Last week’s General Seminar brought back a big haul of artifacts we found when we piled into the janky time machine and went to a near future in which AI had a constitution and was interacting with society in complex and unexpected ways.
We explored the implications of an AI constitution on society, governance, and individual rights, considering both the potential benefits and the challenges that such a future might entail.
Rather than just bringing back a report or musings, we found stuff — the material cultural detritus representing the ways doctrine, policy, governance, and societal norms might evolve in response to an AI constitution.
We didn’t get caught up in hypothetical debates or abstract theories; instead, we focused on the tangible artifacts that reveal how an AI constitution might manifest in everyday life.
These are meant to speak for themselves in a fashion, just as if we had done a kind of reverse archeology — excavating things from the near future we found: street signs, corner store items like lottery tickets, receipts from Erewhon. Each item has imbued within it the implications of some kind of embedded logic based on the rules, norms, and structures that AI “constitution(s)” might impose on society.
One of the main themes in the 90+ minute session was this idea that there may be multitudes: not one embedded set of values/norms/structures that an AI constitution would impose, but rather a variety of possible configurations that could coexist and interact in complex ways across different contexts, regions, locales, zones.
It was an awesome session. Thanks to everyone who brought their curiosity and creativity to the session!
If these kinds of artifact-led speculative creative prototyping workshops look like something you’d like to participate in, stay tuned for future sessions or get in touch to arrange a private session for your team or organization.
A receipt here indicating there is a fee placed on purchcases for ‘alignment’ and what seems like a fund to support universal access to AI compute, perhaps implying that such access is considered a fundamental right or essential resource in this speculative future.
Good one from Marco — there was dicussion about the regulations, rights, and responsibilities associated with having an earned privilege to use certain AIs. For example, could one lose one’s privilege? We are seeing people cut off from models as they are (rightly or wrongly or arbitrarily) suspended from using them for some violation or another.
When I’ve heard tell of this (mostly on Reddit by the supposed victims) the pain expressed is profound — as if a writer was told, for example, that they have been banned from using words to write.
Suppose the variability of models used for autonomous transport and associated rules and regulations leads to a complex landscape of compliance and enforcement where you are banned from an LLM and thus from ancillary services in which it is used, like in vehicles. An occupants license implies that you have been deemed eligible to occupy an autonomous vehicle and utilize its capabilities in accordance with the rules and regulations set forth by the governing authorities. What happens when you get suspended?
Suppose we buy these legal agentic units with a SKILL.md to help represent us for little legal skirmishes we may have — arguing a parking ticket or perhaps a more complicated malfeasence..or felonious misdeed? I mentioned the Ari Birnbaum AI character who appears in TBD Catalog — he/they/it are a legal agentic entity designed to navigate the complexities of legal systems on our behalf (before the current agent craze, I’ll say — that was back in 2012..who said Design Fiction doesn’t work?).
And now suppose if we use these they have to be properly “licensed” and authorized and in compliance with whatever doctrinal requirements a specific locality/government/agency has running on its end? Just like you can’t hire a lawyer to represent you or do work for you who isn’t properly licensed and authorized, such may be the case with this off-the-rack legal agentic contrivances that are somewhere between a burner phone and a lottery ticket.🤷🏽♂️
There was this curious scenario kicked-off by Angela — a short discussion of those little food delivery robots with wheels? Suppose they got intelligent and agentic and started their own business, leveraging their autonomy and capabilities to operate independently in the marketplace. Buying frozen tamales from a wholesaler or negotiating with a local family restaurant when the robot’s AI determine tamales are a hot item, it’s going to be a beautiful weekend, and they see a high probability of making a decent profit by rolling to the city park and selling them directly to consumers.
The whole thing is an agent navigating the complexities of supply, demand, and competition in a way that mirrors human entrepreneurial activity..and that’s weird but plausible and possible even if it is a bit aggrevating. So..then what?
And in some areas there’d be the opposite policy enforcement. Compromise?
I really liked where the policy for agentic activities doctrine in some kind of policy wonkish bureacratic form (“Los Angeles County Agentic Doctrine Policy, Food Service, Section 2, Paragraph (d): Required licensing and compliance for all agentic entities must install the following immutable prompts in their SKILL.md file…” etcetera) led to as the specifics went away from existential things to the mundane everyday things we don’t really think about until we have to think about it.
This one particularly as there has been a warm-ish debate here in Venice Beach about these autonomous delivery robots.
Watching a local taco truck guy fill one with an order, close the lid, and give it a pat on its head before it trundled off made me wonder about some strange logical conclusion in the context of all the (hyperbolic, half-baked, etc., etc.) announcements of fully agentic $1bn business run from a single terminal window.
Suppose some person-less agentic entity determines it can arbitrage food trucks by identifying price differences, demand spikes, or under-served locations and then dynamically deciding on a menu, paying some out-of-work humans to do a bit of food prep, and then agentically routing itself to maximize profit opportunities across the city.
Hey. Hollywood? You still there? There’s a B story of some story or show there in that vein that comes to mind, highlighting the potential and plausible absurdity of such a scenario bubbaleh let’s take a meeting have your guy call my guy..
The broader implication across all of them is the same: constitutions for AI would not appear to most people as philosophical documents. They would appear as fees, labels, permissions, restrictions, warranties, certifications, and exclusions. The artifacts make visible that governance is never only in the model. It shows up in the receipt, the sticker, the sign, the form, the ticket.
See Also