DEVOPSdigest

In certain developer circles right now, the sense of doom is palpable. For many, the writing seems to be on the wall. The question is not if but when AI will render their professions wholly irrelevant. Years of training, of picking up the skills and rising through the ranks — all chucked down the drain through the rise of large language models (LLMs).

It seems like a tragedy. Luckily, it's also a fantasy. It is undeniably true that AI is quite skilled at generating certain kinds of code and is getting better at it by the day. But its current capabilities don't even remotely match those of our most skilled developers, and — despite the doomsaying — will not reach this point for some time, if ever.

Recent developments in AI are a good thing for developers — and for the large software industry. It is no secret that the industry is currently dealing with a talent shortage of unprecedented proportions: for all the talk of AI replacing experienced developers, many companies are actively scrambling to find human ones — and failing.

AI can help plug these gaps. And — as it already has for years — it can make the lives of existing developers much easier, automating repetitive tasks and allowing them to focus on the more imaginative coding work that, at least for now, only humans can hope to produce. At a time when the need for new software has increased exponentially, these are very positive developments indeed.

Not Everything Can Be Automated

This distinction — between the kinds of brute-force work that can be automated and the kinds of imaginative work that can't — is a crucial one. What we're talking about here applies just as much to code as it does to, say, novels or paintings. Yes, an LLM can produce a functional replica of a Cormac McCarthy novel, a Salvador Dali painting, or an essential bit of code. But for the most part it currently lacks comprehensive intelligence that can innovate ideas from the top down.

Certainly, many would like it to. The idea of "developing development" is gaining significant traction in development circles right now — i.e., the notion of AI not just producing code but actually understanding concepts and assembling complete programs. The evidence on the ground suggests we're still some way from this goal and that the more comprehensive, abstract skills only developers can provide are still essential. And even if an AI could build an entire program from the ground up, we'd still have significant hurdles to clear.

Take intellectual property (IP). A given company's IP — its code—is at the very core of its business. But what happens when AI uses software code owned by another entity to create a "new" program?

Who exactly owns the resultant code?

The datasets used to train AI are, in fact, so vast and tangled that sourcing the origins of any AI-generated output may prove impossible.

The copyright implications are significant, and we are far from meaningfully sorting them out. However, a situation in which a software company can't claim to own its own code is clearly untenable. And that's before one considers the security risks presented by code built out of component parts whose origins remain a mystery to developers.

Plugging the Talent Gap — and Democratizing Development

The point here is that AI will not be building programs from the ground up. But it can help broaden the base of people who can work with code — which is very good for software companies.

Again, labor pressures in the industry are about as intense as ever. Key roles are going unfilled for months, harming innovation and holding back the economy. By simplifying the coding process, AI drastically broadens the base of people who can plausibly work in development. No-code/low-code drag-and-drop tools, in particular, present exciting possibilities here, helping to close the gap between developers and motivated laypeople.

There are risks here too, of course. An experienced developer has spent their entire career immersed in best practices; the same cannot be said for someone who started using no-code/low-code options last week. However, we are already seeing companies respond to this fact, with IT departments en masse once again centralizing software development, and we can expect to see many similar innovations on this front in years to come.

The future we're looking at is far less grim than many seem to imagine. Existing developers will work with new efficiency, with AI tools taking on the bulk of drudgework, in the process enabling more big-picture conceptualizing. Software companies need help finding experienced developers to be able to plug crucial gaps through AI tools. And laypeople with no coding experience will be able to contribute meaningfully to their employer's software initiatives. Far from the dystopia some are imagining, we might actually be verging on a golden age for software development — with AI creating the conditions for everyone to flourish.