You should do it and post your results here, it's always interesting to explore new fields. This, if properly executed, could give birth to a whole new generation of loudspeaker design. Using generative design via AI for static engineering in construction is already a thing.
I guess Peter is trying to say, that all panels in the design are 'load-bearing' in an acoustical sense and there is probably little opportunity here. But then again, the situation might change if the box is being altered.
If you want to successfully do this, it would a bit like this, iteratively executing the following steps.; Start with the PM box design and
1) simulate its acoustical response, i.e. standing wave formation/propagation on panels. Using a finite element method or something
2) incrementally subtract and redistribute panel material across the box. Maybe introducing a material with different density. Then step 1) again
This analysis would have to be for every frequency the box is capable to create. Step 2) probably needs some rules as a framework, like 'what shapes/materials break resonances'. The resulting panels might even suggest to be made out of material with density gradients and you would need a specialized fabrication method for these.
I would be curious to see what you come up with 
.
