Drawing from my personal coding escapades, I'd argue: we're not there.
I've put both GPT-3.5 and GPT-4 through their paces across diverse coding scenarios. In my fledgling days of programming, these AI models were akin to a mentor by my side. They played a pivotal role—whether it was breathing life into my maiden Python terminal game, deciphering the quirks of my JavaScript, or streamlining the architecture of my HTML and CSS projects. Their insights even proved invaluable as I dabbled in the world of Java app creation.
However, as my proficiency grew and my linguistic arsenal expanded, my reliance on ChatGPT waned. What once was an invaluable coding companion has transitioned to more of a rudimentary aide, analogous to an intern tasked with mundane chores.
The reason? To put it modestly, ChatGPT's proficiency plateaued below my evolved capabilities. Its prowess in code generation and editing is commendable up to a point—perfect for laying a foundational framework. But as one delves into intricate functionalities, especially when juggling multiple languages, ChatGPT's assistance becomes increasingly limited. Moreover, without a solid grounding in these languages, relying on AI to expand upon a codebase becomes challenging, especially if you're uncertain about necessary modifications to meet specific objectives.
While I don't don the hat of a software engineer, one thing is abundantly clear: the technology hasn't reached a level where it can supersede human expertise. And even if it edges closer, remember, it's the software engineers who'll be at the helm, architecting and elucidating these advancements.
A strong tool, bereft of a user's comprehension of its intricacies and potential enhancements, loses much of its utility, doesn't it?
In essence, innovations like ChatGPT underscore, rather than diminish, the indispensable role of software engineers.