Artificial intelligence technology has grown more powerful and sophisticated as new advances build on old ones, and the pace just keeps accelerating. Experts point to the difference between how long it took AI to beat a game like chess to how long it took AI to beat humans at Go -- years versus days. Now, as 2018 comes to a close and technology's pace continues to accelerate. We are living in a world where if you aren't already doing some kind of agent based technology now, you are probably behind the curve, according to Gartner analyst Janelle Hill, speaking about the adoption of AI in the enterprise at the most recent Gartner Symposium.
But chatbots are just the entry-level technology for many enterprises. There are loads of other advances and innovations on the way, enabled by artificial intelligence and the technologies that comprise it -- machine learning, deep learning, natural language processing, computer vision, and more.
Enterprises have a long way to go to fully leverage these technologies to enable their businesses. It's one thing to have a vision, a plan, or a pilot. It's another thing entirely to put that into practice in a way that makes a meaningful impact on the business, by either improving operations, increasing sales, creating new markets, or some other way.
Advances in technology and the demand for artificial intelligence in businesses and other organizations are driving changes in other places, too. For instance, the high demand for more technology professionals that specialize in AI, machine learning, natural language processing (NLP) and related technologies is continuing and accelerating. Some organizations, recognizing the challenge have changed the way they look at their teams. Maybe they are splitting that data science unicorn job into a team of several workers, each with a specialized expertise -- a statistician, a developer, a business expert. Perhaps they have set their data scientists to work enabling more self-service options for analytics users, bringing the ability to find the business value to the masses.
At the same time, universities and other educational institutions have risen to the challenge to graduate more experts in artificial intelligence disciplines such as machine learning and deep learning. Universities may be expanding the courses they already have available to teach the skills needed. Other universities may be adding these disciplines for the first time. Recognizing the strong job market for these professionals, students are enrolling in these courses.
In this environment, the acceleration of AI innovation is captured in a series of data points and milestones collected from 2017 (as the most recent year that information has been published so far in 2018) and from 2018. The results are published in a paper by Stanford University, The AI Index 2018 Annual Report and compiled by a team of experts from organizations including MIT, Stanford, Harvard, OpenAI, and McKinsey Global Institute, among others. Here are a few of the data points these experts highlighted.
AI development is strong all over the world, but the US lags in peer reviewed research papers. An analysis of Scopus, the largest database of peer reviewed literature, reveals that 83% of 2017 AI papers originate outside the US, with 28% of originating in Europe, the largest percentage of any region. Papers published in China increased 150% between 2007 and 2017.
However, if you look at the number of papers submitted to the Advancement of Artificial Intelligence conference, held in February 2018 in New Orleans, China and the US had the largest number of submitted papers, respectively. A full 29% of US papers were accepted of 268 total submissions. A full 21% of China-affiliated papers were accepted of 265 total. Germany and Italy enjoyed the highest acceptance rate at 41% each, but far fewer papers were submitted by AI researchers affiliated with those countries.
University students across the US and around the world are pursuing educations in AI and machine learning in greater numbers, according to the report. For instance, at five US universities known for their computing programs [University of California, Berkeley; Carnegie Mellon University; Stanford; University of Illinois at Urbana-Champaign; and University of Washington, Seattle], 2017 introductory machine learning course enrollment was 5x greater than it was in 2012.
Interest in these courses is growing internationally, too. For instance, combined AI and machine learning enrollment at Tsinghua University in Beijing was 16x greater in 2017 than it was in 2010. However, the report notes that growth in AI course enrollment really depended on the school.
The vast majority of AI professors at the top schools are male. On average, 80% of AI professors are men at the following universities: UC Berkeley, Stanford, University of Illinois Urbana-Champaign, Carnegie Mellon, University of Washington Seattle, University College London, Oxford University, and ETH Zurich. The report authors caution that this is a relatively limited number of schools, so that the findings represent a small view of a much larger picture.
Attendance at large AI conferences (those with 2,000 or more attendees) has grown dramatically over the past 5 years.
The events with the highest attendance, NeurIPS (originally NIPS), CVPR, and ICML, have experienced the most growth in attendance since 2012. NeurIPS and ICML are growing at the fastest rate -- 4.8x and 6.8x their 2012 attendance, respectively, revealing the high level of interest in machine learning.
Meanwhile, conferences that focus on symbolic reasoning are experiencing relatively little growth, according to the report.
The annual workshop hosted by Women in Machine Learning (WiML), has welcomed 600% more participants than it had in 2014.
AI4All, an AI education initiative designed to increase diversity and inclusion in AI has 900% more alumni than it had in 2015.
Machine learning is the skill cited most frequently as a requirement in AI job ads on Monster.com, but deep learning is growing, according to the report. Deep learning skills as a requirement in job ads has increased by 35x from 2015 to 2017.
Men make up 71% of the applicant pool for AI jobs in the US, according to the report which cites Gartner TalentNeuron. Here's the breakdown of applicants by gender and specialized skills:
TensorFlow currently the most popular technology framework compared to the others, according to the report, based on downloads from GitHub. But the paper's authors also point out that a recent trend is the growing popularity of frameworks backed by major companies. For instance, TensorFlow is backed by Google, Pytorch is backed by Facebook, and mxnet is backed by Amazon.Jessica Davis has spent a career covering the intersection of business and technology at titles including IDG's Infoworld, Ziff Davis Enterprise's eWeek and Channel Insider, and Penton Technology's MSPmentor. She's passionate about the practical use of business intelligence, ... View Full Bio