I think the idea is less "additional sensors don't add value in various scenarios" and more "we are 100% a vision-only system can achieve self-driving, we are not sure about sensor fusion systems".