Gamasutra: The Art & Business of Making Gamesspacer
The Future According to Epic's Tim Sweeney
View All     RSS
October 25, 2014
arrowPress Releases
October 25, 2014
PR Newswire
View All

If you enjoy reading this site, you might also want to check out these UBM Tech sites:

The Future According to Epic's Tim Sweeney

May 7, 2012 Article Start Previous Page 2 of 3 Next

How much emphasis on AI is there at Epic? Because a lot of people, when they think of Epic's tech, they think of nice graphics. What about the AI side?

TS: Well, the gameplay team at Epic puts a lot of thought into simulating characters in our games -- you know, your friends in the game and enemy opponents -- and so we've developed a lot of interesting solutions that help with AI.

There's a navigation mesh which gives each character an overview of how to walk through the level, and how to understand different parts of the level. If an enemy decides he needs to fight you then he'll realize, "Oh, I need to run up these stairs, and then fight him from above where I have a better angle to fire on him from."

There are a lot of basic systems like that, but ultimately it comes down to the gameplay team designing custom AI for each character and the various actions and scenarios that are carried out. It's a pretty special case; it's not a general simulation of intelligence by any means.

Is there any tech that you see happening outside of the games industry that inspires you with your own job?

TS: What's really impressed me? In the last decade, Google search and its uncanny ability to find what you actually mean that you're looking for; there's Siri's voice recognition.

Word Lens, this iPhone app; it's magic to me. You take out your iPhone and you point it in some direction and basically the camera takes a live feed, a video of the world, and it projects it onto the screen in front of you with all of the words translated from one language to another. But not just translated with captions at the bottom, it's translating like if you have a big red stop sign it's translating "Stop" to Spanish and putting it in the proper perspective with the proper lighting in the scene.

And that's a crazy advancement. Just seeing that makes me think that we're on the cusp of seeing a whole generation of augmented reality apps over the next few years that will change the way we interact with computers. Just a few pieces of growth algorithms have shown through like that over the last few years.

Google Goggles lets you do a search by taking a picture. So they're working on image-based inquiries as well.

TS: Oh yeah, and Amazon had a service like that, too; I thought, "Wow, that's an amazing algorithm there!" What you do is there is an Amazon app, you take a picture of an object and a few seconds or a few minutes later it sends you a link to that product on When I saw that it's like, "Wow! They're doing some amazing image recognition tracking technology!" But no, they have an army of people who watch your pictures and then look up the appropriate thing on and email it to you.

What?! [laughs]

TS: I know! It's the joke about Siri, right? That the real trick is that that they have a warehouse full of people in China who are typing away at your translations.

So what about the browser? That's something that relates more to something specific that you guys are working on with Unreal Engine. Where do you see the future of the browser in games?

TS: Well, we would like to see the web browser as another platform. You should be able to take any game -- a PlayStation 3 or iOS game, for example -- and just go to that and play it from any web browser.

We're slowly heading in that direction as an industry. One thing that's happened recently is Adobe Flash. For a decade or more, Adobe Flash was a little scripting language for creating more interactive webpages using a proprietary browser plug-in, but more recently Adobe created a translator.

You give it any C++ program, like Unreal Engine 3, and it translates it to a platform-independent application that can run within Flash, within any web browser or on any platform where Flash runs.

And so now any browser that supports Flash can play any web game that's built with Unreal Engine 3, or any other engine that's cross-compatible with Flash. That's an awesome breakthrough; it shows you the possibilities.

But I think the next step in that is cross-compiling games from C++ or whatever and directly running them as native HTML5 and JavaScript applications within any standard web browser. And you can do that in theory today, but it ends up being slow and unstable just because of the early state of JavaScript implementations, and limited performance, and current web browsers.

In another few years, I think that's going to be a very realistic scenario. And so the web will generally be a platform, and you can have a real application with a full feature set that runs within a web browser; that'll be very welcome. The web is a fairly awkward experience when you use a platform that's not the majority of the install base, and I think we're going to see big improvements there in the next few years.

Article Start Previous Page 2 of 3 Next

Related Jobs

Red 5 Studios
Red 5 Studios — Orange County, California, United States

Graphics Programmer
Red 5 Studios
Red 5 Studios — Orange County, California, United States

Gameplay Programmer
Gearbox Software
Gearbox Software — Plano, Texas, United States

Server Programmer
Forio — San Francisco, California, United States

Web Application Developer Team Lead


Kyle Bue
profile image
Wonderful interview! What I found very interesting is the AI subject. I'm currently taking a Psychology class as well so that helps a bit. But I think it's nearly impossible to replicate the human mind right now if you really think about it. Not only do we just not have the technology to do it, but we barely even know how the mind works at all. We have a miniscule understanding of what the mind can do and why it does what it does, and we want to replicate it? We just don't have enough knowledge of the mind to try to replicate it.

However, even if we did know how it all worked, it would still be inconceivable to try to replicate it in a game (although by the time we actually do know mostly how the mind works, the technology will probably be advanced enough anyway). You have to think of the costs involved and how much coding would be needed just to make one AI character with a human conscious. You would probably have so much code in that one character that it could be a full-fledged game. Of course I'm speaking by today's standards so who knows what the future holds.

Another thing that we have to think about is what really defines a "game." People may define what games are differently, but what most people think of when they hear a game is an interactive experience that has specific rules that you must abide by in order to achieve a specific goal. That could also include milestones along the way to that goal. So what would happen if you gave AI characters in video games an actual human mind? No one really knows of course, but I theorize that it would almost be too random to really be a game. Right now, AI in games are limited to what orders the programmers give to it so they will only perform those certain actions when they are told. But if we give them their own personality, their own state of mind, I think it would just be utter chaos. We have to limit AI so they can do things that make sense in accordance with the game itself; so things don't get out of control. What if the AI were to do something that isn't even coded into the game at all?

We don't usually think about it, but there are just way too many factors to even think of trying to replicate the mind in an AI. Until we know how to overcome those factors, I don't know if we'll even be able to see an AI have even close to the amount of brain power that a person has.

Well, I think I've said enough. I could go on and go into more detail, but it already feels like I've written an essay. It's only a comments section! Sorry for that! It was just a very interesting topic and I had to get it out! Thanks for reading if you did.

Joshua Darlington
profile image
I think it's a false assumption that you would want to "replicate the human mind." There is only the need to represent behavior, which is a much easier problem as you can fake and approximate many processes.

Kyle Bue
profile image
@Joshua Precisely. And that's what I mean. We can't and really shouldn't even try to replicate the mind at all. We can make AI perform smarter actions based on the rules of the game, but in my opinion, giving an AI in a video game true thought would be a train wreck. So we should just stick to giving specific actions to AI for the time being and just try to improve on that concept. Like I said, it probably wouldn't even be game anymore if you did allow the AI to think for themselves.

Joshua Darlington
profile image
"We can't and really shouldn't even try to replicate the mind at all."That's a really weird thing to say. We shouldn't try at all? Please explain.

I agree that we dont have a robust scientific theory of consciousness and we cannot create artificial consciousness, at this point in time. But I dont understand your concern about creating a mechanical (syntactic) automaton that behaves in a convincing way. To excel at such a craft one would likely need to model various consciousness-like dynamics (hierarchical and linear plasticity, task selection, intentionality and etc) using multi-realizable syntax. The brain has something like 200+ structures working in parallel. Even with representations of each of these processes an NPC AI would be something like an advanced coocoo cock.

Stanley de Bruyn
profile image
My opinion is also that AI is runing behind, but also is a daunting task to do right. Also simulating AI like on a human level. Is a brute force aproach. Comes down to simulating biological neuronet. And a human brain excist out of a very large biological neuronet.
But dont forget some part of the brain are just interpreting input from both retina cells from both eyes. It also means to simulate a brain je feed it with a 3D render input into the masive neuronet.

Also the AI needs of games vary. but most of them don't need full human intelligence wich means simulating multiple NPC AI with brute force means a massive amount of computing power wasted, because of huge non relevance to restricted gameplay.

Now they mimic AI with smart practical tricks. With more computing powers those tricks just can takes more scenarios into accaunt get much complexer so AI can behave believable for fraction of brute simulating force.

But I also know that there is a long road to go. Like FPSRPG like Fallout3 or MassEffect3.
realistic interaction with beliavable AI means you interact with your voice and those NPC speak back. Well I do think some barriers need to be broken to archive that.

David Navarro
profile image
A game with any decent approximation of "real" AI would be baffling, incomprehensible, boring or downright immoral, depending on implementation. The real challenge is not to actually improve AI beyond basic mechanical features such as pathfinding and not blocking doorways, but to improve game design so that enemies provide a decent but not overwhelming challenge and are predictable enough to make you feel clever, but not so predictable as to become boring - and friends keep close enough to remind you they're there, but not so close they get in your line of fire.

To wit, most AI issues aren't related to code, but to game design.

Joshua Darlington
profile image
I'm curious to hear more details of your incomprehensible, boring or downright immoral implementation of AI.

Bob Johnson
profile image
Yeah AI in games is sort of a mirage. It wouldn't be fun playing against a "smarter than you" human AI. :)

Most of us don't have fun playing multiplayer online games vs opponents that are much better than us.

It would't be fun to play tennis against Roger Federer or basketball against LeBron James other than the star struck aspect of meeting them and saying you played against them.

Stanley de Bruyn
profile image
I had some thought on this to. I play a game like X series.
Currently X3: albion Pride.
The enemy but also your friendly AI is very limited. Like percived like dumb as a dodo.
Well the result is it aint that of a challage its easy and get killed game over is rare.
even if you not that good at space combat in a fighter.

Smart AI means you die A lot. If your not so good it might brake the game.

If role playing a Hero that combat as a novice instead of a Ace there is a problem.

AI schould adopt to player level. Even with Easy hard extreem you can set the adoption to mild or go up faster. Some gamers play to for the plot and a smooth right through it. Other want to be challanged. Both need a different adotion. One for a easy ride with decent doable AI. While the other need a decent chalange adopted to player Dynamic skill level.

Online FPS Mplay got addictive grinding RPG online feature so you keep playing against better players. Leveling and unlock grinding. Also in today leading FPS online games COD MW3 and BF3. Also it give a big satifaction if you get the better player. But yes loosing is no fun.

Keith Burgun
profile image
>>Is that some kind of holy grail or ultimate goal that video games can accomplish -- an exact simulation of life?

If you think the terms "game" and "simulator" are the same, then yes. If you think that games are actually their own thing, as I do, then absolutely not.

Joshua Darlington
profile image
If you are interested in dramatic entertainment (narrative design), better NPCs would be a powerful tool. In order for NPC dramatic conflict to work, a simulation of physical resources and kinship bonds would be a powerful too. Ideally such a system would be a player independant simulation - as it could provide more substance to the gamespace.

Saying that we need to wait for the Singularity to accomplish decent NPCs seems iffy. Memories compose a great deal of human mentality. So a post singularity AI would still be a superficial approximation of a person.

The OG chatbot Eliza is from 1964 or whatever. The minimalized branching dialog systems used in todays RPGs are not even up to that level of tech. Where are the modular heuristic and semantic tech tools for game NPC chatbots? Is anyone talking to the IBM Watson team regarding game tech? They seem to have the NLU lead.

Bruno Xavier
profile image
WHY does everyone thinks gaming is about socialization or wants to turn it to that nowadays...
Meh if I play games, it is exactly because I dont want to go out and socialize at the very moment I am playing.
If even core console companies like Epic are aiming social bullshit, in few years from now, I think I will need to find another hobby where I can NOT socialize with anyone.
Sometimes you need to be alone with only your own crazy mind, it is very important to maintain sanity.

Titi Naburu
profile image
Sweeney talks about openess. I'd rather talk about freedom. Like, GPL is compatible with the Apple App Store.

Jamie Roberts
profile image
Brains in living organisms are a complex combination of many different specialized systems. A combination of maps that are cross-referenced. There isn't a single model of thought, as brains are unique entities despite their overall similarities--adaptable and using different components based on species and individual condition. Engineers are slowly building digital equivalents of the most basic of cognitive systems: we have facial recognition and basic data retrieval (search), as well as the beginnings of object recognition. They are nowhere near the power and elegance of naturally evolved systems, but they're what we have. Eventually these individual systems will evolve into greater and greater complexity in isolation. The trick will then mixing them into an integrated system. If we ever do achieve a sci-fi type of AI, it will be through the collaboration of several organizations each developing one of these specialized systems, not via a single developer. The task is just too large to complete otherwise. The "singularity" will not be the developing of these isolated systems, but in their integration.

Stanley de Bruyn
profile image
Well it is very common to use middleware in the game industry so it makes a lot of practical sense to specialize.