[oneplane]

It worked out in Paris for quite a while, granted it was smaller than pallet workloads but still, a large physical object transportation network (akin to pneumatic tubes) was quite useful. IIRC the downfall was flooding combined with degraded infrastructure due to lack of maintenance, but those are procedural and planning issues, not foundational issues that cannot be overcome.

Similar smaller-than-container transportation networks exist elsewhere, even temporary ones used for evacuating tunnels while they are being bored. Even coal strip-mining type of situations have extremely long length conveyor systems that work pretty well.

If you forget the nonsense AI picture for a moment, what they are doing is essentially not much more than extra-specialised rail, which when built for a specific purpose makes a lot of sense, especially when you want to factor in autonomy, or per-object routing instead of single-train based routing. It's already done in factories too.

[bluGill]

I'm not arguing you can make something like this work. However it will cost a lot of $$$ and in the end you won't have anything better than known freight trains. Even if you convince every [geographical area the size of Japan] to install one, the scale factors still won't make this better or cheaper than a train.

[oneplane]

But that's the thing, it has already been done, already been proven and it does work. Roads and classic freight rail are not the only thing that are used right now that do this.