China surpassed the U.S. in AI research journal citations last year, according to the latest Artificial Intelligence Index report. Stanford University’s Institute for Human-Centered Artificial Intelligence (HAI) released the report, which outlines various benchmarks and statistics about world progress in AI. We've highlighted the major takeaways below:
- The funds committed to companies and projects in AI rose to $13.8B, more than 4.5 times that of 2019. This was partially driven by the pandemic, as more firms turned to machine-learning techniques to find drug options and help develop vaccines.
- Private AI investment grew by 9.3% last year, higher than the 5.7% from 2018-19. The drug discovery and design sector took in the most private investment of any sector, at $13.8B.
- However, the number of newly funded AI companies fell for the third consecutive year, showing that "we're moving from pure research and exploratory small startups to industrial-stage companies."
- Fewer than 1,000 startups received took in funding in 2020, compared to more than 4,000 in 2017:
- Experts cited in the report mentioned OpenAI’s text-generation program GPT-3 and DeepMind’s AlphaFold program as the most significant AI breakthroughs of 2020.
- There was a near 35% increase in AI research publications from 2019-2020, surpassing the 19.6% growth from 2018-19.
- AI-related papers on pre-print server arXiv grew to nearly 35,000 in 2020, from 5,500 in 2015.
- Mentions of AI/ML during U.S. congressional proceedings have gone up noticeably since 2016:
- In 2019, the last year data was available, 65% of North American AI PhDs moved on to work in the industry, up from 44% in 2010.
- That year, 64.3% of AI PhDs in North America were foreign students, an increase of 4.3% from the year before. 82% who graduated stayed and took jobs in the U.S.
- From 2016-2020, Brazil, India, Canada, Singapore, and South Africa experienced the most AI hiring, according to LinkedIn data.
- AI continues to face a lack of diversity. The majority of new AI PhD students in the U.S. were white in 2019, compared to only 2.4% who were Black and 3.2% who were Hispanic.
- The report found that AI ethics is growing, but still faces "disappointingly small" interest among larger companies. The field's ethical challenges will create "a new frontier of potential privacy violations and other abuses," according to HAI senior fellow Erik Brynjolfsson, director of the Stanford Digital Economy Lab.
- In 2020, AI ethics stories dominated news coverage, including stories about the departure of AI researcher Timnit Gebru from Google and various governments' ethics initiatives.
- The report calls for more testing and benchmarks in ethics, computer vision, and NLP.
- The AI Index is compiled by HAI and an 11-member steering committee, with contributions from SRI International, Harvard University, OECD, and the Partnership on AI.
(20min delay)