Interestingly to me, generative AI is often used to get results that commit the opposite error compared with these statues: they are, essentially, too confident in their choice of details. For any random topic, the average member of the public is likely to believe the AI's results are more accurate than can be backed up by the evidence.

Generative AI exists, but it is very much dependent on the data it has been trained on. Not saying it would not be interesting, but a huge caveat is required.

> If only there were some system that could start from some sparse and noisy observations and weave together a plausible completion...

Humans?

I would much rather see human artist interpretations after they were briefed by the archeological experts on the evidence.