Upcoming Events! Community Event Calendar

Around the Verse: Engineering Intelligence Written Thursday 26th of January 2017 at 01:03pm by CanadianSyrup, Sunjammer and Desmarius, StormyWinters

It's Thursday which means Around the 'Verse. Check out what CIG have been up to.

As per usual, anything said during the show is subject to change by CIG and may not always be accurate at the time of posting. Also any mistakes you see that I may have missed, please let me know so I can correct them. Enjoy the show!

TL;DR (Too Long; Didn't Read)


  • Chris Roberts will be in Europe for the next couple weeks.
  • Martin Whip from Zurich, Switzerland sent some goodies to the team in the form of chocolate and Swiss knives.
  • the 2.6.1. production schedule is up on the RSI site and will update every Friday till its release.

Studio Update

  • Team has been busy with 2.6.1 which will feature fixes, balance passes and more.
  • Misc Razor is currently in the white box stage with Joe Navel
  • Drake Caterpillar and Herald are undergoing VFX polish and optimisation.
  • Surface Outposts are starting to come online now and designers are eager to begin placing them on planets.

Star Citizen: AI

[6:03] Q: What do you oversee at CIG?

  • Tony Zurovec focuses on the architecture - basic capabilities, how it is exposed - this includes working on the editor the designers will use.
  • Francesco Roccucci takes the architecture and implements it. He manages the AI team consisting of Andrega Boni and Mario Sarini in Frankfurt and Rich Walsh in the UK.

[6:51] Q: Current state of AI in Star Citizen

  • They’ve almost completed the initial tier of architectural functionality. Francesco now needs to work with the team to get it working and they’re aiming in the near term to expose functionality to designers for mission creation.
  • Missions will look like what we’ve seen but they will be created quicker and will have an algorithmic component.
    • Locations tend to be fixed right now, this will change.
    • Missions won’t be the same every time - a stranded pilot might be friendly or hostile, alone or with a group, healthy or needing medical attention.
  • There are lots of ways to take a simple mission and make lots of experiences out of it.
  • Francesco is happy with the architecture of the code - it is architected to scale very well. It will be much more fluid and quicker to add stuff.
  • They are discussing the last bits that have been added - reusable pieces of logic - it’s been much easier to add now and everything feels like it’s getting quicker.
  • They are still working to try to create “object-oriented content”.
    • Large missions are broken down into component parts that can be reused. With an initial library of functionality, designers are able to craft stuff much more quickly and do more interesting, varied things.

[12:22] Q: Subsumption

  • First part of subsumption will be the mission system. It’s more straightforward as their isn’t any animation overlap.
  • Animation isn’t typically considered with AI but it’s important as AI needs to be able to anticipate animations to look fluid. A smart AI can still look stupid if it doesn’t work well with the animations. They’ve had conversations about whether they might do a small iterative release of this.
  • Major issue with the Vertical Slice was animations integration: heads orientating incorrectly, etc. A lot of the AI logic was actually operational: activities, sub-activities, secondary sub-activities. They wanted the animation to do justice to the system and the developers.

[17:42] Q: 3.0 NPC AI

  • 3.0 will start make landing zones feel alive: different characters with their own interests and behaviours.
  • The first iteration of combat will include coordinations between AIs e.g. covering fire
  • Players will now encounter AI ships in space and can do more with them

[21:14] Q: For NPC ships, are those under the control of a single AI or will it be multiple AI working together? What are the challenges of this?

  • Multi-crew ships will operate a similarly to the chain of command you'd expect on a vessel. It will be at a high level rather than micromanagement: it's not going to be "Sim Citizen"
  • It will be easy to integrate players into roles otherwise controlled by NPCs. You'll be able to customise as much or as little as you want
  • Moving away from the ship to a character in a seat having behaviour. More realistic: if you kill the pilot the ship doesn't fly automatically
  • You'll be able to buy AI modules (computers) and assign them to some seats. Not all roles, e.g. medical officer, can be substituted

[24:24] Q: What advantages/disadvantages would there be in running an AI-crew versus an all-human crew?

  • An all human crew won't be definitively better: it will depend on their skill
  • How much you are willing to pay for NPCs will determine their skill
  • Players will probably do the roles they are good at or interested in and hire whatever NPCs they can afford for the remainder
  • NPCs will have characteristics that will govern how they deal with tasks
  • Player strategy on how crew assignments are made and what missions are taken will determine their success and if and how their NPCs will improve and learn adding another level of game customization

[27:57] Q: What do you anticipate to be the largest ship capable of operating effectively with an AI crew?

  • Larger ships will require larger crews though it can be mediated by smaller crews with higher skills, but there is no hard coded limits
  • Higher echelon ships will probably require more experience. Assigning a low level quality of crew member to a task would lead to bad decisions and buckling under pressure as you would see in the real world

[29:48] Q: What kind of relationships do you expect us to develop with AI NPCs? Will an NPC react differently to me based on prior actions? If I ruin a relationship with an NPC, will there be ways to repair it afterward?

  • CIG's intention is for the player to form long lasting relationships with all of the in-game characters. Players will develop reputation in a number of different fields that will determine how NPCs will respond to you
  • One to one relationships will be based off your personal interaction and history with said character. As characters get more comfortable with you they will act differently
  • You must continue to perform job/tasks to maintain your reputational status within the community for that job, if you discontinue your reputation will gradually decrease back to baseline
  • Future AI will probably head more towards algorithmic but not purely algorithmic(ability to algorithmically create variations of things)
  • They’ve been pursuing the ability to create all these handcrafted items for awhile so designers can build on top of that and link those pieces together
  • Further iterations giving variety to where choosing the right weapons for the scenario will be vital.
  • Long term, having encounters with AI becoming more meaningful and using AI you’ve encountered as a baseline for future ones. Say you encounter a pirate while hauling and he beats you up pretty good, they could then take that pirates bio, and create scenarios down the road that you may encounter him again in different ways so that you carve out your own little universe of connections amidst everyone else’s.

[39:20] Final Thoughts

  • Adding wildlife is going to be exciting because that’s what can really bring a planet to life and having different kinds of wildlife from primitive frogs, to animals that you’d expect intelligence out of.
  • Getting 3.0 out the door and watching how players interact with it is something they can’t wait to see. They’re curious how players will interact with the various systems and the data they’ll acquire from numbers as well as using the spectator camera to see in person what the player is doing.

Full Transcript

Intro With Chris Roberts (CEO, Director of Star Citizen and Squadron 42), Sandi Gardiner (VP of Marketing). Timestamped Link.

Sandi Gardiner (SG): Hey everyone and welcome to Around the Verse. I’m Sandi Gardiner and with me is Star Citizens’ Game Director, Chris Roberts.

Chris Roberts (CR): Hey Sandi. So thrilled to be here of course. I’m actually going to be heading off to Europe this Sunday, so I’ll be there for a few weeks so you’ll have other hosts here with Sandi going forward for a few weeks until I come back.

SG: Speaking of Europe.

CR: Speaking of Europe look what we got from a fan in Switzerland. He’s sent us lots of sugar obviously. Swiss Toblerones, Swiss Caramels, Swiss knife.

SG: From Zurich, Switzerland, Martin Whip.

CR: Yeah and sent us a t-shirt from his organisation, thank you very much.

SG: Check this out, bold.

CR: Pretty, I’m not that brave to actually get a real Star Citizen tatoo on my arm.

SG: I’m not either, well done.

CR: Well done, very impressive and thank you very much Martin. Anyway back to the show. Today’s show is going to be about something very dear and near to my heart, the game's AI.

SG: That’s right we’ve got Tony Zurovec and Francesco to sit down and answer a few questions about how players will be interacting with NPCs in the game, but that will be coming up later in the show.

CR: Yeah, first up we have our weekly studio update from Nick Elms and our U.K. team.

Studio Update With Nick Elms (Creative Director). Timestamped Link.

Nick Elms (NE): Hi everyone, welcome to another Studio Update from here in the U.K. My name’s Nick Elms, I’m the Creative Director here at Foundry 42. As you might have imagined since the 2.6 release we’ve been very busy trying to get 2.6.1 out for you. There’s been a lot of bug fixing, a lot of feedback that we’ve taken on board from you guys and John Crewe and Andy Nicholson have been looking at incorporating a balance pass now with all the feedback we’ve been gathering from you guys in the community from you playing 2.6. They’re making a comprehensive review that hopefully we’re going to be getting into 2.6.1.

We’re going to look at the Misc Razor next which I think you saw in November for the concept sale. It’s now with the ship team and in particular Joe Navel who’s working on it through its White Box stage. 

As part of the ongoing VFX polish and optimization, we’re going to take a look at the Drake Caterpillar. We’ve been working on the throttle up and down engine effects and with the Herald we’ve been working on the explosion effects. This is the kind of stuff that we send around the office on a daily basis and we thought we would share it would you guys so you can get an idea of what we see. As you’ll see the VFX guys have shared there’s a Vroll of all the ship weapons firing simultaneously. 

Surface outposts is what we’re going to take a look at next which I think we’ve released a couple of snapshots of, but it’s starting to come online a lot further now for us in the form of the building sets that were starting to get through from the artists. The designers are particularly excited about all the gameplay opportunities that these will afford as they can actually lay them out on the planet surfaces now for you guys to investigate. 

That wraps up for another U.K. update. As I mentioned earlier there’s a lot of work still going on behind the scenes with bug fixing and optimisations and feedback, etc. All that remains for me to say is back to L.A. and see you in the verse.

Back to Studio With Chris Roberts (CEO, Director of Star Citizen and Squadron 42), Sandi Gardiner (VP of Marketing). Timestamped Link.

CR: Appreciate the update Nick. Excited to see where the building sets are headed, I mean that's fantastic, a lot of great stuff in there.

SG: Definitely, no kidding, and the work the effects team have been doing is really impressive. Especially the Herald explosion, that was very cool for me, and as we’re seeing all the ship weapons lined up like that.

CR: Yup definitely seeing things shoot and blow up is always very cool. Nick also mentioned all the work going into 2.6.1. If you want to learn more about the up and coming patch make sure to head to the website and look at our full production schedule to get a full sense of what you can expect. We’ll be updating the schedule every Friday until the release so you can track our progress via our website.

SG: And of course a lot of the changes and rebalancing that we’re doing right now are thanks to the fantastic testing that our backers have been doing since 2.6 went live so thank you to all the backers out there.

SG: Absolutely. Your feedback is invaluable to the project so I want to encourage all backers to keep playing Star Marine, and Arena Commander, and Crusader, and to keep letting us know on the forums, Issue Council, as well as on the Spectrum. Please if you haven’t tried out Spectrum, try Spectrum,  It’s beta because we’ll be going live with that soon. It’ll help 2.6.1. and 2.6.2. and after that all the subsequent patches to be as good as they can be.

SG: And as many of you know from our newsletter we recently had many of the directors and leads over at the L.A. office for several weeks of discussions. While that was going on we took advantage of everyone being here to record some deeper conversations about what we’re actively working on.

CR: Yup absolutely, and as the game's AI system is one of the most complicated and ground breaking features that were currently in development on Star Citizen, we figured it would be a great opportunity to have lead AI programmer, Francesco Roccucci and Persistent Universe Director, Tony Zurovec gave an extended discussion about the systems  they’re building.

Star Citizen: AI With Francesco Roccucci (Lead AI Programmer), Tony Zurovec (Persistent Universe Director). Timestamped Link.

Tony Zurovec (TZ): I’m Tony Zurovec, I’m the Director of the Persistent Universe for Star Citizen.

Francesco Roccucci (FR): So I am Francesco Roccucci and I am the Lead AI Programmer for Star Citizen, so yeah, we work a lot together.

TZ: Yeah.

[6:03] What do you oversee at CIG?

TZ: I’m predominantly focused on the architectural side of the stuff - what the basic capabilities are, how we expose that functionality to the designers so that we can get as much leverage from the development process as possible - this includes working on the actual- developing the editor itself which was the designer’s interface.

FR: Well I think about all the architecture from Tony and I make it real in the code, basically. We develop- I put the code- I architect the code and we implement it and then I manage other three people so far on the AI team - so we have Andrega Boni, Mario Sarini, Rich Walsh in the UK studio - so we are a bit in Germany and a bit in UK.

[6:51] AI in Star Citizen: An Update

TZ: The current state of AI in Star Citizen is that we’re basically - I would say we’re cresting the hill with regard to the initial tier of architectural functionality. Francesco has now got a lot of work to do to basically take all of that, distribute it to his guys, get it working within the game and what we’re aiming to do within the fairly near term is to be able to- for the first, in significant fashion to be able to start exposing this functionality to the designers and they should have enough of the base capability to start crafting actual missions within the game.

It’s going to look, to some degree, like the missions you’ve seen before except that they should be able to create them much more quickly which means we won’t be talking about 17 missions - they’ll be able to create- you’ll be talking about much larger potential number of missions - you’ll be talking about a lot larger of number of different things you’ll be trying- problems that you’ll be able to present to the player - you’ll be talking about a lot more content that’s going to have an algorithmic component to it.

For example, right now, locations tend to be fixed - and we’re going to be introducing in the not-too-distant future the ability to go down to planets and stuff and we don’t want you to always encounter a derelict space ship with a stranded pilot that needs to be transported to here always in the exact same location. When you go to pick up that stranded pilot - we don’t want him to always necessarily be friendly; when you go to pick him up, we don’t always want him to be alone - sometimes it’s going to be an ambush, sometimes he’s going to need medical attention.

These are things, all just in the course of that one basic scenario - which is “distress signal, crashed pilot, etcetera” - there is a world of opportunity for us to create a range of potential experiences for the player and this is what- as soon as basically- Francesco’s basically integrated the rest of this and I’m sure there is-

FR: No pressure on me.

TZ: Little bits and pieces as we start to mock-up the first tier of the subsumption missions, but as soon as we’ve got that then we’ll finally start to get a sense as to what the designers can create with this.

FR: I think it’s always like this first step, right? Of getting the core-

TZ: Yep.

FR: -there in a state in which you’re not only happy with the architect of the code as well - because for me it’s a lot, I know I can rely on the vision of Tony for that for the architecture of the design and all the other designers that we have - all very talented people and the code as well has to be architected in a way you can scale as much as possible.

So like, you know the big work we’ve done so far is to make sure that subsumption runs and everything with acceptable amount of memory usage and CPU usage. We have all these concepts of activities that are actually templates that can be re-used between a lot of characters without creating like instance everywhere or having the mission system as well, reuse as much as possible from the subsumption architecture and so even after we create the same logic all over and over again in different systems because it uses different systems and it’s also accessing content from one actor to the mission and passing formation.

It just becomes very much quicker, you don’t have to go through different DLLs - each DLL can expose different functionality without you- don’t have to add code to be doing different- complete different structure and then how do I expose this part to this other system - it just becomes everything much more fluid and much quicker to add stuff. So yeah, that sometimes it takes a bit of time but then I think we are starting to arrive to the point where we are very happy of that.

We’re discussing the last bits that you added - for example, this concept of adding some reusable piece of logic that you can recall and for us it’s very easy now to implement that and every new functionality seems it’s always quicker to add and you can start of feel to say “We’re doing a good job here” and I don’t like to say it, myself, but I think it’s good when you are happy of the work you are doing.

TZ: That’s something I referred to before, which is- we’re very much focused and always have been with regards to the direction we’re going with the concept of this “object-oriented content creation” which is to basically distill all of the different pieces of functionality that we’re going to use to build up these larger missions to break them down into all these component parts and then just like you do in an object-oriented language like C++ to basically allow us to pass inputs in, to extract outputs out so that we can customize them dynamically in the scope of a much larger problem.

And so what this is going to wind up allowing us to do is after we’ve got that initial library of functionality - designers are going to be able to craft this stuff much, much more quickly than they’ve been able to do in the past and what this effectively means to the player is a much larger amount of interesting things to see and experience and challenges to face and solve within the game world.

[12:22] Subsumption

TZ: The first part of subsumption - the easiest part of subsumption to basically expose is actually going to be the mission side of it - the mission side is basically built on top of the, what you want to call the lower level AI layer, there’s a lot of commonality in terms of how we do message passing, the data structures, the concept of functions, reasonable parameters and all this kind of stuff, task archetypes with which we build up the solutions to these problems - the mission stuff in general is a bit more straightforward to expose because it doesn’t directly control any of the animation stuff.

And the animation when you’re talking about a modern 3D game - animation, even though it’s not technically part of what most people would call the AI, it’s plays a massive role with regards to how people perceive it - if you have a bartender and he’s going about doing his business but the path find is-

FR: It’s not smooth or-

TZ: -the path finding doesn’t often very well, he goes right up to other characters and then goes 90 degree angles around them as opposed to doing nice splines - if he basically doesn’t align himself precisely, if things don’t snap right to his hand - if he can’t actually operate things in believable fashion then he doesn’t actually look intelligent, even though under the scenes he’s still able to search for this - he can still basically go over to it, he can still utilize, he can still create items, he can still send messages, received messages, query the environment - all that type of stuff and so one of the big issues that we face before we really expose the AI side of it is we want to make sure that a lot of these - you could call them cosmetic issues because they really predominantly isolated from the AI itself although the solution to them is occasionally quite complex because they are still so tightly integrated.

By way of example, one we’ve talked about before is you’ve got a path find and in combat you’ve got a character and he’s running towards cover and he actually needs to know in the path find that he’s running towards cover in a combat situation and then he’s basically going to modify the end part of that path find he’s running to that destination by sliding into cover and if you’re talking about a character that’s actually going to manipulate something or sit in a chair, they need to know how they are going to need to be aligned so that while the path find - the predecessor step an action, it can be taking that into account so that because- what you don’t want is for them to get to the endpoint and then spin 80 degrees and then basically to play the appropriate animation and so the mission stuff is the first and easiest but given the release schedule and we can’t talk too much about that - we’ve had some conversations about whether or not we might have a small iterative release that might expose some of this for testing purposes but we’ll just have to talk about that as we announce that later.

FR: Subsumption is really global, general name so I think we’re probably- when the mission system is out, well it’s part of subsumption so there will be a part to experience what subsumption is in general, right? But, yes, as Tony says, most of the animation issues are the one that. we are basically working on to … before we can show something. Or not too much before we “can” show but before we want to show ...

TZ: That’s exactly it.

FR: … because I think that’s the critical part. It’s like for me - I think we are both very critical to the work we do - so it’s like always say “ah, this doesn’t look very good: I don’t want to show it”.

TZ: And this was one of the big things with the Squadron 42 Vertical Slice to where one of the largest issues was the animation integration …

FR: Yeah.

TZ: … into the system. As opposed to whether or not you were, under the hood, actually doing all of the appropriate AI logic. The AI logic, a lot of it was …

FR: It’s … yeah, really a good point ...

TZ: … actually operational.

FR: … because the Subsumption logic … it’s already … we have … in our game, our builds, we already use the activities. We already use sub-activities. We already have secondary sub-activities surrounding people say “Hi” to you or greet you, seeing from far, having different activities ...

TZ: Right but we were having problems with the heads weren’t orientating correctly and ...

FR: But … exactly … if you showed something like that ...

TZ: … to any user you show that …

FR:  … yeah.

TZ: … it looks much more primitive than that it really is. And so that was one of the reasons why, up until the very end, we were like “Well how … how quickly can we tie up some of the most egregious animations issues?”

FR: Because you gonna make sure … you want to make sure something that can be very impressive, it actually is. Because otherwise if you ruin the look of something that is actually quite complex, quite advanced, with a bad look then it’s hard to remove that type of first impression. And it’s not smart and it’s like … I don’t think it gives the right credit to the system

TZ: We … we … we

FR: … and the developers that worked on it.

TZ: … we want the quality of the animation to match the ambition …

FR: The system. Yeah.

TZ: … of the actual logic …

FR: The game.

TZ: … and that’s still going to require … that’s still going to require some effort.

[17:42] 3.0 NPC AI

TZ: 3.0 I think we’re actually going to aim for a pretty diverse set of things. Most obviously would just be starting to make the landing zones actually start to feel alive, so that you encounter a variety of different characters all pursuing their own particular set of interests. This will include everything from the prototypical bartender and bar patrons, tourists, vandals, businessmen…

FR: Shop.

TZ … yeah, various shop … shop keepers ... muggers, criminal elements: things of that sort. And then some of the … some of the more interesting stuff in terms of the gameplay potential, I would say, would be the combat aspects.

FR: Because the combat is, of course, is what I really worked on and it’s something I’m really, really happy to work on and I want to push it even further. That is going to be interesting because we try to make this first iteration on it where there is already coordination between the guys that fight. People that can advance while somebody else covers the advancement from a cover spot or if no cover spot are allowed they will just …

I think the main thing with the combat that I’m really trying to push is the fact that they understand the environment. And you are sure. And you understand that what they are doing makes sense for the type of environment which they are in. And that is going to be what I think is cool in this scenario because maybe you’ll find some pirates on the side you want to get a good deal but then they turn on you. And then the fight makes interesting … the situations …

TZ: And this will play into a number of things so that when you encounter … like right now the only ships, as you’re just generally wandering around stuff, you tend to encounter is really other players.

TFR: Yeah.

TZ: And that should be dramatically different in 3.0. Not only different in terms of you encountering them but in terms of what you can do with them. If you disable … if you’re transporting cargo and you basically get attacked by a pirate and he disables your engines then it’s entirely possible that he’s actually going to try to board your ship.

Or vice versa if you’re a mercenary … if you’re a pirate and you’re trying to take out a freighter then part of what you’re going to have to do is - after you disable the ship - is board it and those characters will actually be able to fight …

FR: Fight back.

TZ: … fight back within the confines of their ship. And this is, again, just going to open up a world of possibilities in terms of making the world, space, etc. feel much more alive.

FR: Kind of unique as well. Because I think it creates these kinds of scenarios that are not very easy to find somewhere else. It’s going to be a very interesting to see how the players will play.

[20:36] Back In Studio

CR: Thanks guys. Hopefully that gives you a better sense of the progress we’re making and gives you an idea of how awesome the final version can be. As more and more of the features like the AI and the mission system, locomotion for player and AI come online and are implemented into the game we’re able to unlock even more opportunities to push the game even further and create that first person universe that Star Citizen is going to be.

SG: And we know that AI has been one of our most asked about topics from the community so we decided to open up the conversation to the community and let Tony and Francesco field some of your questions about how it works.

[21:14] AI - Fan Q&A with Tony Zurovec & Francesco Roccucci

Q: For NPC ships, are those under the control of a single AI or will it be multiple AI working together? What are the challenges of this?

TZ: Multi-crew ships will operate fairly similarly to the hierarchy ... the chain of command that you’d see in a typical civilian or military vessel in that you’ll have a captain and the captain will tell the navigator where he wants to go. And the captain will basically tell the engineer what to prioritise: whether he wants … needs the weapon systems back online first or the shield generators or the engines. And he’ll dispatch the chief medical officer to deal with medical problems in this area or that area.

So it’s at a very high level. I’ve said before we don’t want to be “Sim Citizen”. We don’t want to have too much micromanagement in terms of control. That’s not the type of game we are. But the concept of being able to just generally prioritise things and then to have your lieutenant characters address those.

And this actually segues nicely into how players are going to be integrated into the equation. If you’ve got a captain and you’ve got a navigator and you’ve got a chief medical officer and you’ve got a chief engineer and you’ve got a chief weapons officer, and all this other stuff and then all of a sudden the player wants to … one of your player friends wants to deal with weapons or deal with engineering then you’ll have  a very easy means of factoring in how the other characters are going to interface with them.

And what you’re going to wind up getting out of this is a nice concise system to where you can customise as much or as little as you want. You can have … you can recruit NPC characters to do as much of the work, that you and your friends don’t want to do, as is necessary to accomplish the job.

FR: Yeah, gets you really moving away from one behaviour for one ship. So it’s really like each seat will be controlled by the people - just one - to work on that seat. So I think there would be a really the crew feeling working on the spaceship as Tony already said. I think that is the only way to give a bit more realism in the sense like even if you board a ship and the pilot is tries to still bring the ship somewhere and you kill him then you don’t want the spaceship has this external behaviour still like drives automatically, controls the turrets anyway. You just want to really see that something is there.

Another thing we also try is to make these type of AI modules that you can buy and assign to specific seats. So you don’t have an actual crew member sometime but you want to have a small computer at that controls, turret for example. So some stuff, maybe the medical officer, cannot be substituted by computer but the turret can have a controller … automatic controller. And if you just sit down then you can immediately take control over things turn on and off. And I think you as a player you are also able to customise a bit how the ships automatically responds to things. Can be really cool because you can have just like a combination of … that you can have that already makes a huge difference between one ship and another.

[24:24]Q: What advantages/disadvantages would there be in running an AI-crew versus an all-human crew?

TZ: The advantages of AI crew versus human crew? I wouldn’t really say that there’s definative advantages. I would say the devil’s always in the details. What you will have as a player is the ability to recruit NPC characters to fulfil specific jobs on your ship. But when you basically go to recruit them, that’s where the details start to matter. How much are you willing to pay? And that’s going to determine the level of quality, the level of professionalism, that you’re going to get from that particular member.

If you want, a turret gunner that’s really effective then you’re going to have to pay a higher monthly salary. You’re going to have a higher running cost, and so what you’ll see a lot of players... and the exact same scenario exists within players: there’s some will be more adept at a given occupation - a given skill - others will be less capable, and so what I think you’re going to wind up seeing in a lot of cases is players -  you may have a ship that’s ideally crewed by 6 or 8 or 10 people - and the player’s party will take on those roles that they’re best at and then they’ll recruit NPCs to whatever skill level they can actually afford to fill in the remainder.

FR: Also about the interest right? Because maybe some players have interest on piloting, some others have really good interest in shooting. And then you can just use the AI to fit the roles you don’t like too much. But on the other hand the AI won’t rage quit on you, right? You won’t have like a player disconnecting, so.

TZ: Yeah, but they

FR: You have an an advantage with AI? ...

TZ: Well and that's true. All of the whereas most of the things that players will actually do will depend upon their actual capabilities, or their ship capabilities, or their equipment's capabilities to expose things. But how, how well you basically can fire, for example on another player, your accuracy, that's determined by your physical dexterity, and it's influenced by the weapon which may have more or less kickback or a wider area of effect in terms of it's damage, etc. And NPCs will have a variety of characteristics that will model how well they deal with all of these different . . with all of these different things that the player has to . . has to handle themselves.

FR: Yeah the skills will be like definitely be a huge part you know on deciding like this NPC is good at doing every job or jobs. But maybe you have two that are kind of good at the same job, and you still need to fill another role and you put the one. They say, “Well, I still need the shield management.”, and even if none of them is pretty good but maybe after a certain amount of missions. Well he starts to get used to that, and he starts to improve his skills as well. You know you can decide, “Will pay a bit less, but I risk a bit more in the beginning at trying to make simple missions so they get used to these things. I think that is also like interesting to see how the player will use their strategy to put balance their hiring and the other NPC they have with how they can improve and learn.

TZ: Yeah, it becomes another resource to manage. Another way by which players can differentiate their style of gameplay

FR: Yeah.

TZ: versus another. Well it's like do you go with three engineers of lower quality or one really stellar performer.

FR: Yeah.

TZ: So there's going to be a multitude of ways for you to customize the game-play experience via the choices you make in all of these different areas.

[27:57] Q: What do you anticipate to be the largest ship capable of operating effectively with an AI crew?

TZ: There's no real limitation just given the distribution of responsibilities. It would really just be more along the lines of the larger ships will require larger crews to operate effectively and therefore their running costs on average will tend to be higher than crew . . than ships that can be operated with, with a much smaller crew. Obviously you can modulate that to some degree with the quality of the crew you're hiring . .

FR: Yeah.

TZ: . . and things of that sort, but there's no hard coded limit and there's surely no reason why there would be given how all of these tasks and responsibilities that they're going to be executing are going to be ready.

FR: I think that is really like coming from you know to the, to the fact that basically within the individual behavior you know are controlled by a captain or something like that, but then each of these NPCs which is do it's job. So you know there is no real limit on the amount of people that execute their job if you know you have like five, six to control five different turrets. Well, you know, that would be just like five different people. Their coordination of course depends on the amount of targets they have. Who they can target, but then it doesn't matter like if there are three or five or fifteen, because you know the code would just give for that would try and distribute the targets, but so exact you know . .

TZ: But it would like as you go, as you go to upper echelon of ships you are going to probably need more experience . .

FR: Yeah.

TZ: . . better commanders, because they are effectively going to be . . in control of more . . of a larger number of crew members and we would probably . .

FR: Yeah.

TZ: . . just model it such that if you basically get a lower end . . a lower . . a lower quality of chief engineer starts to basically buckle under the pressure and he starts to make bad decisions, etc. .

FR: Yeah . . assigns wrong people . .

TZ: . . once he starts getting beyond three or four or five guys that he is controlling. The exact same thing you'd see in the real world.

[29:48] Q: What kind of relationships do you expect us to develop with AI NPCs? Will an NPC react differently to me based on prior actions? If I ruin a relationship with an NPC, will there be ways to repair it afterward?

TZ: It's always been our intention to allow the player to form long lasting relationships with all of the characters in the game. What this means specifically is that you will develop what we call reputation in a number of different fields. Are you known for your acts of piracy? Are you known for your humanitarian missions? You're basically very medically oriented and you are rescuing people, healing people, solving medical crisis, you know etc. Do you . . are you really well known for how effective a transporter you are in terms of commodity goods and things of that sort? You have a high very . . a very high success rate in terms of getting it done. So there will be those reputational aspects that characters will be able to . . you know of yourself . . that characters will be able to query to determine how much they like you, dislike you, whether they want to offer you the more advanced missions or just the starter missions, what they are willing to pay you in order to do that, etc.

There will also be these what I call these many to one relationships to where it’s not based upon your, you know, piracy rating or your transportation rating… it’s based on your specifically one to one relationship. How do they feel about you? Did you take a specific mission from them? That’s probably going to be weighted much more significantly than stuff that they heard that you did for other characters in the universe. They will also be able to take into account your… the organizations to which you’re a part. Do they like that organization? Do they hate that organization? Are they neutral to that organization?

So all of these things together are basically variables that the characters within their conversation logic can factor into how they respond to you. So… and then we’ll wind up having these things gradually, you know, return to a baseline if you don’t keep doing it. Other words if you were an exceptional… exceptionally good at transporting merchandise, very high success rate. If you haven’t done that in a day, in a week, in a month… then people will no longer be talking about that behind the scenes, therefore you expect that reputational aspect to gradually deteriorate over time and you’re going to have to just… you’re going to have to... basically going to have to keep doing that type of thing and keep doing it well if you want to maintain your standing with the community in that regard.

FR: Yeah, so like for us for me especially now on the AI team what we try to do is I test all of  these functionalities to make sure that the AI can create as much content as they can. So like, you know, once we have these things, the relationship variables or attributes it is really like better to give these to the designers and then the designers can use any of these. Can be the reputation, can be this… you know, what they feel about you and they can just create all the logic that they want. Plus we have… we’re kinda in a unique position because like the game can constantly be updated so these characters can also get new responses. You know, based on the state and, you know, and all the result is new content that we can create with time that just makes, you know, the relation always more smooth and you know, we can go deeper into details and mentioned other stuff that you’ve done, mentioned events that have happened. If I have a good relation with you maybe I gossip about something with you and you know, we can create this and always have more and more and more.             

TZ: There’s the stuff that really matters in terms of the missions they give you whether they’ll talk to you at all and that stuff. Then there’s also just the more cosmetic stuff where they tell you jokes or they’re light hearted or they’re basically bantering on about their history because they feel comfortable with you because you’ve done three missions with them within the last week. Were they frowning at you as you go, were they repulsed by you, were they disgusted by you, you know. All that type of stuff, so I would you say you’ve got different elements coming together which is, you know, not just how they explicitly react to you but more just, you know, they’re just… they’re more subtle behavioural characteristics when, you know, when you’re in proximity.

[33:51] In conclusion, what are you most looking forward to?

TZ: For the future AI I think that clearly going to head more towards the algorithmic, you know, generation of content that does not mean by any stretch of the imagination… we’ve said from the very beginning, it’s like we don’t want to be pure algorithmic. We have no interest in being purely algorithmic, we don’t think that makes for a good gameplay experience. It just makes for good taglines in terms of, ‘ hey look, we’re giving you an arbitrarily large of this’ but it’s all, you know, it’s all bland and uninteresting. There’s nothing really, you know, handcrafted to draw you into it and so what we’ve been pursuing for years now is basically the ability to create all these handcrafted pieces and then to allow the designers to build on top of those to link those pieces together to create very quickly what, you know, what seems like incredibly customized missions but they’re able to do it in much more rapid fashion than what otherwise would be the case. So, I think what we’re going to wind up having is, you know, as you go forward is just the ability to algorithmically create more, you know, variations on those things. You’re not really going to notice it in terms of quality of the content, it's just going to look like… it’s going to look like you went from 12 designers dedicated to this to 24-36 as you basically refine all those algorithms and you start to get your base library functions and stuff all entirely operational.

FR: Yeah, the thing I think we can talk about as well is like the way… the thing I would always like to push forward is the kind of realism is if you want to start to follow one guy, for example, and we start with something like the guy will leave the planet but maybe, you know, eventually you have this guy like change planets, you know and he’ll be able to understand that there are connections between...you know, jumping into space ship, then take the spaceship to fly, quantum jump and then go to this other planet and then go down and search for another part. You can just follow it and then you can be always be more realistic like, you know, I think that is from a tech perspective, you know we’ll have of course some optimisation, what you’re trading, stuff like this that we want to have as quick as possible. So I think there are like these two parallel you know, work that we are doing both on the content production tools to create always more and give more feeling and all the technical problems that we are trying to solve like using objects while you use another object, you know sitting on a spaceship and while you are sitting you actually interact with all the objects and each these interactions triggers something else.

I think these of things that we are going to eventually have... it’s the details, it’s an amount of details that you will eventually have based off all the iteration. So first iteration will have, even on the combat side, first iteration will have a sort of coordination, but then these other iteration will always have more and more coordination using more type of weapons, being smart on which type of weapon is best in this specific environment or in a specific scenario. I think this is just going to give always more realistic feeling.

TZ: Yeah because another long term objective and this is not within the next 12 months it’s more of a 12-24 month window type of thing is basically support for what we refer to as the Persistent Dynamic NPCs and what we mean by that is, of course there will be specific characters at specific areas within even the Persistent Universe, certainly within Squadron 42, but we’ll have those characters even within the Persistent Universe, and then there will be another class of characters that are just created as you want around the world. You encounter, you’re basically transporting a load of commodities from one planet to a refinery and you encounter a pirate and that pirate winds up inflicting significant damage and you wind up having to call for help and once your help arrives you basically high tail it and get out of there.

This is the kind of thing to where what we want to eventually to be able to do is to keep the essence of what that character was, that pirate, what was his name, what does he look like? Was he talking to you on the radio, what voice were we using for him, what was his rank? What was the source of the interaction that you had with him. What’s your history? He basically got the better of you in a battle and then we want to be able to use that as an input into this library of mission content such that next time we’re looking for a pirate to basically be utilized on one of these mission archetypes, we have the possibility to insert him as the specific character and he then has the ability to refer back to that prior history and so what you wind up having is your own little custom version of the world to where these are characters we didn’t specifically go out and hand create and give a history to, but you’re encountering them, your interacting with them made them real, made them persistent for you. They’re also persistent for any party members that you’ve got, it’s like everybody can see them, but those particular characters that you’ve interacted with in a certain way become available for us to utilize when we need an equivalent character as you encounter lots of different scenarios within the game world.

[39:20] Final Thoughts

FR: Well I think that other things I also look forward is when the wildlife on the planets, that’s going to be really interesting because it’s something I worked on a little bit before. In Crysis we had some little wildlife and there’s always this big difference between the smart wildlife and the more simplistic wildlife. If you have a frog, well you don’t expect the frog to be super smart. You can create this kind of, you know the CryEngine word: Boids, animals that can move around and they react to your presence or your proximity much more. Then you have other things like pets or even just bigger animals like a deer or something you know that they’re smarter and they need to do something with you, they’re not as complex as a human, but they’re pretty complex and to give the actual feeling of planet, we’ll have to tackle all of these things and it's going to be very interesting.

TZ: Well, but it’s just like with the cities to where you want to make this world that you’re travelling through look as alive and as interesting as possible because if… like right now, i’m sure bothers us all is the just the fact that we’ve got a great big world, but we’re not yet able to effectively populate it with a lot of interesting material for the player to see, and we’ve talked in the past about how we’ve got 17 missions in the Persistent Universe that you can go and accomplish, but those were implemented in a completely different fashion than what we’ve been aiming to eventually be able to do, but these systems are such that you need to have maybe not the entirely, but a significant majority of the base functionality operational before you have anything and so what makes me most excited about 2017 is just the fact that we should start get this stuff that we’ve been talking about on the content generation side finally into the game and this is where all of a sudden the ability for players to start do a variety of different occupations and for them to just fly around, orbit around a moon, content we didn’t explicitly put there will be able to fill in, and then there will be interesting stories and things to do all littered throughout the entire system.

So I think that once we get 3.0, it’s supposed to be the initial iteration of pushing this stuff out there and then all the successive releases are just going to be refining, refining, basically taking advantage of this system we’ve put in place.

I think that Star Citizen in 3.0x iterations is going to finally start to look like a real finished game to where you can just go in and lose  yourself for hours and hours at a time and that’s not because you’re marveling at just the technological ambition of the project, it’s because there’s actually enough gameplay content in there to keep you interested.

FR: And for me it’s like, because after where we have that, we can actually see how the player plays with the game. The most interesting thing for me working on games is the fact that you produce something that people actually play with and they interact with, it’s not that you know you just give it and it’s not that only, it’s really how you use it, it’s how you play that, and the game plays with you so I think at that point you can also start to tweak things  say, “Ah this is how people like to do, and the way they like to play”, and then you can adjust things and give more content here and I think that is going to be very interesting.

TZ: I can foresee us using spectator mode quite a bit to actually see how players are interacting with all of this initial set of mission concept.

FR: “Stalking Mode”.

Outro With Chris Roberts (CEO, Director of Star Citizen and Squadron 42), Sandi Gardiner (VP of Marketing). Timestamped Link.

SG: Always great to hear more about how NPC crews are going to work. It sounds like it's been a huge undertaking to develop a system this complex.

CR: Yeah, no it’s definitely a huge undertaking and we’re still hard at work at it, especially Francesco and Tony. We really want our NPCs to have complex behaviors you would expect. So the work Tony, Francesco, and the rest of the AI team have been doing is vital to our plan moving forward in creating this immersive Star Citizen, and Squadron 42 first person universe. So once we have the foundation in place we’ll be able to tie it into the mission manager and the mission system itself and create some really emergent and immersive scenarios either by scripted design on the designer's side or through sort of, systemic or procedural generation that's in response to player or AI behavior. So I think it’s going to provide some really cool and unique gameplay.

SG: Cool, with that brings us to this episode of ATV, and as always we’d like to thank our subscribers for contributing to the creation of all our in depth behind the scenes concept.

CR: Yup, thank you guys very much. I thank you for having patience for us being here and doing all of this and a huge thanks to all of our backers who are supporting Star Citizen's development. You guys are a big part of what makes the project unique and special and we definitely could not do it without you guys so thank you very, very much.

SG: We could not, and if you would like to hang out with more of the team make sure to tune in tomorrow at 10am Pacific for the latest Star Citizen Happy Hour stream to watch some live gameplay and discussion.

CR: Yes. I’m not quite sure which one of our devs will be there, but they’ll be there and it’s kind of a cool opportunity to have an informal Q&A and discussion, but until then.

SG: We will see you.

Both: Around the Verse.


Director of Transcripts

A polite Canadian who takes pride in making other peoples day brighter. He enjoys waffles with Maplesyrup, making delicious puns and striving for perfection in screaming at the T.V. during hockey games.



For whatever reason, this author doesn't have a bio yet. Maybe they're a mystery. Maybe they're an ALIEN. Maybe they're about to get a paddlin' for not filling one out. Who knows, it's a myyyyyyystery!



When he's not pun-ishing his patients or moderating Twitch streams, he's at Relay pun-ishing their patience as a member of the transcription staff. Otherwise, he can be found under a rock somewhere in deep East Texas listening to the dulcet tones of a banjo and pondering the meaning of life.

"If you can't do something smart, do something right." - Sheperd Book


Director of Fiction

Moonlighting as a writer in her spare time StormyWinters combines her passion for the written word and love of science fiction resulting in innumerable works of fiction. As the Director of Fiction, she works with a fantastic team of writers to bring you amazing stories that transport you to new places week after week.