Identifying and measuring developments in artificial intelligence

Abstract

OECD Working Papers should not be reported as representing the official views of the OECD or of its member countries. The opinions expressed and arguments employed are those of the authors. Working Papers describe preliminary results or research in progress by the author(s) and are published to stimulate discussion on a broad range of issues on which the OECD works. Comments on Working Papers are welcomed, and may be sent to OECD

Advances in De Novo Drug Design: From Conventional to Machine Learning Methods

“…QMRFs are an essential part of the validation and acceptance of QSARs for use in regulatory decision making, and thus, similar approaches for DL models would be an essential next step. While the OECD have not yet developed specific guidance for DL models, they have published a set of values-based principles for the development of AI methods [147] and position papers on using AI to help combat COVID-19 [148], and on identifying and measuring developments in AI [149].…”

Measuring the AI content of government-funded R&D projects

“…Key term matching presents a number of challenges, since there is at present no consensus on a standard set of key terms that comprehensively and unambiguously represent AI-related R&D; moreover, such a set is bound to be specific to different corpora (scientific publications, R&D project proposals, patent claims, job descriptions, company reports, etc…) and vary over time. As noted in Baruffaldi et al (2020), failing to capture all potentially relevant key terms risks overlooking many AI-related projects, thus underestimating their total number. This problem can also arise when the title and abstract of project applications do not contain sufficient information on the research methodology to be used in a given project.…”

“…There is a wide and fast growing literature dealing with field-specific topic extraction from several corpora, mostly publications, with some efforts looking at AI in particular, as documented in (Cockburn et al, 2018[17]) and previous OECD work aimed at identifying and measuring Artificial Intelligence (AI)-related developments in science, as captured in scientific publications; technological developments, as proxied by patents; and software, particularly open source software (Baruffaldi et al, 2020[18]). However, there are fewer precedents when it comes to R&D funding data (Abadi, He and Pecht, 2020 [19]; Annapureddy et al, 2020 [20]) 5 . Project funding data are products of the administrative processes of R&D funding organisations (ministries, agencies, etc.…”

“…Visualising this larger set of terms in a two-dimensional space is more challenging but still possible through a more sparing use of labels, as shown in Figure 2.3. 20 This extended visualisation differs from the previous one as it represents connections across a broader set of terms, all derived from an entirely AI-related corpus (the AI journals).…”

Artificial Intelligence and International System Structure

“…On the one hand, patents evolution, artificial intelligence and human-computer interaction publications by country. Those metrics approximate human capital and innovation and are the standard metrics used by several global AI reports (Capgemini Consulting 2018;McKinsey Global Institute 2018;Baruffaldi et al 2020;Tortoise 2020;Nature 2020;Zhang et al 2021) and by researchers studying innovation development and its future (Acemoglu et al 2016;Pugliese et al 2019).…”