ChatGPT writes code, but won't replace developers

ChatGPT, an AI-powered chatbot that can simulate human conversation and write code, will metamorphose application development -- and the developer profession -- into a different animal, according to industry experts.

More than 1 million users have signed up for ChatGPT's free research preview since OpenAI released the chatbot on Nov. 30. The bot can write simple webpages and applications in programming languages such as JavaScript, Python and React. It can also find bugs in code and help create new programming languages.

The model interacts in a conversational way and can admit its mistakes, answer follow-up questions, challenge incorrect premises and reject inappropriate requests, according to OpenAI.

While it cannot yet write complex code, such as what's required for banking applications, ChatGPT will become a proficient coder within the next decade, said Rob Zazueta, a freelance technical consultant in Concord, Calif.

The last iteration of GPT was surprisingly good, but this seems to leave it in the dust. It doesn't take much imagination at this point to see where this is all going.

"I have a weird mix of existential dread about it ... but also a ton of excitement," he said. "The last iteration of GPT was surprisingly good, but this seems to leave it in the dust. It doesn't take much imagination at this point to see where this is all going."

Charlotte Dunlap, an analyst at British market research firm GlobalData, echoed Zazueta's prediction, but believes the effects will happen much sooner.

"Consider how we've gone almost overnight from GitHub's Copilot autocomplete prompting method of AI-injected coding to a dialog format via ChatGPT," she said. "Advancements are happening such that we'll get there in two to three years."

But no one knows what form those advancements will take, said Abhishek Gupta, founder and principal researcher at the Montreal AI Ethics Institute. Twelve months ago, no one could have accurately predicted the ubiquity of generative AI systems and the many forms that it is taking today, he said. Similarly, it's impossible to predict what will happen in three to 10 years.

"Nobody has a crystal ball," Gupta said. "Trying to make a forecast that's more than 18 months ahead in time -- you might as well just flip a coin."

Software engineering jobs of the future

One thing that is in the cards: Developers will no longer have to write boilerplate code, Gupta said. Instead, they can focus on areas such as complex application architecture or cybersecurity.

Zazueta agreed, saying ChatGPT already does a good job at writing working code.

"I can take that, modify it to fit my needs and cut through boilerplate stuff quickly, allowing me to focus on the more intensive kind of work the AI is not yet ready to handle," he said.

While ChatGPT might replace some aspects of coding, such as writing generic functions or boilerplate code, it won't supplant programmers altogether, Zazueta said. That's because a programmer's job requires more than coding.

"It takes skill to be a programmer -- to be able to structure a program, follow the logic and generate something greater than the sum of its parts," Zazueta said.

Still, ChatGPT could pave the way for new job titles. Prompt engineering, for example, will become an in-demand skill set in the AI era, he said.

Prompt engineers understand the principles and techniques for writing model inputs to get the best possible results from chatbots. As such, it should be called dark artistry, Gupta said.

"You have to have the right incantation in place that triggers what you want from it," he said.

The rise of AI coders such as ChatGPT will also lead to an increased demand for software developers versed in data science principles, GlobalData's Dunlap said. For example, engineers who can design, build and test applications using data science platforms and languages such as Go and Python.