Regulatory Agencies Reaffirm Commitment to "Responsible Innovation" in AI
Commentary by Steven Lofchie
The CFPB, the DOJ, FTC and the Equal Employment Opportunity Commission reaffirmed their commitment to monitoring the development and use of artificial intelligence ("AI") to ensure "responsible innovation" and protect individuals’ rights.
In a joint statement, the agencies said that while AI tools promise advancements, their use has the potential for unlawful bias. The agencies highlighted several ways in which discriminatory practices can stem from such tools:
- unrepresentative or imbalanced datasets resulting in skewed outcomes;
- model opacity (a/k/a "black boxes") of an automated system’s inner workings which often times are not clear even to the developer of the tool;
- a lack of understanding as to how automated systems tools will be used by private or public entities; and
- systems designed by developers based on "flawed assumptions" about its users.
FTC Chair Lina Khan said that "[t]echnological advances can deliver critical innovation – but claims of innovation must not be cover for lawbreaking."
CFPB Chair Rohit Chopra added that new methods to improve lending and marketing are not "inherently bad," but that "when done in irresponsible ways, such as creating black box models or not carefully studying the data input for bias, these products and services pose real threats to . . . nascent firms and entrepreneurs trying to compete with those who violate the law."
Commentary
The worry is that the regulators will make the same types of mistakes with the regulation of AI as regulators made with digital assets--fail to adopt regulations that are appropriate to the product. In the case of AI, the demand that developers fully understand the logic of the systems is not possible. AI is not simply a more complicated logic tree, like an expert system with multiple branches. Much of AI is, at least at this point in its development, inherently a black box. If the regulators can not modify their requirements to meet the inherent limits of the product, it puts everyone outside the law, regardless of their efforts at compliance.