Session 1
Should we adopt
an AIC-DC licence?
Questions for the strategic workshop
DownloadPowerPointPDFOpenDocumentFontsInstall the fonts before opening the PowerPoint or OpenDocument file
Why this session
Do we want what the framework does?
- AI was made with the combined knowledge of all of us.
- So the benefits of putting AI to use should return to all.
- Everything the framework does follows from this stance.
- Without this stance, the framework is hard to make sense of.
What a licence establishes
We answer to the people we commit to.
- The Standard sets what each commitment requires. We choose which to declare; we do not write our own.
- Each commitment is a promise to particular people: those our deployment draws on or affects.
- They can see whether we keep it, and they lose if we do not, so the licence gives them the right to enforce it.
- No central body approves us.
The questions
The questions that follow will help us see whether the framework fits our organization.
Question 1 of 16
What do we make of the view that AI was made with the combined knowledge of humanity, and that the work it powers should therefore give back to the commons? Where do we agree with it, and where do we hesitate?
Question 2 of 16
How would it change us to turn part of what we do with AI into a commons, no longer ours alone? What would we gain by it, and what would we find hard to give up?
Question 3 of 16
What do we hope establishing such a commons in our organization would change, and for whom?
Question 4 of 16
Why are we considering this now? Would those reasons survive a difficult year?
Two terms before we continue with the questions
A deployment and its deliverable
- A deployment is a working arrangement of processes that puts AI capabilities to work to produce one deliverable.
- The deliverable is what we declare the deployment to produce: a product, a service, an offer, or an AI capability itself.
- The deliverable draws the boundary of the deployment.
- We may run many deployments.
Question 5 of 16
Based on the framework's definition of a deployment, which deployments do we run in our organization, and which of them could we imagine opening to others, so that they are no longer ours alone? Which of them would we start with?
Question 6 of 16Trying on: Reciprocity (AIC-DC-R)
If this deployment already brings in revenue*, then looking at that revenue and the costs of running the deployment, could we arrange for at least 5 per cent of that revenue to flow continuously to everyone whose completed work went into it, once they are no longer paid for it, with anyone free to come forward and claim recognition, whether or not we knew of their work? Would we see that arrangement working here?
* This is a condition for profile AIC-DC-R.
Question 7 of 16Trying on: Sustainability (AIC-DC-S)
How far are the environmental impacts of this deployment measured and accounted for today? What would it take to go from there to restoring, in the place where it occurred, what each of them takes from or does to the environment, on every planetary boundary and without buying credits, until the balance stands at least at neutral? Could this deployment be run so that it holds that balance itself, and could it go beyond it, to regenerative?
Question 8 of 16Trying on: Openness (AIC-DC-O)
How much of our own machinery that puts AI to work in this deployment - the code we write and any weights we train or change, the sources of the data the deployment draws on, and how we specify, evaluate, and record the failures of the AI we use, whether built by us or supplied - is documented today? Could we open all of it to anyone, withholding only what the law requires?
Question 9 of 16Trying on: Governance (AIC-DC-G)
Who does this deployment materially affect, in what it draws from them or sends into their world? What would it take to identify those people and seat their representatives, on an equal footing, in a body made up of them alone with the power to decide whether this deployment continues? That body would also have to approve any weakening of the commitments we have declared for it, and could of its own motion bind us to stronger or further ones. Could we govern this deployment that way?
Question 10 of 16Trying on: Access (AIC-DC-A)
Since AI was made with the combined knowledge of humanity, would it be feasible, given what it takes to produce, to make the deliverable of this deployment available to everyone free of charge?
Question 11 of 16Trying on: Value (AIC-DC-V)
Where this deployment earns revenue*, would we be willing to direct at least 5 per cent of it into an unconditional basic income for everyone living in a region of our choosing where extreme poverty is widespread, building it into what the deliverable costs to produce, as an obligation rather than a gift? Would we see in it a fitting return for the combined knowledge of humanity?
* This is a condition for profile AIC-DC-V.
Question 12 of 16
Looking back over the six profiles we have tried on (questions 6 to 11), which of them, even if only one, would best fit our organization, what is feasible for it, and the way we see the world?
Question 13 of 16
How would we live with a commitment that outlasts a change of management, of strategy, or of owners, since it can be weakened only on notice, never quietly withdrawn, and, where a governance body holds that power (question 9), only with its consent?
Question 14 of 16
Who among us would need to stand behind a decision to adopt the licence, and which of them is not yet in the room?
Question 15 of 16
Whom outside our organization, among those our work with AI touches, should we hear before we decide, and how would we reach them?
Question 16 of 16
What would count, in a year's time and in five years' time, as the licence working as we intended, and what would tell us we had misjudged it?
Reading our answers
Adoption fits if we
- have at least one deployment, and at least one profile for it (questions 6 to 11), that we are willing and able to take on in full
- accept that the people we commit to can enforce it
- will declare it in public and keep it through changes of management, strategy, or owners
- have, or can win, everyone who must stand behind it
It does not fit if we want
- an assurance of responsible conduct
- an assurance that our AI is safe
- a third-party seal of approval
- a commitment we could quietly withdraw later
- any end other than a dependable commons within our operation
decks.aic-dc.net/session-1