Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
Sure, ChatGPT and other tools like it can miscalculate math, the drawing of hands, and even love connections between human and bot, but they are increasingly helpful to developers as enterprises wonder how to put AI to work.
ChatGPT offers a time-saving first sketch for coders—one that can be placed into an editor, but still requires a human touch.
“One should never be copy-pasting something out of ChatGPT or somewhere else and just running that code in production; that’s just a recipe for disaster,” said Randy Lariar, practice director for big data and analytics at the consultancy Optiv. “But if you’re an experienced developer…you can certainly get a lot of acceleration.”
This is your copilot speaking. Mike Lempner, VP of engineering at the credit card and financial-services company Mission Lane, needed to move data from the project-management platform Jira to a new system.
A data-migration tool could do the job, but the developers turned to a cheaper option, ChatGPT, to get the code going.
They prompted the bot: Can you write python code that will extract tasks, subtasks, comments, and attachments from Jira.
“Within seconds, there were dozens of lines of code in Python that they were literally able to copy and paste and put it into an editor,” Lempner told IT Brew.
A GPT-free approach would have required research of each system’s API specs, and how to navigate between them—a task that could take hours. With some tweaks to the AI-assisted output—like adding in the company’s unique custom fields—the team got the data moving.
“It’s not the finished product. But it’s something that really kind of gets you to a good starting point,” said Lempner.
Git goin’. Steve Bennett, senior technical lead at Soliant Consulting, used the tool to implement a React-based signature pad—a feature that allows a digital drawing of one’s handwritten name.
While the output arrived in seconds, Bennett did have some troubleshooting to do himself, like updating libraries pulled from ChatGPT’s training data, which stops in 2021.
While some of Bennett’s crew is less than impressed with the ChatGPT’s coding capabilities, the team lead is nudging everyone to keep an open mind in their open code, given how the tool demonstrates API integrations and syntax options.
“I’m kind of encouraging people to say, ‘Hey, even if you feel like you know what you’re doing, just spend 10 seconds, type out what the integration is that you’re going to be doing for a new library, and see what it spits back to you.’ Because it’s going to save you the time of analyzing a framework or library,” Bennett said.
ChatGPT is a potentially valuable assistant, but it does not have the training of a veteran human copilot who’s seen a few things.
“In no way does it replace a human developer,” said Larier. “There’s a lot of judgment and experience that a human will understand with that code to make that actually useful.”—BH