Desire and capability to build Dyson spheres isn’t necessary for dangerous misalignment. A sufficiently advanced LLM-based AI could kill humans by accident while not being able to (or interested) in doing anything much after that. A more goal directed AI could kill humanity and then have other goals besides building Dyson spheres.
There’s also the possibility of killing ourselves before building ASI (possibly with AI assistance).
Desire and capability to build Dyson spheres isn’t necessary for dangerous misalignment. A sufficiently advanced LLM-based AI could kill humans by accident while not being able to (or interested) in doing anything much after that. A more goal directed AI could kill humanity and then have other goals besides building Dyson spheres.
There’s also the possibility of killing ourselves before building ASI (possibly with AI assistance).