The fanfare surrounding ChatGPT3 has attracted (stolen) my attention and my mind eventually wandered to think about how software provides value. When developing traditional applications, you start with a data model and design functions that will create, read, update, and delete the data. Validation and business rules were defined to act upon the data and produce results for users. There was a linear relationship between the value provided by software and the business rules encoded. Want the application to do more, then create more application logic.
AI and generative models have turned this approach on its head. Instead of starting with a small data set, you start with a very large data source for “training”. Instead of spending the bulk of your time developing business rules, in an AI world, you will likely spend your time designing pre-processing algorithms, selecting AI models for learning, and tweaking model parameters for better and better results. For developers that are not going to pivot away from traditional development, see this post from Brooks Hamilton about how AI can help you with traditional development.