Microsoft Research Asia launches a large model for industrial scene, which uses GPT-4 to control air conditioning system.

CTOnews.com, August 11, Microsoft Research Asia recently proposed a method to use the GPT-4 model to control the air conditioning system (HVAC) in industrial scenes. It is said that "this method requires only a small number of sample samples and can be better than the traditional industrial control system in cost and efficiency."

Microsoft, a related paper released by ▲ Research Asia, said that for industrial scenarios, traditional industrial control software requires a lot of time and budget costs in dealing with heterogeneous tasks, inefficient samples and adapting to new scenarios, while using a large pre-trained model to control relevant industrial control equipment can ensure that the relevant deployment costs can be reduced with high accuracy.

It is reported that Microsoft used the GPT-4 model and created a virtual air conditioning system environment for experiments, developed a method that can use the basic model for industrial control without too much training, and finally got "positive results".

Microsoft also said that the goal of the study is to explore the potential of directly using large pre-training models for industrial control tasks, and to design ways to solve industrial control tasks with low "technical debt". So as to gradually replace the industrial environment with high fault tolerance.

At present, the relevant research results have been published on ArXiv, interested CTOnews.com friends can learn about it.