On the Road with GPT-4V(ision): Early Explorations of Visual-Language Model on Autonomous Driving