Purdue University researchers discovered that large language model chatbots like ChatGPT can be used in autonomous vehicles to provide a more personalized experience for passengers, according to a Purdue News press release.

Using AI, a team in the Lyles School of Construction Engineering is working on an autonomous vehicle that allows for human-like, indirect prompts like, “I need to be at the diner as soon as possible,” instead of direct prompts such as, “Increase the speed of the vehicle,”  the press release says.

Researchers trained their chatbot with various direct and indirect prompts and defined a set of parameters, including traffic rules, road conditions and weather conditions, according to the press release. These parameters were uploaded to the cloud so that an experimental vehicle could easily access them.

The team is testing the capability of their trained chatbots in Columbus, Indiana on an old airstrip to collect data on how the vehicle responds to commands from a passenger regarding high speeds, two-way intersections and parking.