloperidol) on reinforcement learning in a reward context, as well as the proposed mechanistic account of an en