There is some excitement about the likes of ChatGPT and its ability to write or analyse code and for good reason: An AI writing code as autonomously would have been hard for many of us to imagine even a year ago.

But there is a fundamental limitation: A big software project is full of thousands of little decisions and interactions, typically made by a human. Those decisions might be made based on anticipation of the behaviour of the user. They might have a social context. They might need to be in line with the vision of the company. Some decisions might lead to issues if they contradict other design aspects of the system.

Fully specifying literally everything would be near impossible and really, it is all these little decisions that are the hard part of making software, not the act of transforming them into executable code.

If natural language becomes the programming language that’s turned into code by ChatGPT, then we are not tackling what actually makes software difficult. What it would take is for the AI to be able to write the specification itself.

Want to scale up your IoT solution?

Get in touch today to set up a preliminary discussion with our IoT wireless mesh experts to improve your position for the future.