Luke Metz
Luke Metz is a co-author of DCGAN, a 2015 paper on deep convolutional generative adversarial networks, and of VeLO, a learned optimizer from 2022. The ICML 2021 conference named a paper on Persistent Evolution Strategies an Outstanding Paper, and ICLR gave DCGAN a Test of Time award in 2026. Together with Mira Murati, they co-founded Thinking Machines Lab, which announced 29 employees in February 2025. In January 2026, they returned to OpenAI, and since August 2026 they work at Meta Superintelligence Labs, reporting to Alexandr Wang.
Timeline
-
2015
Co-writing the DCGAN paper
In 2015, they co-write Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks, the DCGAN paper, alongside Alec Radford and Soumith Chintala.
-
2021
An ICML Outstanding Paper
Unbiased Gradient Estimation in Unrolled Computation Graphs with Persistent Evolution Strategies, written with Paul Vicol and Jascha Sohl-Dickstein, appears in the proceedings of ICML 2021. The conference names it an Outstanding Paper.
-
2022
The VeLO learned optimizer
In 2022, they co-write the VeLO paper, which trains an optimizer for deep learning that is itself a small neural network.
-
February 2025
Co-founding Thinking Machines Lab
They co-found Thinking Machines Lab with Mira Murati. In February 2025, the lab announces 29 employees, among them former OpenAI research scientists.
-
January 2026
Returning to OpenAI
In January 2026, OpenAI welcomes Luke Metz back, along with Barret Zoph and Sam Schoenholz.
-
2026
A Test of Time award
The ICLR 2026 program chairs select DCGAN for a Test of Time award, which recognizes papers published at ICLR 2016.
-
August 2026
Joining Meta Superintelligence Labs
Meta Superintelligence Labs hires them in August 2026, and they report to Alexandr Wang.
Sources and credits
Photo: We have no published photo of Luke Metz yet. Send us a free-licensed photo.
- [1511.06434] Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
- ICML 2021 Awards
- Unbiased Gradient Estimation in Unrolled Computation Graphs with Persistent Evolution Strategies
- [2211.09760] VeLO: Training Versatile Learned Optimizers by Scaling Up
- Former OpenAI CTO Mira Murati unveils Thinking Machines Lab details and leadership team stacked with former OpenAI colleagues | Fortune
- Mira Murati's startup, Thinking Machines Lab, is losing two of its co-founders to OpenAI | TechCrunch
- Announcing the Test of Time Awards from ICLR 2016 – ICLR Blog
- Meta hired Luke Metz seven months after OpenAI hired him back