AI Definitions Affect Policymaking

Abstract

The task of artificial intelligence policymaking is complex and challenging, made all the more difficult by such a rapidly evolving technology. In order to address the security and economic implications of AI, policymakers must be able to viably define, categorize and assess AI research and technology. In this issue brief, CSET puts forward a functional definition of AI, based on three core principles, that significantly outperforms methods developed over the last decade.

Trends in AI Research for the Visual Surveillance of Populations

“…For instance, past research has identified China as the largest producer of AI papers, particularly in computer vision. 10 Yet computer vision is a broad field, comprising hundreds of thousands of papers and approximately one-third of all published AI research. 11 To better understand trends in computer vision research, we analyzed publications in terms of research tasks: the particular problems that papers address.…”

“…10 Yet computer vision is a broad field, comprising hundreds of thousands of papers and approximately one-third of all published AI research. 11 To better understand trends in computer vision research, we analyzed publications in terms of research tasks: the particular problems that papers address. Papers With Code, a site that tracks AI benchmarks, lists more than a thousand research tasks in computer vision, from general ones like object detection to specific tasks such as handwritten digit recognition.…”

The Public AI Research Portfolio of China’s Security Forces

“…This work includes, but is not limited to: target detection, object tracking, hyperspectral imaging, facial recognition, gait recognition, vehicle detection, forgery detection, and fingerprint recognition. This emphasis on AI-based computer vision is also seen more broadly in Chinese paper and patenting activity across the entire country 14. Natural language processing garners considerably less attention overall, and often focuses on areas such as sentiment analysis, analysis of online social media, and text classification.…”