This is pretty much why Brad doesn't refer to the strategic scripts as AI. He refers to them as 'scripts', and the term your looking for is not 'Artificial Dummy' but rather 'automation'. The strategic AI is roughly as complex as the logon scripts I wrote for the computers at my day job; the thing I use to map printers and network drives. If anyone referred to that as AI, I'd think they were thoroughly odd. Scripts are pretty much just a list of things for a computer to do one after the other. I write dozens of them a day, and I would not call myself an AI developer; I'm a systems engineer. In games development, scripting is a low-skill entry-level job (like QA); many of the larger companies offload it on interns and new hires who are still being trained up. Some places I've worked even have running jokes about scripters being outranked by janitorial staff.
Players, on the other hand, do refer to it as AI, but largely because they think of the computer players as each being one entity that is called 'AI'. This is fundamentally incorrect; what they're actually playing against is a series of often completely unrelated subroutines, some of which are AI (the way the computer players decide what to do with their ships, for example), and some of which are just basic scripts, occasionally with some entropy thrown in to make them slightly less predictable (for example, the way the AI chooses which ships to build is largely just a random weighted array defined in a script). Therefore, even those of us who should know better do refer to script improvements as 'AI', simply so that others know what we're talking about.
Gal Civ 3's AI does respond to human interference. It makes decisions based on what is happening around it. IT checks what it has to hand, and then attempts to choose the best course of action. And it does so very well, provided it is given the resources to do so. Brad is quite proud of his AI, and for good reasons - it's very good for a commercial product. It's not that special compared to the kind of thing they run on Deep Blue, but he's not really in the business of writing LISP for supercomputers. He is entirely aware that the GC3 AI will never be up to that standard, and has been entirely honest about that since GC3 was announced. The scripts, being scripts, will never actually do that. However, if there's enough different triggers defined in the scripting language, then you can generate many, many, many different ones that can be triggered based on different circumstances.
Many of the problems on this forum come down to this basic misunderstanding - when we tell Brad his AI sucks, he thinks we're referring to the actual AI, takes a look at it, and sees that it's fairly robust. But we're usually referring to the scripted aspects, which are horrible.
