Published on the 27/06/2023 | Written by Heather Wright

But curiosity starting to outweigh scepticism…

There’s a healthy scepticism amongst developers when it comes to artificial intelligence – despite 70 percent of developers already using, or planning to use, the AI tools in their development work.

Stack Overflow’s 2023 Developer Survey shows there’s still plenty of work needed to win developers over and gain their trust, with 27 percent saying they ‘somewhat’ or ‘highly’ distrust the accuracy of the AI output from tools. Thirty-one percent are on the fence on the issues of trust.

Despite those concerns, shows 44 percent of developers are already using artificial intelligence (AI) in their development; 25 percent plan to do so soon. Even more of those currently learning to code are already using AI tools (55 percent).

“These new tools are not an easy button and the axiom ‘trust but verify’ is the best approach.”

Of those using it, 83 percent say they are currently using AI tools for writing code, with 49 percent saying they’re using it for debugging and getting help. Documenting code, learning about a codebase and testing code were also tasks AI is currently being deployed for, according to the report, with blockchain developers, data scientists and front-end or fullstack developers most likely to be using the AI tools already. Embedded application developers, desktop/enterprise developers and hardware engineers were less likely to be using them, and less inclined to use them in future.

The annual report is based on a survey of more than 89,000 software developers, both professional and learners, from 185 countries.

Despite 21 AI tools being included in the survey, the vast majority were just using ChatGPT (83 percent) and GitHub Copilot (56 percent).

Erin Yepis, author of the report, says professional developers may need time to adjust existing workflows and will most likely be motivated by their junior colleagues who are using AI tools while learning to code.

“It’s early days in the hype cycle for these newer AI technologies. We expect that a little more time may need to pass before we see developers using more AI tools more broadly,” Yepis says.

The trust factor is likely to be contributing to the slow adoption of AI tools, she says. Of those using or planning to use the tools, just three percent highly trust the accuracy they provide. Twice that number highly distrust the tools.

“When it comes to important matters like school or work, these new tools are not an easy button and the axiom ‘trust but verify’ is most likely the best approach to integrating new tools in the development process,” Yepis adds.

Another potential barrier is the ‘complexity cliff’, Yepis notes.

“Much like Helmsman’s complexity cliff, a concept found in project management, after a certain point, the ability for AI to handle all the nuances and interdependencies of a solution drops off.

“That’s when humans, their adaptability to apply judgement and have original thought, saves the day.”

Interestingly, those who are not interested in using AI tools say AI assistance in writing code would be the least beneficial use of AI – they’re more interested in it for collaborating with team mates, deployment and monitoring, and committing and reviewing code.

“This disconnect most likely is with the fundamental difference of type of developers not interested in using these tools with those that are interested and have more applicable use cases for the current functionality available,” the report notes.

Rgardless, both professional developers and those still learning believe their development workflow will be different in a year, thanks to AI tools, with the impact on documenting code expected to be greatest.

So what benefits are developers looking for with the tools? Unsurprisingly, increased productivity comes out on top, with speeding up learning and greater efficiency tied for secondary benefits.

A recent GitHub developer report found 70 percent of US developers surveyed believed AI coding tools would provide better code quality, completion time and resolving incidents. Seventy-percent of the developers believe AI tools’ top benefit is in upskilling, which can be directly integrated into a developers workflow with AI coding tools, followed by productivity gains.

But they’re also eyeing AI coding tools as a means of making teams more collaborative, something they’re keen to see more of – and to see as a top metric in performance reviews. Collaboration and communication is only used as a performance metric in 33 percent of the companies surveyed.

Indeed, the GitHub survey of 500 developers, 92 percent of whom said they were using AI coding tools both in and outside of work, suggests AI will require an evolution of performance metrics for developers.

“The way developers are currently evaluated doesn’t align with how they think their performance should be measured,” Inbal Shani, GitHub chief product officer, says.

“The developers we surveyed say they’re currently measured by the number of incidents they resolve. But developers believe that how they handle those bugs and issues is more important to performance. This aligns with the belief that code quality over code quantity should remain a top performance metric,” Shani says.

Notably, developers believe AI coding tools will also give them more time to focus on solution design, something that has direct organisational benefits and means developers believe they’ll spend more time designing new features and products with AI instead of writing boilerplate code.