Scientists have assessed the success of AI for 2018

Massachusetts Institute of technology, Stanford and Harvard universities, jointly with OpenAI and a number of others, published the second annual AI Index study, which assessed the level of development, distribution, investment and other indicators in the field of artificial intelligence (AI).

The aim of the scientists was to assess the progress in the field of AI on specific quantitative indicators and to understand this progress (.pdf). A variety of indicators were evaluated: from the number of published works on AI to media and public interest in the topic.

For example, the number of scientific papers on AI has increased eight-fold since 1996. 28% of such work falls on Europe, 25% — on China (here the number of works in this area has increased since 2007 by 150%), 17% — on the United States. Interest in AI also shows an increase in the number of students attending introductory courses in related specialties. In the US alone, their number increased 3.4 times compared to 2012.

The number of active startups engaged in AI in the US increased 2.1 times from 2015 to 2018, while the total number of startups only 1.3 times. Venture capital investments in these companies increased in 2013 by 4.5 times (venture capital investment as a whole — 2.1 times).

The introduction of AI in companies has increased significantly. We are talking about a variety of areas (from the pharmaceutical industry to retail) and different applications (process automation, machine learning, interactive interfaces, machine understanding of natural languages, unmanned vehicles, etc.). So, more and more companies are introducing industrial robots. For example, in China in 2017 they were installed six times more than in 2012.

The authors collected data on AI-related SOFTWARE: the number of downloads of operating systems for robots has increased by 352% since 2014.

The report also assesses the speed of AI training. For example, the quality of machine translation from German into English (according to the so-called BLEU rating system) has increased three and a half times since 2008.

The report assesses whether AI is mentioned in the media in a positive, negative or neutral way. The number of positive references has increased by 2.5 times since 2016. The number of references to AI and machine learning in government and parliamentary statements and documents in the US, Canada, and the UK has also been calculated. In all three countries, the number of such references increased dramatically in 2018.

Experts believe that soon it will be necessary to evaluate the use of AI in all new areas and develop new techniques. "If we want AI to multiply human intelligence, and not replace it, we must move from the environment of "man against machine" to the environment of "man plus machine". The parameter that will demonstrate progress in this can be the number of non-Autonomous systems in which the final decision is made by man, but which are supported by machines", - says Professor of the University of Padua and Deputy editor-in-chief of the Journal of Artificial Intelligence Research Francesca Rossi.

According to Professor Toby Walsh of the Australian University of New South Wales, the index should also assess the growth of public investment in AI by country and the expansion of its use in the military sphere. "This is often called the "AI race" in which countries such as China demonstrate their desire to achieve dominance through AI," - said Mr. Walsh.