© Written by Rob Ager July 2014
CHAPTER TEN
ARTIFICIAL INTELLIGENCE
I mentioned earlier the lack of technological progress in terms of making computers able to hold convincing conversations with people. And the cause of this limitation is a direct off shoot of the snail pace progress of artificial intelligence as a whole. I’ve also recently written a review of the Spielberg / Kubrick collaboration film AI: Artifical Intelligence. While researching the film I came across a lot of literature and video interviews in which scientists and techno-enthusiasts drastically over-estimate mankind’s ability to imitate itself with computers. In fact I found the subject so interesting I posted a separate article about it titled Androids and Artificial Intelligence: A Modern Myth and then expanded that article into a much more in-depth video.
Back in the 1960’s AI was hailed as a great emerging technology with experts assuming we would have robots within as little as thirty years that would be able to perform all of the same functions the human brain does. The assumption that man will soon be able to imitate his own mind with machines has since become so widespread that thousands of sci-fi movies and novels have predicted talking robots that perceive and communicate like everyday people. But it’s an illusion. We’re now in 2014 and we are nowhere near achieving this God-like dream, yet there are still ridiculously hopeful claims being made such as one scientist anticipating that within another thirty years we’ll be able to upload our entire minds to computers.
Sure we’ve built some incredible robots and machines that do impressive things and sometimes we design them to appear life-like, but not once have we made such a mechanism that can actually fool a human into accepting it as being alive. Despite the technical skill that goes into machines like these, they are no more self-aware or alive than a common toaster. They’re just mechanisms. And even though these machines perform some initially impressive feats, if you actually stop and compare what they’re doing to the vast complexity and intricacy of what an average human being does effortlessly from moment to moment the winner is obvious.
Because of our tendency to be unimpressed with things that we are familiar with, crowds of people will gasp in awe at advanced robotics demonstrations, while forgetting that the human beings doing the presenting are actually demonstrating abilities in their momentary behaviour that are far more sophisticated than the robots they’ve created.
This failure of AI in replicating the human mind is the main obstacle that limits video game interactivity. Although NPCs in games are often coded to do some quite intricate things they are no long-term match for the gamer, who simply observes the behavioural patterns of those characters and then devises reliable ways to defeat them in battle. In effect the gamer creates a little piece of coding in their own brain to counteract the coding of the NPC. Even little kids succeed in doing this. To compensate, we now have online gaming where we can interact and compete with other human minds in order to take on a real challenge. And, for some gamers, going back to gaming worlds that only consist of NPCs simply holds no appeal.
One form of compensation for the failure in replicating human thought and behaviour, which was touched upon earlier in this article, is that game designers have real people perform small chunks of spoken dialogue and physical movement that are then recorded into the game environment for context-specific playback during gameplay. The results are often marvellous, but would it be if those NPC actions hadn’t been performed by a living breathing human in the first place?
Such limitations in AI are also the central reason why video game narratives are so predefined. The overall assembly of quests, plot points, and character motivations within the game story are exclusively thought out by human game designers because the intricacies of creating a compelling fictional story, even in the frequently dumbed down generic forms found in most video games, are beyond the scope of artificial intelligence. There isn't a single computer on Earth today that can create, from scratch, a fiction story, much less one that humans will find compelling.
This imposes a dilemma on game designers. How do we create fiction stories that are truly responsive to the gamer? Given the thousands of different behavioural choices a gamer might engage in within the game environment, how can a story be made to truly respond in a way that there is always going to be some sort of satisfying conclusion that does not limit the choices of the gamer during play? What if the player chooses to kill a central character whose presence would otherwise be found in the predefined end of the story? What if the player chooses loyalties that conflict with the predefined ending? What if the player avoids key gaming environment locations that are integral to the predefined ending? And what if the player fails to take particular actions or complete specific sub-plot tasks that are required to justify the ending the designers have in mind?
To truly make game stories responsive to the player would require that the complexity of NPCs (and ranges of their actions) be increased by thousands of times. And I'm not just talking about processing power. Human game designers themselves would have to unlock some of the mysteries of how the human mind actually functions in order to infuse human-like responsive qualities into their predefined NPCs. Basically, there would have to be a combined quantum leap in general human understanding of our own minds and how to reproduce it's versatility in computer program form. Only then can video game stories truly respond to the psychology of the player.
But even if the aforementioned AI advances were achieved, there is an additional problem. The concept of the satisfying story ending itself defies reality. In real life there is only one story ending - the moment of death. Everything else is just ups and downs on a continuum. So the game design would have to incorporate ways to both emulate the vast behavioural choice and consequence structure of reality AND contain the means to always facilitate a satisfying story ending that meets the psychological requirements of the player.
As far as I can tell, there is only one practical and achievable way to deal with this game design dilemma. And that is for a complex myriad of plot twists and varied endings to be conceived and encoded into the game by its human designers - a sort of Groundhog Day design involving hundreds of different chronology outcomes. These scripted variations could never truly reflect the range of outcomes that people's actions can take in the real world, but if enough of the gamer's potential story-affecting actions were anticipated then it could lead to a game story architecture far more satisfying than the stale one-dimensional stories still dominating most new game releases today.
As well as the time and cost of designing such a complex web of game story variations, the task is made even harder by the current trend of expensive cienmatic cut-sequences in games. Key transitional points in main quests are voice-recorded and physically acted out by actual human actors to give a stronger emotional engagment for the player (though I personally take exception to it, as I find such bottle-necked cut-sequences disengaging because they temporarily reduce me from gamer to choiceless observer). If cinematic cut-sequences were to be used in a game whose narrative featured a multitude of responsive narrative pathways and endings, then it would result in an equivalent escalation in the time and effort of working with real world actors during game production. This is impractical. Such a game would have to drop the currently over-valued cinematic cut-sequence almost entirely and probably replace it with on screen text-communicated plot transitions.
In video games interactivity ironically brings out the total artificiality of the experience, but with movies it’s a different ball game. There is no input from the viewer so the movie doesn’t have to respond. As long as the script, acting, direction and other aesthetics contain enough internal logic to persuade the audience to emotionally and intellectually engage, the movie has succeeded on the artificial intelligence front. But imagine if we could interact with a movie. Like with AI in video games, the lack of convincing responses by actors and the off screen film making crew would utterly reveal the staged nature of the experience, as often happens with stage plays.
And here’s an interesting little notion to put the whole AI issue into perspective. If we are ever able to truly imitate the human mind with machines then a good test of whether we had succeeded would be to ask a robot to write and direct a fiction movie that would be accepted as a work of art by human audiences. Asking such a machine to also create, from scratch, a video game that humans would enjoy playing as much as the average commercially released game would be another worthy test. Can you honestly anticipate either of these scenarios ever happening? Personally, I can’t.