I didn’t attend this seminar, but if I had I would have asked why the proposed system uses kiosks rather than service to arbitrary addresses within the safely drivable region. Restricting the routes to specific streets on which low speeds are acceptable seems like a sensible starting point for such a service, but I don’t see any reason for restricting to a limited number of specific stopping points.
A separate issue raised in the opening remarks is that of identifying different levels of AIDriving. My own preference would be to have just two acceptable levels: namely driver monitoring systems that take over in cases of apparent operator incompetence such as when an imminent collision or loss of traction is detected, and full driverless operation (at maybe a restricted speed on a limited selection of routes). The intermediate levels, I think, all encourage inattention of the “responsible” operator which is just asking for accidents to happen. (This does not preclude the traditional “cruise control” so long as it is just to relieve the operator of the physical requirement of having to exert constant pressure on the accelerator to maintain speed, but should perhaps be subject to switching itself off and reducing speed whenever a loss of operator attention is detected.)