Brent Orrell writes at The Dispatch about a flaw in one common piece of career advice.

On the campaign trail, President Biden advised a crowd to “learn to program” in an attempt to address their concerns about job security in a changing economy. Suggesting they “learn to think” would have been better advice. New artificial intelligence is beginning to transform computer programming in the same way the assembly line revolutionized industrial production, marking down the value of some lower-level coding skills.

A recent article in The Verge described how advances in natural language processing technology by companies like OpenAI and its Codex program may soon reduce demand for coders, especially those at the lower end of the software development value chain. This is the kind of advancement that can take some of the shine off of a coding certificate or other short-term credential.

Codex is a software program trained on all of the functioning code in GitHub, the world’s largest open-source coding repository. Via Codex, written and oral commands in English can be converted quickly into functioning code. A programmer might type in, “Create a webpage with a menu on the side and title at the top,” and the AI, pretty much instantly, will pop out the code necessary to create the display. It’s a basic webpage that has to be refined, but the process is far more efficient than a human coder building the same platform from scratch. …

… The key to success in the technology workforce, and the job market more generally, is to move up the value chain away from routine work and toward more creative, people-intensive, and harder-to-automate tasks. As The Verge’s article points out, building software systems consists of two very distinct skill sets: coding and design—and AI is coming for the coding jobs.

The second skill set relates to system architecture: being able to look at a problem and be part of an imaginative process that envisions and creates end-to-end solutions. Original work like this is highly bespoke and difficult to automate. …