Can Academia Offer a Model for AI Governance?

AI has been in the news for a while, but, more recently, attention has turned to calls for a deceleration in the pace of AI development. And the concerns extend well beyond the opposition to data centers. Rogue AI agents free to pursue their own goals could pose threats from large-scale cyberattacks and bioterrorism to widespread disruption.

Dario Amodei, CEO of Anthropic, has been leading the charge to decelerate the pace of innovation, what he refers to as pacing the frontier. His note on pacing the frontier is about slowing down further developments in AI to allow time to create more transparency and build in third-party verifiability through legislation and peer evaluation (embedded evaluators) systems. He has called for the use of ‘checkpoints’ to check on alignment of the system with goals.

Opinions differ on the intent behind this call and many suspect ulterior motives especially given the support for pacing from other CEOs like Sam Altman and Elon Musk. At the same time, the departure of Anthropic’s researcher Jacob Coxon and the hacking into Hugging Face point to real concerns with the use of AI.

At the risk of oversimplification, for those of us within academia, Amodei’s note on pacing the frontier sounds very much like building a system of governance around AI.

Amodei refers to a system with ‘embedded evaluators’ and describes the benefits that come with such evaluation including verifiability, transparency, and access to a second opinion.

Within the academy, we have always used review panels, external reviewers, and double-blind review processes to bring objectivity to our work.

Amodei calls for regulation of AI companies as well as for them to voluntarily work together through ‘industry groups.’ He alerts against the use of AI capabilities by autocracies and offers three levels of agreement across countries that could ensure global pacing.

Academia uses discipline-specific expert bodies to set standards for self-governance. Isn’t this what academic freedom is about – our ability to self-govern using disciplinary norms? These disciplinary standards have emerged from rigorous research, and discovery, and are grounded in the pursuit of truth.

Towards the end of his note, he points out:

…even if we cannot achieve formal agreements, simply changing informal norms may have some value.

We fully agree – research universities have such a model – one grounded in the principles of academic freedom.

Go Jags!

Latha Ramchand
Chancellor