Former OpenAI researcher Tian Yonglong joins Tencent to develop visual language models.

OpenAI talent continues to bleed away. Mainland media reports that former OpenAI researcher Tian Yonglong has recently joined Tencent (00700)’s large language model department, where he will participate in R&D related to vision-language models (VLM).

According to The Paper, Tian Yonglong’s core research areas at OpenAI previously included computer vision, visual representation learning, and generative models. He joined Google Research in Cambridge as a senior research scientist toward the end of 2022, then in May 2024 he moved to Google DeepMind to work on the development of visual perception models. In October of the same year, he transferred to OpenAI as a technical researcher, focusing on machine vision and multimodal representation learning research, among other areas.

For alumnus and colleague Yao Shunyu

This is another instance of Tencent poaching from OpenAI. Late last year, Tencent announced an upgrade to its large model R&D framework. It newly established an AI Infra Department, an AI Data Department, and a Data Computing Platform Department, appointing former OpenAI senior researcher Yao Shunyu as Chief Artificial Intelligence (AI) Scientist and the head of the AI Infra and Large Language Model department. According to Securities Times, Tian Yonglong and Yao Shunyu are both Tsinghua undergraduate alumni, and they were also core researchers at OpenAI in the same period. Yao Shunyu joined Tencent last September.

OpenAI Chief Futurist to resign after the end of July

Separately, OpenAI’s Chief Futurist Joshua Achiam also posted on his social website announcing that he will resign after the 24th of this month. “There are no concrete reasons to leave, and there are no special reasons for choosing to leave now,” he said, adding that he had considered it for some time and that now is the time. He said that in 2017, he joined OpenAI as an intern at age 25. “Back then, computers couldn’t speak; now computers can solve cutting-edge scientific problems. So much has happened over these 10 years—it feels like several centuries.”