Indeed, that’s very true.
Nonetheless, it already indicates the direction in which the journey is headed. Code is frequently suboptimal or requires reworking to function properly, but it can still provide hints about where to focus your attention. In any case, tools like Tabnine, CoPilot, Rix etc often do already a fairly good job.
At present, the term “intelligence” is somewhat misplaced. If these tools were truly intelligent, they wouldn’t merely provide answers, but would also ask critical questions and offer proactive suggestions. For example, they might say, “Why are you solving the problem this way? Here’s an entirely different approach,” “You haven’t considered the following options,” or “I have security concerns regarding your intended architecture,” and so on.
Moreover, it would be more advantageous, ethical, and less hazardous if AI tools revealed the sources they used. This would allow novices to evaluate whether specific solutions might be effective, but possibly outdated, which is often the case, regrettably.
Lastly, AI should be able to recognize when manufacturers actively supply the system with data (which GP acknowledges doing in the panel) and, consequently, promote the formation of opinions that, if the provider is clever and cunning, could cause the purely statistical “objectivity” to become increasingly absurd.
I fear that after SEO, AIO will soon become the next “big thing.”