How To Fine-Tune GPT-J
- If you want to fine-tune a large language model, GPT-J is a good place to start (at least now, in July 2021). I wrote a guide on how to do so, part of the official repo here.
This Bot Crime Did Not Occur
- StyleGAN 2 ADA, trained on 10,000 mugshots from the Illinois Department of Corrections, only a fraction of the nearly 70,000 mugshots in their database.
- The US imprisons too many people; mugshots shouldn’t be so widely available. More in the pinned tweet thread.
Works In Progress
- a novel
- a game with N.H.
- more wyrd art
- data cleaning stuff
- clean data, see above
- A “spiritual guide” written in conjunction with GPT-3. My first AI-assisted publication.