Software development will fundamentally change in the next few years, according to Matt Garman, AWS’ new chief executive, who argued that the rapidly maturing AI programming ecosystem could mean “most developers are not coding” within two years.
Leaked audio obtained by Business Insider revealed Garman told AWS employees that coding was not the be-all and end-all of a software developer’s skill set, stating that “being a developer in 2025 may be different than what it was as a developer in 2020”.
Garman asserted that the ability to write code is not the central skill embodied by a good software developer, noting instead the core capability required by developers is being able to ‘innovate’.
“The skill in and of itself is like, how do I innovate? How do I build something that’s interesting for my end users?” Garman added.
Garman’s comments come amidst a period of flux for developers globally, with the emergence of generative AI in late 2022 prompting some industry stakeholders to warn the rise of AI developer tools could lead to workforce cuts in years ahead.
The AWS chief executive isn’t the first prominent business leader in the technology sector to make such claims either. Nvidia CEO Jensen Huang made similar predictions earlier this year, for example.
Speaking at the World Government Summit in Dubai, Huang stated that, in light of the advancements made in generative AI space, learning to code should no longer be the priority for aspiring developers.
“It is our job to create computing technology such that nobody has to program. And that the programming language is human, everybody in the world is now a programmer. This is the miracle of artificial intelligence,” Huang said.
A month later at Nvidia’s GTC 2024 event in San Jose, Huang clarified these comments, interpreted by many as portending the ‘death of coding’. Huan said that while many will no longer need to have a strong understanding of specific programming languages, there will still be roles for human programmers.
“[P]rogramming is not going to be essential for you to be a successful person…but if somebody wants to learn to do so (program), please do – because we’re hiring programmers,” he reassured attendees.
Sid Sijbrandij, CEO at GitLab, pushed back on the discourse surrounding AI’s role in software development, stating he feels discussions around whether AI will replace software developers is asking the wrong question.
In a LinkedIn post referencing GitLab being named as a leader in Gartner’s Magic Quadrant for AI coding tools, Sijbrandij said a more appropriate starting point for these discussions lies in how businesses can leverage AI to produce real value for their software development teams.
“How can businesses harness the power of AI across the software development lifecycle to accelerate innovation and drive tangible business impact for customers?”, Sijbrandij argued.
Speaking to ITPro, Peter Schneider, senior product director at software design firm Qt Group, said although their clients were reporting productivity gains in particular areas of the development lifecycle, coding tools still have some way to go before enterprises can start considering shrinking development teams.
“What we are hearing from cross-platform application developers is indeed that coding assistants can provide significant productivity gains in creating test cases and code documentation. Furthermore, we hear from professional developers that they love using general-purpose Large Language Models to give insight and examples of using new programming languages or their features,” he explained.
“However, we are also seeing some indications that general purpose coding assistants are struggling at implementing industry-specific code. This struggle leads to a lot of time being spent on reviewing code suggestions which overall does not improve the productivity for the use case of writing code.”
As a result, Schneider stated he doesn’t believe generative AI coding tools will replace human operatives any time soon, adding that even if their accuracy takes a significant leap forward there are still strong reasons for keeping humans tightly in the loop
“I don’t think any of these GenAI tools are likely to become substitutes for real programmers, unless the accuracy of coding answers supplied by models increases to within an acceptable margin of error (i.e 98-100%),” he argued.
“Even if GenAI does reach this margin of error, the four-eyes principle is still one of the most important mechanisms of internal risk control – in other words, any activity of significant risk like shipping software has to be reviewed and double-checked by a second, independent, and competent individual.”
Experts challenge Huang’s ‘everyone is a programmer now’ claim
Peter van der Putten, director of the AI Lab at Pegasystems and assistant professor at Leiden University, told ITPro that a good understanding of the major programming languages will be an important skill moving forward, despite the development of AI coding tools.
Van der Putten was not convinced by Huang’s argument that generative AI coding assistants mean developers will no longer need to be proficient in various languages.
He argued that without skilled coders, businesses will be left with significant amounts of technical debt generated by their automated coding tools, leaving them exposed to scores of potential vulnerabilities.
“[S]ome argue that coding assistants can unlock generation of software for non-coders such as domain experts. This will generally not work because these tools generate, well, code, so someone needs to be able to assess its validity. If you leave that to non-coders you will generate loads of technical debt, or worse unsafe, non performant or simply software that doesn’t work.”
