GPT-4 is a new disruptive technology. For the first time, machines start to understand the nuance and tone of our asks. Through the chat bot, we can carry out interesting conversations. Machines can take and succeed at our standardized tests (SAT, AP, GRE, GMAT). Through the use of plugins, the bot can find precise and accurate information, much like humans can.
This new technology promises to be disruptive to many existing and established processes. The technology learns from existing artifacts on the internet, and then is able to apply that knowledge to other asks.
Building software products remains hard. Even though a simple function in a program can be changed easily and at will, it becomes harder to coordinate large changes in potentially many programs together and make it work well at scale.
The way we have set up our computing environment closely mirrors our own natural world. We set up small programs that respond to input and output information, then we stitch these programs together across many machines. The stitching introduces complexity that remains hard. Despite focus on design and architecture, software engineering remains an art and not a science.
In the world of GPT-4 and beyond, software architects become more productive, as current bots can make simple changes to a single program, and architects can stitch together multiple programs. For example, building user experience screens or writing database queries is repetitive and mostly straightforward. In a way, this could be a way to help standardize the software engineering discipline.