A fascinating discussion at today’s QMUL/SCL/WorldBank event on AI Ethics and Regulations on how we should develop such ethics and regulations. There was general agreement that an ethical approach is essential if any new technology is to be trusted; also, probably, that researchers and developers should lead this through professionalising their practice. First steps are already being taken, with journals requiring that papers consider ethical issues: both those that can be addressed in design and implementation, and those that need to be monitored as the technology or context develops.
However for some fields, applications and contexts, society will not trust AI (and there are already plenty of people ready to lead a backlash against it) unless there are legal sanctions as well as ethical codes. The process of developing those laws must identify and address the concerns of all stakeholders, not just legal or technology experts. That needs a lot of genuine consultation (the multi-stakeholder process of Internet governance was cited with approval), with stakeholders, legislators and regulators all being enabled to make their contributions. Participation must represent all those who may be affected: lack of prior technological or legal knowledge must not be a barrier. Some technologies may be found to be socially unacceptable.
Successful regulation should be a benefit to those who want to enter these markets, by stabilising society’s attitudes to them. This was illustrated by two contrasting “new technologies” of the past: GM crops, and the Warnock Committee on Human Fertilisation and Embryology.