As discussed with @Halfak.
The Scoring Platform Team's products consist of two primary audiences: downstream developers who use our services and APIs (including tool developers and Foundation teams), and the users of those downstream products.
In theory, as the upstream developers we coordinate with downstream developers who are then responsible for coordinating with _their_ downstream users. However, in practice:
- The end consumers are often interested in the inner workings of the system
- Downstream developers may not have the interest or knowledge in explaining how the underlying mechanisms work, preferring to treat Scoring Platform products as magic black boxes.
- We want to keep many groups of people in the loop, including aforementioned downstream devs and end-users but also researchers and others working in the AI and ML spaces.
As part of a broader objective of increasing/improving reuse of our products – an objective that includes things like improved documentation – there should be a nexus of activity that serves as the central organizing point for our activities. "Wikimedia AI" comes from the IRC channel, is understandable as a name, and is not overly specific like "ORES" (which is a specific product).
By the end of Q1 2018–19 we should be able to have a clear answer to the question "where should people go if they are interested in AI-related work on the Wikimedia projects?"