creators

Jonathan Ho - DDPM author, Ideogram co-founder

Jonathan Ho wrote the 2020 Denoising Diffusion Probabilistic Models paper and is named on Ideogram's founding team, after a Berkeley PhD under Pieter Abbeel.

Jonathan Ho received his PhD in computer science from UC Berkeley in 2020, advised by Pieter Abbeel. His early work was in reinforcement and imitation learning, including generative adversarial imitation learning, written with Stefano Ermon in 2016. He is the first author of "Denoising Diffusion Probabilistic Models" (2020), the paper that made diffusion a workable way to generate images. He followed it with classifier-free guidance in 2022, cascaded diffusion, video diffusion models and Imagen, Google's text-to-image system. He co-founded Ideogram with three other Google Brain researchers, Mohammad Norouzi, William Chan and Chitwan Saharia; the company's own launch page of 22 August 2023 names the founding team and $16.5 million in seed funding, and Norouzi announced it on LinkedIn the same week. Ideogram is still going and published open weights for its Ideogram 4 model in 2026. No page he or a company controls states what he does now: his own site stops at 2020 and neither Ideogram nor any other employer has published anything about him since.

Specialization

diffusion models, generative modelling, text-to-image, reinforcement learning