The Lost Innovations

by Rick Powers - March 2, 2005, 9:28 pm PST

What aspects of gaming are still open to innovation by developers? Rick explores the possibilities ...

It’s becoming well-known in the industry that we are quickly approaching a level of graphics fidelity that will put us into a negative trend which Masahiro Mori called the “uncanny valley”, a level of detail where humans look real enough that we are disturbed by the sight of them. We’ve seen this in games already, where characters look startlingly real, except for a lack of life. You can never quite pinpoint the exact problem, but you know instinctively that it just doesn’t feel right. Even body scans, motion capture, and sophisticated graphics technology is incapable of creating a believable character that is indistinguishable from reality … we are able to tell the difference in mere milliseconds.

Yet with every generation, we’re employing ever-greater graphical technology in the hopes that we can sidestep this hurdle, but it simply can’t be done. Not in a single console generation (roughly about five years apart), and possibly not even in two. Some wise developers have already clued in to this, “The Wind Waker” being a notable example of backing away from the precipice of the uncanny valley and making characters more abstract, and thus, easier for players to anthropomorphize*, or imprint themselves upon. (It’s worth noting that the problem with fans accepting Wind Waker’s graphic style was not with the style itself, but was a problem of customer expectations.) In fact, the Japanese developers are keenly aware of this concept. Have you ever noticed that many popular forms of animated Japanese entertainment (manga, movies, games, etc.) all use simple or cartoon-like characters? Have you ever wondered why you seem to enjoy them so much? It’s because when there is a lack of definition in a character, you can imprint some of your own personality and feelings on the character. Beyond Good and Evil is another example of a game where the characters felt more real to us than more “mature” games employing a photorealistic graphic style.

The only way to bridge this gap (or at the very least, shorten it) is to invest in technology of a different kind, technology that will improve artificial intelligence and animation. While there have been minor strides made in terms of motion capture, animation blending, and other techniques, a character in motion never reacts in a realistic fashion to their surroundings. The only game in recent memory that really attempted to solve this problem was Half-Life 2, and even so, this doesn’t ring true in every case. Carefully scripted events can give the illusion of sophisticated artificial intelligence, but if the player does the unexpected, the illusion fails. NPC characters need to have the ability to react to their surroundings, react to the unexpected, and generally react as you would expect their personalities to react. World Driver Championship on the Nintendo 64 had a rudimentary AI system that established rivalries with the player’s character, and seemed to challenge the player during a race. Even just the hint of these “feelings” had players e-mailing the company with stories of how a particular driver “had it out for them”, when it was really a pre-programmed series of events. This worked because player behavior was somewhat able to be accounted for, since there are more limited actions a player can accomplish in a driving game. An entire generation later, Gran Turismo 4 employs no AI whatsoever, as the cars simply follow lines on the track and ignore the player entirely.

In terms of animation, a character needs to move in a more fluid fashion, which is always going to be a problem when you have a limited number of fields-per-second (or frames-per-second) to work with. Animation blending is a small piece of the puzzle, but to reach a more believable level, characters need to have micro-movements, the subtle indicators of life that we subconsciously notice. The way muscles move under our skin, the way our eyes react to light and to motion, the way our breathing can change under stress, these are all cues that we notice without paying particular attention to them. To date, very few games have even attempted to tackle these problems, and with varying degrees of success. All we’ve really seen are tech demos which are impressive in controlled situations, but still don’t fool even the casual observer.

Part of the problem is the adversity to risk that mainstream publishers have, not wanting to put money on an unproven technology or ideas. This necessitates a need for private investment in AI and animation technology. Sony promised advances with their “Emotion Engine” in the PS2, and they have largely been unrealized. Having a processor capable of handling complex data is nothing more than a blank slate, but developing software algorithms and structures are a significant time and money sink, and that is something that absolutely must be encouraged and supported. Theoretically, Sony’s Cell processor structure might make it possible to devote more resources to AI, but we would be back to the same problem that we have with the current generation.

A possible answer would be tools developed by the first parties, who have the resources and incentive to create more realistic characters, which could be licensed to developers. Rendering engines, sound technology, and many other tools are already being licensed in order to shorten development time and cost, and this reveals a unique opportunity for first-parties to secure a compelling feature set for their console.

There is something that can be done now, however, to help alleviate the problem and work towards creating more realistic characters. Animators need to study traditional acting techniques and human psychology, more than just kinematics and movement. Pixar’s animated films are largely successful not just because of the story they tell, but because of the effort the animators put into acting the part and using the animated character as a proxy for that performance. If video games are really going to become the dominant form of entertainment going into the next generation, we need to vastly improve upon the current B-movie level of acting performance and rely less on motion capture to tell the story. The characters need to be acting all the time, constantly communicating information even when they simply exist in the game space. What is the character thinking when not interacting with the player? What is their motivation, not just in terms of the plot sequences, but from moment to moment?

Surely there are even more options, and perhaps backing away from photorealism as an aid to create more believable characters is a necessary step while we work on technology that will allow us to put more emotion and feeling into the games. Most importantly, developers need to take AI and animation technology seriously as a necessary component towards creating compelling worlds and engaging stories, and even advancing the entire industry in the coming generation.

*Anthropomorphization is the act of seeing human traits (feelings, thoughts, and actions) in objects and animals not fully capable of those processes. Commonly, people tend to do this with their pets, assuming that the animal is communicating on a much more complex level than they are truly capable.

Please visit our Talkback forum to discuss this editorial with other readers and staff members.

Got a news tip? Send it in!
Advertisement
Advertisement
Advertisement