I had a coffee (actually, I had a burger) recently with a colleague discussing the issue of “right to access” in the context of AI. I remember discussions like this during the early (really early) internet when it was all dial-up and there was this word “broadband” that had entered the public policy discourse.

Broadband was this magical thing that would allow us to do all sorts of things online that we couldn’t do with dial-up. It was a promise of a future where the internet would be more accessible, more inclusive, and more powerful.

Nowadays, there's a new equally evocative/confusing/exciting/antagonizing term — “constitutional AI”.

And these constitutions are being crafted right now. Actually I should say that they have been crafted.

What kind of role would someone have if they were responsible for the development and enforcement of the constitution of an AI system? What would that job look like? What would the responsibilities be? What qualifications would be required?

In anticipation of next week's General Seminar, speculative job posting for a role that would be responsible for the development, “red-teaming” and enforcement of the constitution of an AI system.

You can see the artifact I created here.

I'm no expert on the specifics of these constitutions, although sure — there's good PR material and blog posts from the foundation AI companies that are creating these.

It's all so...confusing. And it raises a lot of questions about who gets to decide what values and principles are important, and how those values and principles will be enforced in the development of AI systems.

That's why I'm hosting this next General Seminar on the topic of constitutional AI.

For those who have never been to a General Seminar the approach is simple: we go into possible futures to find the material cultural artifacts that indicate some corner of some aspect of this world. I refer to them as like archeological fragments — pieces of material culture that one might find in this world.

In General Seminar we create speculative prototypes — artifacts that are designed to be provocative and to spark conversation. They are not predictions, nor are they prognostications. They are artifacts that are meant to be used as a starting point for discussion about policy and governance and artificial intelligence.

Which I think is the point. We make worlds through the stories we tell, the artifacts we create, and the conversations we have. So let's have a conversation about constitutional AI.