AI for Mass-Scale Code Refactoring and Analysis

Chapter 3. AI Large Language Models: Beyond the Chatbot

When tackling a problem you wish to solve with AI, one of the first questions you should ask yourself is what kind of AI model you should use to solve the problem. An AI model can act as a trained, interactive knowledge base to add to your toolbox. However, that’s like saying a dog is simply a four-legged animal that barks. After all, you’re likely considering which breed works best with the family. Is the dog trained, and trained on things that matter to you? And what is the right sized dog for your environment?

While a chatbot can be useful for generating code, LLMs can help beyond a mere chatbot experience. There’s a wide selection of AI models available today, both private and open source, that are great for coding use cases. The types of models you’ll want to use are highly dependent on what your organization is trying to achieve. But you’ll also want to consider the training and performance of the model for your use case.

In this chapter, we’ll offer you general guidelines on model selection and some of our specific thinking about models regarding implementations for mass-scale automated code refactoring.

AI Models Pretrained and Ready for Adoption

Because of the heavy lift of AI model training, most of us will be “renting” pretrained AI models, which we’ll tune with other AI tools and techniques, such as RAG, to refine results.

AI models learn by looking at previous examples, scoring their output, and updating their ...