Stark wrote:Totally unrealistic expectations are unrealistic. Increase map size, quality and interactivity while retaining the entire world as persistence?
A lot of the stuff he asked for is just hardware limitations. Memory is still fairly tight on current consoles, 512 Mb isn't that much when you've used most of it up with the frame buffers and streaming HD models and textures. With double-digit gigabytes of memory, the cost of persisting even an entire city (to GTA4 resolution) becomes a trivial fraction of your available memory, so it becomes a design decision for how long damage etc should persist. Similarly, dense traffic and crowds is essentially a model count issue, the next console generation should have no problem rendering them (particularly since resolution will stagnate at 1080p for a decade or so).
Dwarf Fortress relies on emergent behaviour and that's plenty nice, but this is not going to magically create a story for you.
Automatic generation of stories was a major research topic in symbolic AI in the 1980s;
TALESPIN is the prototypical example that kicked it off. A significant amount of work has been done with some surprising hidden successes, and a lot of that could be updated with modern techniques (e.g. Bayesian nets, millions of times more brute force available for search/constraint satisfaction) to produce at least passable narratives. There's virtually no interest though, because (a) you need creative people to make games and they want their 'vision' realised, even if it is shitty fanfic grade writing and (b) no one in the industry has the cash for such blue-sky stuff. Graphics research has been pushed by many contributors; the CAD community, movie CGI, compsci academia, hardware researchers, scientific visualisation etc. Game physics similarly benefits from masses of research on physical simulation for engineering and scientific purposes. Game AI just gets a few scraps from academic AI, and automatic narrative generation hasn't really been trendy for twenty years or so.
PeZook wrote:Wooo! Anyway, my point is that you don't need super-duper AI technology to avoid stupid crap like psychic people: just write it better.
Game writing isn't magically going to improve any more than movie writing is, so this is a non-solution. Once a form of media reaches mass popularity and enough budget to have dedicated writers, you run into the basic limitations of the available talent base. Game writers already get plenty of attention and have pretty much all the tools they could ask for (for telling static canned narratives). You get the odd title with really good writing or neat gameplay innovations of course, and rising budgets can add either polish or length (usually polish, if you're lucky, marketing if you aren't) but fundamental improvements in gameplay experience require technological progress.
(OMG I WANT TO ENTER EVERY BUILDING!!!)?
I don't see why that's unrealistic. It just requires more storage space and some (probably offline) procedural content generation. Suspension of disbelief hinges on different things for different people, but for me certainly in GTA it's annoying that you can't blow holes in buildings (and reveal interiors) the way you can in the Battlefield games.
Vendetta wrote:Having a witness management and communication system that would handle events like that propagating throughout the world would be shiny, but probably not sufficiently shiny to pay for it's development cost.
Not just the dev time, also the significant memory consumption and worst major increase in tester load / debugging time. As I said before, if you want this done well, it would have to be done in reusable middleware. If you imagine how the software in Star Trek holodecks must work, their 'games' seem to based on a completely common, extremely advanced engine that does everything from ray-tracing the graphics, through physics and content generation to near-human-level AI. The level of skill required to create a new game is that of an amateur movie director. That is the end point that we are slowly converging towards.
