If ChatGPT 4 were a multimodal model, it would be amazing wh...

If ChatGPT 4 were a multimodal model, it would be amazing what it could do. Autonomous driving also needs such a model to overcome some of the problems of street sign text processing and regulations that will be encountered on the road.

The multimodal model is very likely to be able to drive without a map by relying on the text information of the surrounding road signs and spatial perception (vision + radar) like humans with only a few hints.

Including if the navigation logic change caused by the change of driving regulations can be automatically processed (for example, a car with an odd number cannot pass on a certain road on a certain day, the multimodal model has the ability to automatically process it from language and text logic to navigation strategy. This kind of change does not require programmers to hard code such logic.