AI has emerged as the defining technology of our era, as transformative over time as the steam engine, electricity, and the Internet. While still in its early stages, AI is already reshaping the economy, society, our personal lives and the very nature of work, with much more to come as the technology and applications mature, including the scientific research necessary to anticipate and fight future pandemics. But, what’s the current state of AI and where is it heading?
To help address these important questions, the annual Artificial Intelligence Index was created under the auspices of the Stanford Institute for Human-Centered Artificial Intelligence, - guided by a Steering Committee of experts from academia and industry, and in collaboration with a number of sponsoring partners and data contributors. The AI Index aims to “provide unbiased, rigorous, and comprehensive data for policymakers, researchers, journalists, executives, and the general public to develop a deeper understanding of the complex field of AI.”
Now in its third year, the 2019 AI Index Report was released this past December. So was the associated 2019 AI Index Data, which includes three times as many data sets as the 2018 edition. The report comprises nine chapters. Let me summarize a few of its highlights.
Technical Performance. Advances in AI computational power are truly impressive. Prior to 2012, AI performance doubled every two years, closely tracking Moore’s Law. Post-2012, AI performance has been doubling every 3.4 months, advancing by more than 300,000x, instead of the 7x increase that would have resulted from the previous 2-year doubling period. This is largely due to the development of a variety of AI accelerators, - such as GPUs and TPUs, - specifically designed to boost the performance of the algorithms used in machine learning, convolutional neural networks, machine vision and other AI applications.
As a result, “the time required to train a large image classification system on cloud infrastructure has fallen from about three hours in October 2017 to about 88 seconds in July, 2019. During the same period, the cost to train such a system has fallen similarly.” These advances are enabling selected AI applications to achieve or surpass human levels of performance, including image and speech recognition, skin cancer classification, breast cancer detection, and prostate cancer grading. But, the report cautions that these tasks are highly specific, “and the achievements, while impressive, say nothing about the ability of the systems to generalize to other tasks.”
The Economy. While there’s has been an increase in hiring for all categories of AI jobs over the past few years, they remain a small share of total jobs. In the US, the share of jobs in AI-related topics increased from 0.26% of total jobs posted in 2010 to 1.32% in October of 2019, with the highest share in machine learning at .51%. The demand for AI jobs has increased across all industry sectors. As the share of total jobs posted within each industry, the highest percentage is in the IT sector at 2.3% of the total IT jobs posted, followed by Professional Services (2%), Finance and Insurance (1.3%) and Manufacturing (1.1%). Demand for AI related positions is also increasing around the world, with Singapore, Brazil, Australia and Canada experiencing the fastest growth.
Global investments in AI startups continue to increase, from a total of $1.3B raised in 2010 to over $40.4B in 2018, an average annual growth rate of nearly 50%. The number of AI companies that received funding increased from under 1,000 in 2014 to over 3,000 in 2018. The US is the overall leader in AI investments, followed by China, but their investment dynamics are quite different. US investments are spread over a relatively large number of startups, while Chinese investments are focused on a select few large firms. Relative to their size, Israel and Singapore had the largest investments in AI startups on a per capita basis.
Over the last year, the largest share of global investment went to Autonomous Vehicles (9.9% of the total), followed by Drug, Cancer and Therapy (6.1%), Facial Recognition (6.0%), Video Content (4.5%) and Fraud Detection and Finance (3.9%).
Education. “Enrollment continues to grow rapidly in AI and related subjects, both at traditional universities in the US and internationally, and in online offerings.” In North America, over 21% of computer science PhD students specialize in AI/ML, over twice the percentage of the second most popular specialization. In 2018, over 60% of AI PhD graduates from US universities went to industry, compared to 20% in 2004, and over twice the number that took academic jobs in the US.
In addition, the number of international AI PhD students graduating from North American universities has increased from less than 40% in 2010 to over 60% in 2018, with only around 10% leaving the US after graduating. However, gender diversity has not shown much progress in the US, as the percent of female AI PhD recipients has remained virtually constant at around 20% since 2010.
Societal Considerations. “AI systems raise a broad variety of ethical challenges that are now the concern of government, public interest organizations, NGO’s, academia, and industry. Efforts to identify these challenges and to develop guiding principles for ethically and socially responsible AI systems are emerging from each of these sectors.”
An analysis of nearly 60 documents dealing with AI ethical issues revealed that attention has been focused on 12 key ethical challenges. Fairness, Interpretability and Explainability; and Transparency were mentioned in over 80% of the documents analyzed. Accountability; Data Privacy and Reliability; and Robustness and Security were mentioned in over 60% of the documents. Human Control; Safety; and Diversity and Inclusion were mentioned in over 40%. Lawfulness and Compliance; Multi Stakeholder Engagement; and Sustainability were the least mentioned challenges.
“Global news coverage of Artificial Intelligence has increasingly shifted toward discussions about its ethical use.” To better understand how these narratives are taking shape, the AI Index initiative analyzed news articles from around 60,000 global English news sources and over 500,000 blogs on AI ethics published between August, 2018 and August, 2019. The analysis found seven major themes across all the various articles. Nearly a third (32%) of the articles dealt with the various AI Frameworks and Guidelines proposed by governments and policy institutions around the world. Other major themes were Data Privacy Issues, found in 14% of all articles; Facial Recognition (13%); Algorithm Bias (11%); Tech Ethics (11%); Ethics in Robotics and Driverless Cars (9%); and AI Transparency (6.7%).
To help researchers navigate its large volumes of data, the 2019 AI Index developed the interactive Global AI Vibrancy Tool. The tool makes it easier to compare the activities of different countries, including cross-country perspectives as well as country-specific drill downs. “Though it is tempting to provide a single ranking of countries, such comparisons are notoriously tricky. Instead, we’ve provided a tool for the reader to set the parameters and obtain the perspective they find most relevant when comparing countries.”
“This tool helps dispel the common impression that AI development is largely a tussle between the US and China. Reality is much more nuanced. Our data shows that local centers of AI excellence are emerging across the globe. For example, Finland excels in AI education, India demonstrates great AI skill penetration, Singapore has well-organized government support for AI, and Israel shows a lot of private investment in AI startups per capita.”
Comments