AI will have been the primary access point to history, the means by which art is seen and understood by its audience; it is the avenue by which art retains an educational dimension.
If these are in line with an artist's concerns, then engagement with AI is probably a necessity. This is definitely not to say you have to make any money, but you do have to contend with the mechanisms by which art becomes part of the art world.
So this is project of data normalization— for how AI should deal with data and uncertainty.
The downside to Wikipedia is collaboration, unfortunately. Specifically, editors must marginalize all the variables in the model to a filter log. Thus, the main computational task in algorithmic governance is gatekeeping, whereas in traditional AI it is the absence of a listening device.
As such, a focus on innovation per se in computing and development may be misplaced to some extent. Cultural products are essentially general-purpose optimization machines, built to handle a specific task domain, when they are technically computers.