A group of AI experts have published the the Artificial Intelligence 2018 annual report, detailing the growth in AI academic research, use by industry, mentions by government, patents and technical performance in computer vision and natural language processing.
While the first report last year focused on North American AI activities, this year's report includes efforts in Europe, China, South Korea and Japan.
One measure of AI activity across regions was by output of academic papers. On this count, Europe was leading, accounting for 28 percent of AI papers last year, followed by China, which accounted for 25 percent, and the US with 17 percent.
The most widely covered topics were machine learning and probabilistic learning, neural networks, and computer vision.
AI research in different countries focused on different fields, with papers from China concentrating on engineering, technology, and agricultural sciences, while the US and Europe focused on humanities and medical and health sciences.
Another notable difference between countries was that in 2017 the Chinese government produced four times as many AI papers than Chinese corporations, whereas in the US corporate AI papers lead the way.
Growing interest in AI is also reflected in course enrollments. In the US, enrollments in introductory AI were 3.4 times larger in 2017 than in 2012. In the same period introductory machine-learning enrollments (ML) grew by five times. AI and ML course enrollments at China's Tingshua University in 2017 were 16 times greater than they were in 2010.
Downloads of robot software have also taken off across the globe, particularly in China. China is also leading the world in the number of robots that are installed each, and has seen a 500 percent rise since 2012. Last year almost 150,000 robots were installed in China compared to fewer than 50,000 in North America.
The report digs into mentions of ML and AI in Canadian and UK parliaments, as well as mentions in the US Congressional Record. From 1995 to 2015, there were less than 25 mentions of the technology each year in US Congress. In 2018 there were 100 mentions. In the UK, the technologies were barely mentioned until 2015, while in 2018 mentions skyrocketed to nearly 300.
The report also tracks human-level performance milestones of AI. In 1997 IBM's DeepBlue beat chess champion Gary Kasparov, and in 2011 IBM Watson won Jeopardy. By 2016, Google DeepMind's AlphaGo beat leading Go player Lee Sedol. This year, a DeepMind agent reached human level performance in 3D multiplayer first person game, Quake III Arena Capture the Flag.
Notably absent from the report is any analysis of military use of AI and government spending on the technology. As noted by UNSW Sydney AI researcher Toby Walsh, some governments including the UK, France, and Germany have committed billions to AI. Walsh would like to see a breakdown of investments by country as well use of AI by military.
PREVIOUS AND RELATED COVERAGE
San Francisco-based cloud software startup People.ai is building data "pipelines" using machine learning that will automate much of the drudgery of updating contacts in systems such as Salesforce.com.
The self-portrait raises questions about the nature of creativity and human expression.
Bridging the gap between the edge and the cloud in order to bring accelerated vision features and AI performance to the edge of the Internet of Things.
The search giant says it needs to work through "important technology and policy questions" first.
David Karandish, founder and CEO of Jane.ai, discusses the next step for machine learning and AI through chatbots and natural language processing.
Major breakthroughs will lead enterprises to increasingly advance, scale, and trust artificial intelligence, according to IBM.