I am once again not holding my breath in hopes that model collapse saves us from genAI
Reposted from
Tim Kellogg
Model collapse doesn’t happen in practice
The short story is that synthetic data is used contextually, not blindly, so there’s no actual evidence that it’s happening, or will happen
arxiv.org/abs/2410.16713
The short story is that synthetic data is used contextually, not blindly, so there’s no actual evidence that it’s happening, or will happen
arxiv.org/abs/2410.16713
Comments