Artificial intelligence and its effect on employment and skilling

Abstract

Autonomous systems have traditionally focused on performing either routine manual tasks requiring well-defined spatial motions or routine cognitive tasks that could be unpacked into explicit instructions for computers. Advances in automation technology specifically in the field of Artificial Intelligence (AI) have attempted to perform non-routine tasks by utilizing large and representative examples from the problem domain to train algorithms which can make predictions over the domain objects. Stakeholders such as industries and governments can choose to deploy AI in diverse forms, such as — develop entirely new products and services, amplify human abilities in existing tasks or automate tasks by replacing humans with algorithms and robots. The task of choosing among these alternatives is non-trivial as it demands balancing objectives that can be fuzzy, shifting and conflicting. These choices can affect workers who either become more productive due to their skills being complemented by AI or made redundant as they relinquish their comparative advantage to machines. In the above stated context, this chapter discusses the state-of-the-art performance of AI specifically in tasks associated with vision and language. We examine literature that unravels the relationship between task characteristics and their suitability for AI techniques with examples from sectors such as healthcare, financial services, manufacturing and retail. We then discuss the different economic incentives for firms to adopt AI and review its impact on labor markets through contemporary economic models of AI-based automation. Subsequently, we propose a framework that formulates AI adoption as a multi-objective optimization process in which pluralistic perspectives of agents such as firms and workers can be modeled to aid policymakers evaluate AI deployment choices.