IA on AI

On EverQuest Next and AI-Driven MMOs


When I heard the news that Daybreak had cancelled development of EverQuest Next, I admit that I wasn’t surprised. I had expected it for quite some time — as did many others. There was no real sense of loss or remorse. Any grief process had been drawn out over the course of well over a year. In a way, it is similar to watching someone die slowly of a debilitating condition of some sort. When you finally hear the news of their passing, your reaction is often simply, “Oh. OK.” You knew it was coming and you had already both resigned yourself to the eventuality and prepared yourself for life afterwards. This was the same… I had already grieved slowly over the course of 2 years — not just the end of my direct role in the development, not just the knowledge that no one else had picked up on what I had started, but the knowledge that even the concept of the game was going away. Just as with the example of a loved one, EverQuest Next, had been dying or dead for a long time.

At least in this form…

And that’s why I’m writing this.

Emergent AI was not born at SOE/Daybreak or born of Everquest Next. It also did not die with it.

One of the things that has been difficult in the 2 years since I left the project, and even more so in the year plus since Storybricks left (and subsequently shut down), was watching the commentary of the fans. So many times, I saw speculation about the AI and what it could do. Most of the times, they were correct. Yes, my AI system could do a lot of things that people had been wanting. At other times, I saw the opposite — detractors swearing that the world-spanning, emergent AI that we (I?) had promised was impossible. Some people even posited that it was the AI that was the reason for the protracted development cycle and even the ultimate cancellation of the project. While I don’t know exactly what went on at SOE/Daybreak after I left, I can almost guarantee that the AI was not the problem… or at least it wasn’t when Storybricks left a year ago.

With the announcement of the cancellation of Everquest Next, another faction has cropped up (likely overlapping significantly with the first one above)… those that lament that we will now never see an MMO with that sort of “living world” and “emergent AI” that was promised in the EQN reveal at SOE live almost 3 years ago. It is to those people that I write…

Please hear me when I say, emergent AI was not born at SOE/Daybreak or born of Everquest Next. It also did not die with it.

What many people do not realize is that most of those concepts came from the mind of the Storybricks team — myself included. For that matter, many of the different features actually had nothing to do with Storybricks, but rather, were my own. In fact, the examples that were used by Dave Georgeson in the reveal presentation — including the “Orcs by the road” — were lifted almost exactly from a lecture I gave at the 2009 GDC Austin (subsequently GDC Online)… almost 3 years before anyone from Storybricks (including me) pitched anything to SOE at all. Dave came upon that idea after Stéphane Bura and I were presenting bits and pieces of our prior lectures for him and other principals on the team. I didn’t even know he was using that example in the reveal until I saw him present it. Admittedly, it was a good example of the dynamic nature of the world we were creating.

That lecture was meant, at the time, to be a thought experiment — what would happen if we actually did put decent AI in MMOs? In it, I challenged a lot of tropes that the MMO scene had fallen into: the Holy Trinity (specifically as it applied to ridiculously contrived aggro systems), static, unchanging (and unchangeable) worlds, simplistic “big bag of hit points” enemies that operated more like a mechanical wind-up toy than an intelligent being, etc. My point was that we could do better in many facets.

Thankfully, I wasn’t the only one who felt that way. Dave G., Jeff Butler, Darrin McPherson, and others at SOE often said, “we are making the MMO we always wanted to play but couldn’t do technically until now.” That is true for many of us. More importantly, there was one person who was in a person to push to make it happen — John Smedley, CEO of SOE. John had for years been interested in creating an AI-driven MMO. Sure, there were gameplay reasons for doing it, but also it makes good business sense. In the traditional “theme park” model of MMOs in the past 15 years, you can only create as much content as your army of designers and artists can crank out. It is well-established — not unlike the cliched Thanksgiving dinner cook’s lament — that it takes far less time for players to consume that new content than it did for the devs to create it. With the size and scope of today’s MMOs, it is generally impractical to hire enough people to generate a non-stop conveyor belt of content at a rate that ensures the every time a player logs on, something in their world is different.

No designer intervention necessary.

But what if you could create a world that, through its systems, was inherently dynamic? To revive a phrase from my 2009 lecture, and one that I used to pitch the idea to SOE, what if all of that happened with “no designer intervention necessary?” THAT was the power of the “Orcs by the road” example… the world would change by itself — in reasonable, believable ways — without having to have designers authoring those changes. It’s fun for the players as well because they had the potential to be the impetus behind those changes. The orcs didn’t abandon their ambush location along the road because a patch came out with a new scenario… they left because the players cleared them out. For real. They weren’t going to respawn in 2 hours… they were gone. Either dead, fled, or rebasing somewhere else safer and potentially more lucrative for them. No designer intervention necessary.

We were breaking things off the pedestals and rails…

One other thing I brought to the project was my ability to create an AI system that provided not only interesting group behaviors, but also compelling, dynamic, realistic individual behaviors. These included combat behaviors that were unlike what people were used to in MMOs and also interesting non-static ambient behaviors in cities, towns, villages, and in the wild. We were breaking things off the pedestals and rails to which so many MMO NPCs are affixed. We were setting them loose with their own needs and desires and an awareness of their environment that allowed them to seek out their own solutions. And yes, we were giving them an awareness of you, the player.

All of those things that were promised with the AI were true and deliverable. It was not vaporware. In fact, after leaving the project (something I did not want to do), I ended up going straight to ArenaNet (thanks to my dear friend Mike Lewis) and spent a year implementing much of the same architecture and design. Some of it was used in the Heart of Thorns expansion to GuildWars 2the rest was prototyped up for possible use in future ArenaNet titles. (Much of the reason that some things were not used is that they were too much of a departure from the established gameplay of the GuildWars 2 franchise.) If you want to see how some of the system I implemented at ArenaNet works, watch this lecture from the 2015 GDC AI Summit. It is more dev focussed and it certainly isn’t all of my multi-faceted AI architecture, but it will give you a glimpse.

There is someone who believes in the power of AI-driven MMOs

As of last fall, I’m also working with John Smedley at his new company, Pixelmage Games, on Hero’s Song. Yes, I’m implementing the same sorts of things that where intended for Everquest Next… a large, dynamic, living world driven by emergent AI from the ground up. When I found out that John had started his own company, I knew that he was going to be pushing for a game like that. As I mentioned, he was the one who wanted an AI-driven game when he started EQNext. I sought him out and offered my services even before I knew what the game was. 4 months in, I can’t tell which of the 2 of us is more tickled to be working with the other — him that there is someone who can implement his ideas or me that there is someone who believes in the power of AI-driven MMOs enough to enable the development of one.

So while I certainly understand people’s disappointment at the cancelling of EverQuest Next, please don’t believe that this is the death knell for MMOs with living, breathing worlds, with NPCs who are engaging, challenging, and often surprising. I am not the only talented AI dev out there. As the co-founder of the AI Game Programmers Guild, I can ensure you that there are plenty of them.

Unfortunately, many companies are simply afraid of non-theme park MMOs and “smart” AI. Believe it or not, they truly posit such premises as “players don’t want smart enemies” and “players want predictable puzzles” and “if we don’t tell the player what to do, they will be lost”.

This flies in the face of one of the most memorable exchanges that I ever had in a game… and it was in the first month of the seminal MMO, Ultima Online. A new player trotted up to me in town and asked, “so what do I do?”

“Huh?” I responded.

“What am I supposed to do in the game?” he persisted.

“Anything you want to do.”

“No, I mean, what’s the story?”

“Anything you want.” I replied, not really trying to be as cryptic as he thought I was being.

“There’s no plot? No story?”

“Story? Go walk outside that city gate over there.”

“That’s where the first story is?”

“You walk outside that gate, trust me… you will have a story to tell. Anything that you chose to do and anything that happens to you… THAT is the story.”

No designer intervention necessary.

Emergent AI is possible.

So while it isn’t necessarily in your power as players and fans to create an MMO that will have the types of AI-driven experiences that EverQuest Next provided, you can make a difference. First, don’t give up hope. Emergent AI is possible. As an AI consultant, I’m doing it every day — currently for Hero’s Song, as I mentioned above. While not as bold and sweeping as EQNext (and maybe that’s a good thing), it will have a lot of the same features that you were expecting from it.

Second, continue to make your voices heard… tell game companies and the game media that you want something more than the latest themepark MMO with pretty graphics and meticulously crafted stories on a pedestal and quests on rails that you consume in weeks, days, or even hours.

Tell them that what you really want is to walk out that city gate, chose to do something and have something happen to you, so you can have a story to tell. YOUR story. As I said back during the EQNext reveal week (and Dave G. also parrotted)…

“If you don’t go out and find our content, our content may very well come and find you!”

No designer intervention necessary.

Tags: , ,

18 Responses to “On EverQuest Next and AI-Driven MMOs”

  1. Warlander says:

    After I first was story bricks well before EQN got their mits on it I knew there was something very special about it. Thanks for trying and hope to see your work in the future

  2. Syraleaf says:

    That was a very interesting read! As someone who originally followed Heroes & EQN this was a pretty good read, and not only that, but listening to the storys about a ‘good’ ai… its just great! I honestly cant wait until we can get our hands on it!!! 😀

  3. Ayon says:

    Hi David
    Great to see your AI work/concepts being used in ArenaNet and with Pixelmage Games.
    it was one of the draw cards that got me laying LM while we awaited EQN

    Looks like I’ll be playing Hero’s song in the foreseeable future 🙂
    Can’t wait to see your AI in Action as the current Themepark MMO’s really are getting old.

  4. Xanoxis Do'Urden says:

    Good luck with Hero’s Song, I will keep my eye on it.

    Good AI and fluid storymaking was what drew my attention in EQ Next, and I’m dissapointed its not being made, but this gives me some hope. Make good game, sell very well, and make your own Everquest Next in new company! That would be great 🙂 Remember to include cool tiger race.

  5. Oakenfal says:

    The AI was the meat and potatoes of EQN. Guess they will not be pioneers.

    The Pixelmage team is impressive. Patrick Rothfuss – World builder and design…awesome.

    I can’t wait to experience the AI.

  6. Interesting to learn you’re working on Hero’s Song as well. I have to admit that I hadn’t given the title a fair shot simply based on a distaste for pixel art games, but its impossible for strength of ideas to surpass that. If nothing else, I hope it’s a massive success so that MMOs with modern graphics will inevitably steal all of its ideas. 😛

    Thanks for sharing your thoughts Dave, I consider this to be the one must read post-mortem discussion of EQ:N so far.

  7. Eric says:

    I just went through your presentation and am very interested in prototyping some of these ideas in my indie strategy game. May I use what I saw in the video and just credit you or do you need to approve the game first, etc. ? Not sure how that part works, but I definitely want to implement the organic flee component and when people ask about it (if any) direct them to your site.

    • Dave Mark says:

      If I didn’t want people to use it, I wouldn’t have spoken about it at a public conference!

      Which lecture are you referring to, by the way? The AI in MMOs one from 2009 or the one on GuildWars 2?

  8. Karnex says:

    That was a very interesting read. After the cancellation of EQN, I gave up on MMO. But after reading this, I think EQN was trying to do a change too big, and players are not ready for it. Decades of following yellow exclamation mark is a hard habit to break. But I would love to see this dream come true. There are many talented people in the community, and I am sure with something similar to the open development process that EQN had promised, we can make it happen (just look at github for what open development can do). I am so glad the idea is not dead. All the best on your future endeavors

  9. TrueGamer says:

    I don’t think it’s the game development companies that need to be convinced of anything. It’s the publishers and parent companies for those game companies that need convincing in order to spend a nickel on anything with a risk factor tied to it. As you know, everything is about money anymore. Corporations don’t care about their customers, not really. They care about acquiring their customer’s money. So while the people in the development trenches may be trying to make fun and engaging games in order for people to have fun, their superiors don’t care about that. All they want is something that will make people reach for their wallets. You may think the two are the same, but I don’t.

    I had written off Smedley and the retro-looking Hero’s Song. I still don’t think a 16-bit 2D scroller has much chance for true success and there will need to be something VERY fun and special about the game to convince me to buy it/play it. In any case, it will not be Everquest Next. And Smedley will fuck it up like he’s known to do.

  10. Leopardpm says:

    Dave, I have been in contact with you (slightly, through forum posts and purchased your Behavioral Mathematics book) and have always been impressed with your depth of knowledge and manner. I just viewed the GDC-15 video on your AI component system (couldn’t find an actual name for the conglomeration of all four major components, I will call it ‘AI System’ further down) and it is brilliant. My question to you is that I have been of the mind that some version of GOAP would be beneficial to creating truly dynamic and emergent AI, and in our past conversations, you seemed to believe that such planning was also important. But, in your AI System, I don’t exactly see how it would mesh with any sort of a planner – in fact, it appears to focus on the ‘immediate’ action(goal), and not so much on any sort of series of actions to accomplish tasks/behaviors. Did I miss something? Or are you planning on expanding into planning in the future? just curious your thoughts here…

    Also, regarding your AI System – is this going to be something that you will wrap up nice and neat in a package that can be sold? I appears that you have developed a general enough, easy to use system ta\hat can be used with a variety of different genres and would be beneficial to many developers, including (and especially) Indie developers. Your thoughts?

    Again, thank you – I have been waiting for a truly dynamic world, driven by autonomous AI, ever since Ultima Online, and you are my one remaining ray of hope that this will be accomplished in the near future!

    • Leopardpm says:

      sorry I missed the name, thought this was the name of one of the four components: Infinite Axis Utility System (IAUS) – also wanted to say your interface looks good and easy to use… I bet both AI programmers and Designers are salivating to se your system in their projects!

  11. Dr Shabbs says:

    Fuck that shit. Yall can rationalize it all you want. You just dropped the ball. even if you coudnt get the ai to work exactly as you wanted. it still could have been something……..you just had the wrong project leaders who didnt know what the hell to do with it. It was an amazing concept and i think it could have been DONE if all of the companies resources were put into it, instead they branched out into a half assed me too zombie game. and other things……..Daybreak fucked this up. and sounds to me like once again instead of standing up and saying this COULD be something big, we may not get everything we want into it right away but we could still TRY.. nope. everyone LET it die. thats the thing. you gave up on it. yall gave up on the game way too soon. Landmark should have never happened it was a waste of resources and it distracted the team from focusing on Next. i think thats what kiilled it.

  12. Ana Julia says:

    That was a very interesting read. After the cancellation of EQN, I gave up on MMO. But after reading this, I think EQN was trying to do a change too big, and players are not ready for it. Decades of following yellow exclamation mark is a hard habit to break

Leave a Reply to Ethan "Isarii" Macfie