Massachusetts Institute of Technology, 2000
Over the next five years video games will have survived a full ‘Family’ life cycle. Starting around the mid 1970’s games have been around now for about 25 years… New parents today with children, have themselves grown up in this video game era. These ‘gamer’ parents are very comfortable around them. They ‘get-it’ and I fully expect to see them commonly playing on the newest machines, with their own children, welcoming video games into their homes.
With this new hardware comes incredibly better visuals… Today, all 3D video game graphics are still displayed in polygons, the more polygons, the more incredible the images. The problem is that on PC games for example there are endless possible configurations of video cards, sound cards and processors. So we just don’t know how powerful any single person’s machine actually is.
At Shiny Entertainment for the last 3 years we have invested heavily planning for the extremely high polygon future, but also keeping the technology compatible with all machines today. So we call this a scaleable real-time deformation and tessellation engine. This means that our games constantly monitor the performance of your hardware… They actually stress test your machine WHILE you are playing the game… As they see your machine start to choke, then the engine eases back by removing polygons from the scene in real-time. If you drop another complex character into the scene or cause a massive explosion, it automatically steals polygons from the background world and from the other characters to keep the game speed stable. This is an exciting technology because it means that no matter what hardware combination you have, our engine will always push it to 100% of it’s ability 100% of the time. You will see the richest display that your personal machine can handle. This is important because most of our competitors have to try to guess the average power out there around the world and design for that. So on some machines, their game runs very slowly indeed, on high-end machines they don’t significantly push the hardware at all. Business wise this means our games are designed to happily support Moore’s law. So by the time the game finally hits the shelves (usually about 2 years after starting), gamers will see higher quality displays from the game than we ever saw during development.
My favorite kinds of technology are the ones that don’t just improve visual or audio quality but that also give us a chance to design better games. If we have to worry about some theoretical average worldwide PC and design the game for it, limiting gameplay to maybe 6 large characters on screen at once, then the games design is already very constrained before we even begin.
The normal solution is to use level of detail models, for example 3 versions of the same character using different polygon counts. One for if the character is up close, one for the character when he is quite far away and one for when he is very far away. Using our scaleable technology which we call our Messiah engine, due to the algorithm we get infinite levels of detail, so as a character approaches, he very smoothly has new detail added to his body. This allows me to easily add 100 or 200 3D characters into a scene. This lifts a giant design shackle that we have all been faced with for years, especially if you try to make an epic War game but can only have six 3D soldiers on screen at once.
Another interesting part of this technology is that by re-creating the polygonal character in real time, we can also add influences with little penalty. So we can have subtle volumetric lighting on each polygon, and skin deformation as bones push against the skin. It even allows us to add random artifacts that make people look different. For example right now in video games characters of a certain class like an archer all look absolutely identical. That’s because they ‘are’ identical, it’s the cheapest way to store them. Like in real life, by just changing a few parameters we can make characters that are the same class at least look like individuals by giving them attributes like longer legs, a larger head, big belly etc… Peter’s next game is proving this point to an extreme.
Some research friends of mine at Caltech are also developing a system for adding more personality to each characters motion. They came up with a way to blend motion capture to combine for example a happy walk with a large person or light person, or an un-happy person stepping over a dead body… It looks great. Here lies a problem. Academics luckily mostly work without the restrictions of games (except funding that is!)… They constantly solve problems that we would LOVE to benefit from, that could easily change video games in many ways… But when my friends presented me with their system, it was in Matlab, it took lots of storage space and could not handle looping animations. That said, I expect to see the video game community embrace people like this in the future and I expect to see some remarkable physics, generated dynamic human locomotion, real-time pore tracking facial a animation and things like that in the future. Somehow we just need to communicate our restrictions better!
The extra quality in game displays and the constantly improving processor power also help move us forward in the immersive experience. In the old Pac Man days, very crude characters kept many adults thinking video games were just some childish toy. Nowadays we can Cyberware scan both exact body shape and texture using lasers and and then capture an actors subtle motion with the latest visual capture studios, so combined with future display hardware we will have have photo-realistic actors in the game. To see Keanu Reeves or Tom Cruise or Leonardo di Caprio looking real and speaking and acting in real 3D I think will help video games appeal to an even wider audience. You never know… that might even get us heavy interest from more of the other 50% of the potential world gaming market… Girls.
The problem however is that the better characters look and the better they sound, then the production costs are going to sky rocket. You will see many video game companies go bust over the next decade and you will see a lot of symbiotic relationships formed with talent agencies, Hollywood special effects studios and Movie studios to help share the ever-escalating costs.
These relationships might also help us reach our ultimate goal… to make it more common to sincerely ‘care’ about the characters in the game. We strive to keep increasing emotion states…. Making you scream at the screen or burst into tears under our control. Today we make you put so much work into developing a character so that when they die, you do feel a loss. That loss is often just for all the work you did, not because you are in love or have compassionate feelings for that character.
This is where we can learn from movies and books and psychologists about getting better at character development… Making characters that truly have endearing personalities. Making them ‘fun’ to spend time with.
That leads me to AI. 90% of video game AI is still appalling. I think that’s why it’s so much fun to shoot characters in games… They are so damn annoying and stupid… They just keep saying the same stuff over and over and getting in your way. Compared to real people, these characters are in no way artificially intelligent, even some brand new games today still contain artificial idiots and it seems like we are many years away as an industry from getting this right. In the real life AI field I believe MIT has had the same problem with robotics research… One of the main issues is that basic common sense alone has millions of parameters, and many of the decisions we make are from emotions and knowledge, not a script or logic tree. So I don’t think you will see characters in games demonstrate Real ‘real’ life until we totally change our way of developing them, but I would expect a ‘kinda like’ real life. That is why games stay so focused to genres… You can make a pretty good Soccer game with players that know what they need to do but you can never leave the stadium. In driving games, you are stuck to simple tracks/roads or in war games to the current battlefield. To remove these restrictions would provide absolute freedom and be an absolute programming and resource nightmare!… For me, this true Freedom is the holy grail of interactive entertainment that I hope to see at least in massively multiplayer games before I am old and grey. Either that, or I hope someone at MIT finds a way to extend my life so that I can still be there waiting. 🙂
The 3rd part of the equation is audio… Currently we have 5.1 surround sound, Dolby, DTS & Environmental audio but you can expect to see proper sound rendering next not just sound placement in 3D space. This will require echo and material absorption so if you smash out a window in an office, the sound in the room would correctly change etc… Again this will just add to the immersive experience. The ambiance in a video game is worth it’s weight in gold.
One thing that MIT has demonstrated but is still not working well in games is real Voice recognition. There is no standard on computers and no consoles even have microphones. The problem is that video games need to run fast, smooth and consistent …. This creates the crucial ‘feel’ of the game. We cannot have a great but slow voice recognition system randomly hogging our always limited system resources. Our hope here is that this will be handled on chip in the future… Allowing us to write fast and efficient systems to get the words we need. Today we only detect simple words that sound quite different or the stress on the voice…
Actually on that note… sensing stress on the voice can be coupled with sensing frustration from the gamer… We can ‘watch’ you from within the game, we can see and feel you stabbing a key when you can’t get the door handle to open. We can see you are lost and you keep walking past a key object… There are 1000’s of ways to SENSE what you are experiencing and what emotions you have… This can add great tools to the designers tool kit, especially when we see you do something good and we know you are feeling good, we can back that up… The game can respond… We can manipulate your feelings in real time because we know when you just fought off 10 guys, and have low health that you are nervous. This is a major focus of the next game we are working on.
Anyway, back to the voice recognition… Our problem is being forced to keeping the word library small, to keep the memory footprint low and to able to spot the words easily. This way the game code does not have to ‘understand’ anything… However if I am flying an fighter aircraft I can say “Air to Air missile”, “Lock target”, “Fire” whilst concentrating on flying the plane. The problem is that if your algorithm accidentally misunderstands “go back” as “Attack” you’re screwed!
To wrap up my intro… A message for people here that don’t play video games or think they ever will… Don’t worry, we WILL get you! There is no escape!!!… We will get you playing something through your cell phone, PDA, hotel room, airplane seat, wireless devices, watches, cable boxes or email… So you might as well start practicing now!