Researchers call for more accountability in AI systems

The AI Now Institute at the New York University published its second annual report with recommendations for the field of artificial intelligence (AI). The report pays particular attention to the use of AI algorithms in areas such as criminal justice, healthcare, welfare, and education, and draws attention to the dangers of bias and inaccuracy in such algorithms. Among the recommendations, core public agencies are advised to no longer use black box AI and algorithmic systems; such systems should be available for public auditing, testing, and review, and subject to accountability standards. AI companies should rigorously test their AI system before release, to ensure that they do not amplify bias and errors due to any issues with the training data, algorithms, or other elements of system design. It is also recommended that AI bias research and mitigation strategies are expanded beyond a narrowly technical approach, and that ethical codes meant to steer the AI field should be accompanied by strong oversight and accountability mechanisms.