Shulou(Shulou.com)11/24 Report--

The 2023 AI Index Report has been released! The report shows that China leads the world in AI summit papers, but the number of citations is lower than that of the United States. In addition, among the top ten institutions in the world in AI papers, China accounted for 9 seats, surpassing MIT one after another.

Stanford released its AI Index Report 2023 today.

It is worth noting that the Stanford AI Index report lists the top ten institutions in the world for the number of AI papers published, 9 of which are all from China and have surpassed MIT.

They are: Chinese Academy of Sciences, Tsinghua University, University of Chinese Academy of Sciences, Shanghai Jiaotong University, Zhejiang University, Harbin Institute of Technology, Beijing University of Aeronautics and Astronautics, University of Electronic Science and Technology, Peking University, and MIT.

This year's report is divided into eight sections: Research and Development, Technology Performance, Ethics of AI Technology, Economics, Education, Policy and Governance, Diversity, and Public Perspectives.

The following extracts some of the salient points of the report.

From 2010 to 2021, although the pace of cross-border collaboration on AI papers has slowed, the number of AI research collaborations between the United States and China has increased by about four times since 2010, 2.5 times more than the total number of collaborations between China and the United Kingdom.

However, from 2020 to 2021, the total number of China-US cooperation increased by only 2.1%, the smallest year-on-year growth rate since 2010.

In addition, the total number of AI papers has more than doubled since 2010. From 200,000 articles in 2010 to nearly 500,000 articles (49601) in 2021.

In terms of the type of AI papers published, 60% of all AI papers published in 2021 were journal articles, 17% were conference papers, and 13% came from repositories.

While journal and repository papers have grown 3-fold and 26.6-fold over the past 12 years, respectively, the number of conference papers has declined since 2019.

Pattern recognition, machine learning, and computer vision remain hot topics in AI research.

China continues to lead in the volume of journal, conference and repository papers.

The United States still leads in AI conferences and repository citations, but those leads are slowly eroding. Nevertheless, the majority of the world's large language models and multimodal models (54% in 2022) are produced by U.S. institutions.

China dominated the AI list, but the citation volume was lower than that of American AI journal papers. China always maintained the leading position, with 39.8% in 2021, followed by the EU and the UK (15.1%), and then the United States (10.0%).

Since 2010, the proportion of cited papers in Chinese artificial intelligence journals has gradually increased, while the European Union, the United Kingdom and the United States have all declined. China, EU, UK and USA accounted for 65.7% of the total global citations.

So, what about the publication of the World Summit papers?

In 2021, China accounted for the largest share of papers published at the global AI summit at 26.15%, followed by the EU and the UK at 20.29%, and the United States at 17.23%.

Judging from the citations of top conference papers, although China has a high yield, the citations are lower than those of the United States. The citation rate of American papers was 23.9%, and that of Chinese papers was 22.02%.

It can be seen from the side that China has the largest number of papers published, but the quality is not as high as that of the United States.

The United States leads the world in AI paper repository submissions with 23.48%. China has the lowest, 11.87%.

9 institutions in China, AI papers published catch up with MIT2021, China accounts for 9 of the top 10 institutions in the world in terms of the total number of papers published. The total number of papers published by different institutions is shown in the figure below. MIT ranks 10th, publishing 1745 papers.

In terms of computer vision (CV), ten institutions in China rank among the top ten in the world: Chinese Academy of Sciences, Shanghai Jiaotong University, University of Chinese Academy of Sciences, Tsinghua University, Zhejiang University, Beijing University of Aeronautics and Astronautics, Wuhan University, Beijing University of Technology, Harbin Institute of Technology, and Tianjin University.

In the field of natural language processing (NLP), things are different.

The top ten institutions/companies in the world are: Chinese Academy of Sciences, Carnegie Mellon University, Microsoft, Tsinghua University, Carnegie Mellon University-Australia, Google, Peking University, University of Chinese Academy of Sciences, Ali, Amazon.

Speech recognition is ranked as follows:

Among the major AI machine learning systems released by industry leading academia in 2022, language systems accounted for the most, with 23, six times the number of multimodal systems.

Industry is ahead of academia in terms of paper output.

Until 2014, most important models were published by academia. Since then, industry has reversed course. By 2022, 32 important machine learning models will be born in industry, compared to just three in academia.

Thus, building the most advanced AI systems increasingly requires a lot of data, computer power, and financial resources compared to nonprofit organizations and academia, and industry players certainly have more financial resources to do it.

In 2022, the United States produced the largest number of significant machine learning systems, with 16, followed by the United Kingdom (8) and China (3).

In addition, since 2002, the United States has surpassed the United Kingdom, the European Union and China in terms of the total number of important machine learning systems created.

Looking at the distribution of researchers behind these important AI systems, the United States has the largest number of researchers, 285, more than twice as many as Britain and nearly six times as many as China.

Large language and multimodal models, sometimes referred to as base models, are an emerging and increasingly popular type of AI model that is trained on large amounts of data and suitable for a variety of downstream applications.

Large-scale language and multimodal models such as ChatGPT, DALL-E 2, and MakeA-Video have demonstrated impressive capabilities and are beginning to be widely deployed in the real world.

By analyzing the national affiliations of the authors of these models, the majority of these researchers came from US institutions (54.2%).

The Stanford AI Index report also lists timelines for large language and multimodal model releases.

Large language models are getting bigger and more expensive.

The first large-scale language model, GPT-2, was released in 2019 with 1.5 billion parameters and a training cost of about $50000. Google PaLM is one of the large language models launched in 2022, with 540 billion parameters and a cost of up to $8 million.

In terms of parameters and training costs, PalM is 360 times larger and 160 times more expensive than GPT-2.

Not just PalM, but large language and multimodal models are getting bigger and more expensive overall.

For example, DeepMind's large-scale language model Chinchilla, launched in May 2022, cost an estimated $2.1 million, while BLOOM's training cost about $2.3 million.

Over time, GAN progressed in face generation, with the last image generated by Diffusion-GAN, a model that achieved the latest SOTA on STL-10.

Last year, with the release of models such as OpenAI's DALL-E 2, Stability AI's Stable Diffusion, Midjourney, Meta's Make-AScene, and Google's Imagen, text-to-image generation models gradually entered the public eye.

Enter the same prompt,"A panda playing the piano on a warm Paris night," generated by three publicly accessible AI text-to-image systems, DALL-E 2, Stable Diffusion, and Midjourney, respectively.

Of all the recently released text-to-image generation models, Google's Imagen performed best on the COCO benchmark.

This year, the Google researchers who created Imagen also released DrawBench, a harder text-to-image benchmark designed to challenge increasingly powerful text-to-image models.

In addition, the report also introduces some biases in current generative AI models, such as when the DELLE-2 is prompted to CEO, everyone seems to adopt a confident posture with crossed arms.

In Midjourney, when prompted to generate "Influential People," it generates four images of white men who look older.

For the full report see:

https://aiindex.stanford.edu/wp-content/uploads/2023/04/HAI_AI-Index_Report_2023.pdf

This article comes from Weixin Official Accounts: Xinzhiyuan (ID: AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.