China’s Tech Industry Faces AI Brain Exodus: Associate Dean of Alibaba’s AI Powerhouse Departs; Former JD.com VP Joins Duke University; How AI Erases Peter Parker from Spider-Man

Weekly China AI News from May 16 to May 22

Top news of the Week

Associate Dean of Alibaba’s Damo Academy Leaves

What’s new: Chinese e-commerce giant Alibaba just lost another top AI mastermind against the backdrop of a massive AI brain drain from China’s embattled tech industry.

Rong Jin, an Alibaba Group vice president and Associate Dean of Damo Academy (Alibaba’s research arm) has reportedly departed from the company. Jin is the latest well-known scientist who left Alibaba in the past six months, joining (Alan) Qi Yuan, former Chief AI Scientist and Vice President of Alibaba’s Ant Group, and Gang Wang, Head of Autonomous Driving Lab.

Who is Rong Jin? Jin is one founding member of Alibaba’s Damo Academy and Head of Damo’s Machine Intelligence Lab, where he developed Pailitao, a visual search application that allows tens of millions of users to search commodities with photos every day. Before joining Alibaba in 2015, Jin was a faculty member of the Computer and Science Engineering Department at Michigan State University. Jin received a Ph.D. degree in computer science from Carnegie Mellon University, Pittsburgh, PA, USA, in 2003.

Other quits: Another recent noteworthy departure is Jianqiang Yu, Vice President of JD.com and Head of R&D at JD Logistics. Last November, the Alibaba’s rival lost its most important technology lead, Bowen Zhou, who was the president of JD Cloud & AI Chair of JD.com’s Technology Committee.

What caused the AI brain exodus?

• China’s tech industry starts to feel the pain of the year-long crackdown and a macroeconomic slowdown. With slashed budgets and dropped stock prices, Chinese tech companies are not seen as an appealing place for top talents.
• Top AI talents are mostly scientists who excel at high-flying research projects rather than down-to-the-earth products.
• Most departed scientists have completed their missions at tech companies by incorporating the AI infrastructure into corporate businesses.
• AI research is demanding more theoretical innovations for the next breakthrough so we are seeing a brain flowing from industry to academia

Former Huawei AI Chief Scientist, JD.com VP Joins Duke University

What’s new: Jian Pei, a former JD.com VP and once an AI chief scientist at Huawei, will join the faculty of the Department of Electrical and Computer Engineering in Duke University’s Pratt School of Engineering beginning July 1, 2022. Pei shares joint appointments at Duke in the Department of Computer Science and the Department of Biostatistics and Bioinformatics, according to Duke University.

Who’s Jian Pei? Pei is one of the most well-known experts in the data science and mining field. Before joining Duke University, Pei has been a professor of computing science since 2004 at Simon Fraser University. His industry career includes a stint at both Huawei and JD.com to help build AI infrastructure and R&D. He earned a P.hD at Simon Fraser University and undergraduate and master’s degrees in the field from Shanghai Jiao Tong University.

Pei is one of the most-cited authors in data mining, database systems, and information retrieval. The textbook Data Mining: Concepts and Techniques he co-authored is the most popular guide to the principal ideas, techniques, and technologies of data mining. His published papers and books have accumulated a citation more than 92,000 times.

Pei is a Fellow of the Royal Society of Canada, a Fellow of the Canadian Academy of Engineering, a Fellow of the Association of Computing Machinery (ACM), and a Fellow of the Institute of Electrical and Electronics Engineers (IEEE).

Paper & Projects

Towards An End-to-End Framework for Flow-Guided Video Inpainting

Researchers from Nankai University and Hisilicon Technologies recently proposed an End-to-End framework for Flow-Guided Video Inpainting (E2FGVI) through elaborately designed three trainable modules, namely, flow completion, feature propagation, and content hallucination modules. The three modules correspond with the three stages of previous flow-based methods but can be jointly optimized, leading to a more efficient and effective inpainting process. Experimental results demonstrate that the proposed method outperforms state-of-the-art methods both qualitatively and quantitatively and shows promising efficiency. You can find the code here.

Language Models Can See: Plugging Visual Controls in Text Generation

Researchers from the University of Cambridge, Tencent AI Lab, University of Trento, DeepMind, and the University of Hong Kong, recently proposed a training-free framework called MAGIC (iMAge-Guided text generatIon with CLIP) which can plug in visual controls in the generation process and enable language models to perform multimodal tasks, such as image captioning and visually grounded story generation. Unlike existing multimodal models that rely on image-text pairs, MAGIC combines GPT-2 and CLIP for image-grounded text generation. MAGIC outperforms the state-of-the-art method by notable margins with a nearly 27 times decoding speedup. You can find the code on GitHub.

AI Recreates Symphony Music for InterStellar

A piece of symphony excerpt generated by AI recently gained traction on Reddit, which sounded just like the epic soundtrack in the movie InterStellar. The AI is called SymphonyNet, a symbolic symphony music generation solution presented by researchers from China’s Central Conservatory of Music, Oxford University, and Tsinghua University. To bridge the gap between text generation and symphony generation tasks, researchers proposed a novel Multi-track Multi-instrument Repeatable (MMR) representation with particular 3-D positional embedding and a modified Byte Pair Encoding algorithm (Music BPE) for music tokens. Their empirical results show that the proposed approach can generate coherent, novel, complex, and harmonious symphony compared to human composition. You can find the code on GitHub.

Rising Startups

Quillion Technology, a CPU provider, has raised RMB600 million (~$90 million) in its Series A funding round led by Lightspeed China Partners. Founded in 2021, the Guangdong-based company provides self-controllable, energy-efficient general-purpose CPU chips, enabling users with information system integration and technology consulting.

PhiGent Robotics, an autonomous driving solution provider, has raised $30 million in its Series A funding round led by INCE Capital. Founded in 2021, the Beijing-based startup focuses on research and development, and commercialization of AI-based visual imaging and 3D visual radar.

ZMO.AI, an AI-powered e-commerce software provider, has raised $8 million in its Series A funding round led by Hillhouse Capital. Founded by Chinese entrepreneurs, the company creates computer-generated clothing models to present a growing number of stock-keeping units (SKUs). Learn more about the company from this TechCrunch article.