Well .. I want to argue and offer the following two points:

1 - proposed solution uses optical nerve that already in place, it is a legit way for BCI when talking about vision (as in eyes)... invasive way would still end up replacing optic nerve and in that sense would be anyway more efficient.

2 - this idea of invasive is very silo / extrapolation of today's tech, what if in 20 years what we would end up actually using would be a bunch of nano machines that work in the brain and then send / receive signals as an interface, it would be completely different from Neurolink approach as well, we should not get fixated on invasive vs non invasive we should fixate on the quality of solution.