Google’s parent company, Alphabet, published the world’s most cited paper in AI.Credit: Smith Collection/Gadot/Getty
US tech giants Alphabet and Microsoft have more citations to research papers on artificial intelligence (AI) than any other company, but Chinese companies Baidu and Tencent lead in patents.
This is according to PARAT, the Private Sector AI-related Activity Tracker, a tool run by the Emerging Technologies Observatory (ETO), which collects data on AI trends and has been undergoing significant updates.
PARAT’s latest data also includes data on the number of AI-related jobs at companies, as well as publication and patent outcomes. AI research and its products, especially generative models that create text and images, have become big business. Governments are considering how to regulate AI technology as it disrupts industries and raises questions about safety.
How does ChatGPT “think”? Psychology and neuroscience explain large-scale AI language models
In a field where cutting-edge research is happening in industry as much as in universities, keeping an eye on commercial activity is important, says Ngor Luong, who tracks AI investments and corporate activity at the Center for Security and Emerging Technologies, an AI-focused think tank at Georgetown University in Washington, D.C., that hosts ETO. Companies are at the forefront of AI innovation, she says.
International Expertise
Zachary Arnold, the institute’s principal analyst, says the data shows that China’s big players are highly competitive in AI, even when accounting for the quality of the research they produce. Three of China’s tech giants — Tencent, Alibaba and Huawei — are in the top 10 when you break down companies by the number of highly cited AI papers and preprints they produce (see “AI giants”). “There’s still a bias in Washington, D.C., and maybe elsewhere, that China is big and can do a lot, but it’s not really at the top of the heap,” Arnold says. But the ETO calculates several quality-adjusted metrics, and Chinese companies “are putting up impressive numbers” on these, he says.
Source: PARAT/ETO
According to PARAT data, the most cited paper in all of AI research is the 2017 paper1 “Attention is all you need”. The paper was written by researchers at US-based Alphabet subsidiary Google and is notable for describing the “Transformer” architecture that is now the basis of many generative AI models. Another highly cited example that includes Chinese authors is the paper2 co-authored by Tencent researchers, “ICNet for Real-Time Semantic Segmentation on High-Resolution Images”, which describes an improved method for identifying objects in images.
Of the 10 companies that filed the most AI patents over the past decade, only three were U.S. companies, with the rest coming from China, Germany, and South Korea. The Chinese government has long encouraged patenting, but in recent years has increasingly opposed granting patents free of charge, Luong adds.
Top Employers
The data also highlights the diversity of the field, Arnold says, with many more companies than the “big five” of Alphabet, Amazon, Apple, Meta and Microsoft. And when ranked by most citations to AI research, well-known companies like OpenAI and Apple appear alongside companies less known for AI innovation, such as Japanese conglomerate Mitsubishi and U.S. entertainment company Disney. Luong points out that PARAT’s paper and patent data only goes up to the end of 2023, missing more recent trends.
Other metrics reveal AI activity that is often overlooked, Arnold said. PARAT now includes figures on the number of AI-related jobs at companies, a metric based on data from social media platform LinkedIn and that is the most accurate for U.S. companies. He said the metric identifies companies with “a certain amount of AI talent.” Amazon leads the metric with 14,000 jobs, closely followed by multinational consulting firm Accenture. Large consulting firms now act as “hired gunmen” for other companies’ and government AI projects, Arnold said.
Being able to analyze a company’s activities from different perspectives is important, he adds: “There’s a lot of discussion about ‘who’s leading in AI?’ And being data nerds, we know that the data that can be used to answer that question is very diverse, and doesn’t necessarily point in exactly the same direction.”
