Govts need to balance ethics, AI innovation

The biggest governance dilemma in AI is setting guidelines for the technology’s ethical use without unduly weakening the incentive to innovate.

- J. P. Singh

The concept of responsible AI (RAI) exemplifies this idealistic rhetoric. The principles it espouses — from ensuring that algorithms are not based on faulty data sets to preventing privacy and human-rights violations — are worthy. Where RAI falls short is in showing how these ideals should be incorporated into AI governance, and how to balance regulation with incentives for continued innovation.

Nonetheless, RAI has been embraced by many governments, which have incorporated relevant language into national AI policies. International organisations have also championed RAI — Unesco's Global AI Ethics and Governance Observatory is a leading example — with the goal of shaping international norms and national policy. But such top-down approaches contrast sharply with the deliberative, bottom-up decision-making that has proven most effective in addressing problems requiring collective action and coordination.

Meanwhile, corporations are touting their supposed commitment to RAT, often while resisting the regulations. Even universities have jumped on the RAI bandwagon, offering Al ethics courses in computer science departments. AI governance courses, however, are usually offered in other departments, so computer science students may not take them.