Instant travel is a game killing for me.
It is clearly an awfull lazy move from dev considering the billions of more believable options to speed up travel.
It isn't like the only two options are walk slowly like a turtle with crippled legs vs teleport everywhere anytime.
Instant for whom?
I agree that unless the PC has some teleportation skills or is stepping into a teleporter or something like that, travel should not be instant for the PC. However, travel being instant for the player is often desirable. (Or at least really fast for the player - i.e., a "proper RPG" is going to be simulating the game world for all of that travel time, so it's not going to be instant for the player but only as fast as their computer can go. The travel can be accomplished by having an AI temporarily take control of the PC with the simple goal of "get to destination", and the AI does so merely with the normal means available to the PC, and should return control of the PC back to the player early if something unexpected should happen during that travel.)
I am a firm believer that time should matter in the game world, so the fact that the PC is spending time traveling should also matter, and if the player makes the PC engage in a bunch of unnecessary traveling thereby wasting a bunch of game-world time, that may very well have significant consequences in the game world.
Creating an AI for this type of thing is several orders of magnitude easier than for DF.
No, it's not really that different.
- Techniques such as hierarchical decomposition can, when actually needed, be used to bring larger game worlds under control. It trades off some solution optimization for that massive reduction in complexity, but not only is it very much worth it, humans do the same thing (though generally not in as mechanically a fashion as simpler AIs).
- Most NPCs have zero need to analyze the whole world when making decisions. Do you consider the traffic patterns on the Po Nga road in the Tai Po market district of Hong Kong when you are trying to decide what to eat for breakfast? Of course not. That would be stupid. Think about what information you actually need to get through your day, and you will find it is some vanishingly small fraction of what exists in the world (much less what exists in the universe). Even if you are just considering your home town, 99.99% of what happens there is likely none of your concern.
- NPCs should not even have any way to analyze the whole world, because they should not have access to all of that knowledge. Information should not just teleport into every AI's brain anymore than IBM's unreleased quarterly numbers teleport into mine. Knowledge should exist only in the minds of the player and the AIs, and in explicit manifestations (books, the sound of two people talking, the actual item some knowledge is about, footprints, etc.). Transfer of such knowledge should only happen by means of in-game-world mechanics. So not only should AIs not consider the whole world when deciding what to do, they can't because they don't have such vast information - they only have the information they have acquired via their senses (whether mundane or magical) and via their own mental skills (e.g., deduction and synthesis). (And even most sensory information gets discarded fairly rapidly. There's some truth to the expression "in one ear and out the other".)
- There is no need for you to consider every possible action. Do you, even for the smallest of moments in your subconscious mind, consider dropping crushed sunflower seeds into the postal box each morning while considering what to do next? Again the answer is "of course not". If your AI is written such that you are going down a giant case statement of possible actions and considering "yes or no" for each action, then your AI is total crap. That's just about the worst possible state-space search strategy in the entire universe. Instead the possible actions considered should be guided/directed/limited such that they lead to a much more efficient state-space search strategy, which is why some concepts in our mind are related to others, and even when they don't pull those related concepts into our conscious minds (which they sometimes do), they still tend to "activate" them as "things to consider in current context". What this means is that having more and more possible actions does not in fact increase one's mental load to the same degree. It requires a bit more memory to store those possible actions and their mappings to other concepts, but it doesn't require a lot more CPU. (And it's not like DF is even using all available CPU right now - the game is still single-threaded.)
That's why you typically see a corresponding curve between game complexity and AI problems.
I get the distinct impression that your knowledge of AI is derived pretty much entirely from what you see in the games you play rather than a direct examination of what those in the AI field have shown to be possible. When game AI is regularly short-changed in terms of development effort (even while PR types regularly give a bunch of lip-service about how "advanced" the AI is going to be),
of course it's not going to be very good. Since (at least per my own personal "theory of fun", which doesn't apply to everyone) good AI is the source of the bulk of the entertainment games provide, it should be where much of a game's development resources are focused (rather than getting somewhere around 1/20th or 1/30th of the art budget). When you squeeze the AI development effort to such a degree, you end up spending a large percentage of that development time on simple nuts-and-bolts issues rather than making the AI actually good, and subsequently the result is even worse than a linear extrapolation would suggest. (Some of the issues that you see may be due to AI dev time being cut so short that even some of the nuts-and-bolts are missing.)
Are you kidding me? If someone develops a generic AI that can run DF, that will be the Singularity, won't it?
No, I'm not kidding. Generic AIs have already been created, in all sorts of flavors. Breadth-first state-space-search is a very simple example of a generic non-learning AI "engine". Some of them would likely be capable of playing DF so long as they were provided an accurate encoding of the game mechanics and a means to represent game state (and some sort of evaluation function), though the inefficiency of many of them would be completely unacceptable (and not practical in terms of memory usage). The most challenging parts of working on AI have nothing to do with making something that could work in theory, but optimizing representation, search and evaluation efficiency enough to make it work in practice (thus things like
alpha-beta pruning are invented which, while being a relatively crude way of optimizing search and reducing the need for evaluation, is still effective in simpler cases). More advanced AIs, the kind of AI needed to make the type of RPG I would like to see made, require more advanced techniques - being able to work with concepts, hierarchies, abstractions, etc. But none of these are new to the AI field - they've been done. Even the best strategy game AIs do not appear to be anywhere near state-of-the-art, and saying that RPGs (NPC non-combat AIs) are far behind those strategy game AIs is a giant understatement. Whether or not you believe a certain level of AI can be achieved in the near future (something which I'm not really interested in debating here), even you should be able to see that there is a crapload of improvement that could be made.
And no, it won't be singularity. Programming an effective AI is a massively harder problem than playing DF. (For more than ample evidence of this, see the large numbers of DF players who are entirely incapable of writing any sort of decent AI.) An AI being able to play DF does not imply that said AI could create something as capable as itself, much less an even smarter AI, so no singularity. (BTW,
here's a video of an AI that plays Dwarf Fortress, in fortress mode no less. GIT is
here. Now where's my promised singularity?)
You do realize that DF is in the middle of its development right now, right?
Irrelevant. The question is not whether it might be worth playing 20 years from now, but is it worth playing
now. I can't play the future version right now, only the one that exists now. That the future version might be worth playing in no way makes the current version so.
Have you considered the possibility that maybe you are just too intelligent and edgy to enjoy video games?
Have you considered the possibility that you have reading problems? I pointed out multiple videogames that I explicitly stated are fun for me, in this very thread, in posts that you responded to.
Everything you say is wrong.
You don't have to prevent the player characters from having to fight easy stuff. It may be that's just the way the game world is. What you can prevent is forcing the player to sit through boring combat - just let it go to auto-combat. And it doesn't matter if the game dev had no idea whether the player party would be weak or strong - just make the determination at runtime by comparing the player party with the mobs in question. Many games past and present have made such determinations (e.g., Warlords 3), so it's not like it's anything new. (And if the RNG gods are feeling mean and things go unexpectedly bad during an auto-combat, the game can then drop out of auto-combat and let the player take over so it doesn't get worse.)
You don't have to do weird level design to try to avoid backtracking. Just have fast-travel (fast for player, not fast for PC) so backtracking is relatively painless for the player.
You don't have to eliminate all "vendor trash" in order to avoid presenting the player with such tedious decisions. Let the player set filters and/or let the player define vendor trash by selecting items and saying "anything like this or worse goes in the vendor trash bag". Then for all future loot collected just automatically take care of throwing such trash in the trash bag for the player.
And absolute perfection is not the goal here. If we can get boring%:interesting% to go from 90:10 to 70:30, that's a win. If we can go to 30:70 that's even better. Just because 0:100 may in fact be impossible does not mean we should just be happy with something like 90:10 for the rest of our lives, especially when we know RPGs
can be made better
now and games in other genres are achieving better ratios.
Is there a reason why this isn't achievable?
0:100 probably is impossible because neither the game nor the player can with perfect accuracy predict
in advance what the player will or will not find interesting
in retrospect. In fact the player's opinion of what was and was not interesting may change later in the game - especially if the game is giving the player clues now that won't even make sense to the player until later in the story, clues that in some cases the player might not even figure out until years after they have finished playing the game. There might also be problems trying to present a coherent story while not saying anything about any events uninteresting to the player - some of those events may be key to understanding what's going on even if the player doesn't find them especially interesting, so in some sense the player
is interested in them because they want the story to make sense yet at the same time the player may never consciously realize that and so think they were uninteresting. (Also consider the movie genre - movies have total control over everything that happens in them. Yet what % of movies manage to be interesting to you all the way through - every last moment? I know for me that %, if not zero, is pretty close to zero. But that doesn't mean there aren't movies that I like, because I don't expect movies to be 0:100 any more than I expect games to be.)