British self-driving car startup Wayve has developed an AI model that can be trained on one type of vehicle and then transferred to another without any adaptation. Last year the company demonstrated its AI model that trained in London could then be used successfully in other UK cities without any retraining or further knowledge acquisition.

According to Wayve, its latest AI development is the first time that an automotive AI has been able to successfully transfer from one type of vehicle, a passenger car, to another, a delivery van, without any retraining.

The company has developed a generalised AI model that appears able to scale very significantly without need for re-engineering.

Wayve is part of a new generation of self-driving startups, sometimes described as AV 2.0, that eschew the multi-modular sensor configurations of the first generation in favour of a single focus on very deep neural network approaches.