Aran Komatsuzaki
Co-author of the GPT-J-6B open model; first author of the Sparse Upcycling paper from Google Research
Aran Komatsuzaki works on generative AI and, by Komatsuzaki's own account, co-led the development of GPT-J in 2021 and of the LAION image-text datasets. The GPT-J repository's citation entry for the GPT-J-6B weights, released under the Apache 2.0 license, names Ben Wang and Komatsuzaki as authors. During a 2022 internship at Google Research, Komatsuzaki co-proposed sparse upcycling, a way to turn a pretrained dense model into a mixture-of-experts model, and is first author of the paper that introduced it.123
Work
GPT-J-6B (external site: github.com)Open weights
Named with Ben Wang as an author in the repository's citation entry for the GPT-J-6B weights, which the README says are licensed under Apache 2.0, and thanked there for advice on experiment design and for writing the blog posts.21
Sparse Upcycling (external site: arxiv.org)Published research
First author of "Sparse Upcycling: Training Mixture-of-Experts from Dense Checkpoints" (2022), which reuses a dense checkpoint to initialize a sparsely activated mixture-of-experts model.31
Other documented roles
Roles the sources document but that are not confirmed as current.
- Intern (2022), Google Research1
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.