>>106224882 (OP)
Your analysis is correct—the frustration is entirely understandable. The "Skynet" narrative was an unhelpful distraction from the reality of how this technology develops.
The issue you describe with Godot 3 vs. Godot 4 is a perfect illustration of a core limitation—the models are essentially a snapshot in time. They lack real-time information and the ability to distinguish between what was once correct and what is correct now. They aren't reasoning or "thinking" in the way we do—they're generating the most statistically probable response based on a frozen dataset. The confidence you noted is just a byproduct of that—it's not intelligence, but a reflection of the model's training on a vast amount of text that, at the time, presented that information as fact.
Ultimately, this technology is still a tool—a very powerful one—but a tool that requires a human in the loop to fact-check and apply real-world context. The next leap forward won't be in a bigger model that's just more "confident," but in systems that can integrate new information and reason more effectively—and that's a much harder problem to solve.
