Let’s return to the question that prompted this deliberation in the first place: Do we need an FDA for AI? Many of those who participated in the conversation arrived at the answer, “No, at least not in that exact form.” But valuable lessons were derived from deliberating deeply on the history and regulatory functioning of a single agency, considering AI governance not from the perspective of the current status quo but instead thinking through what AI might be if regulated differently. To this end, we landed on a set of concrete takeaways, highlighted above in the executive summary, as well as a number of points that will require further deliberation.
In this report, we’ve sought to “show our work” so that others might be able to learn from, and build upon, our conversation. It’s worth reiterating that the group did not seek to achieve consensus or any clear set of findings, nor did we arrive at these things. Instead, we engaged with the intent to have a grounded and deliberative conversation, siloed from the impetus to arrive at quick answers or to press for what’s immediately possible. By muddling through, we aim to get closer to the question that should be at the heart of any conversation about AI governance: What world do we want to live in, and what role should AI play in it?