Update #32: Meet the Developers - Steve Weatherly

[anek]

Q: What feature are you most looking forward to programming?

 

A:

I got into the game industry wanting to be an AI programmer. So, while I enjoy the challenges of engineering other systems, AI is where my heart is. I can't wait until we have a full complement of spells so I can write some cool wizard AI and give people an epic wizard battle!

 

 

Awesome!

 

I encountered too many a' wizard in the Infinity Engine games who wasted his entire Time Stop placing multiple redundant high-level spells on some wimpy summoned creature of mine that wouldn't have lived much longer anyways, while he could have easily killed off or disabled several party members instead.

 

...or stubbornly emptying her whole spell repertoire on the one party member that happens to be highly magic-resistant.

 

...or failing to put up a common on-demand protection spell that could have saved his life.

 

...or killing her own allies with area-of-effect spells.

 

But as other have mentioned, the biggest weakness of the Infinity Engine AI's was not their simplistic spell tactics, but rather their helplessness against cheese such as:

• setting traps around a blue-circled monster right in front of its eyes, without it turning hostile, or - after you attack it - avoiding the traps
  

• luring away and fighting a single enemy, while his buddies - although only a few feet away and thus clearly in hearing range - stay where they are, as long as no party member has entered their line of sight yet
  

• throwing a multi-round area-of-effect spell or a bunch of summoned creatures into a room and closing the doors
  

• hit-and-run, so the monster persistently follows the character that attacked it first, instead of giving up the chase and turning to more immediate threads when it is clear it isn't fast enough
  

Edited November 22, 2012 by anek

[Gyges]

It'll be cool to learn a little bit about each team member.

 

I'd just like to say that whiskey isn't good, even if I drink it at very rare occasions.

 

 

I'm more of a vodka man. Although I don't drink much at all anymore. I'll settle for a few beers. But if I had to sit around all days programming, I'd drink a lot too. (Had to do a bit of programming at university some years ago, but did not find it fun. I had more fun with 3D modelling.)

 

While were on it, you've got another vodka man here. Grey Goose and Beluga are the favorites with GG taking the first spot. I usually mix up something weird with lower tiers of vodka, but the good ones go down bare.

 

Also, yes. Gotto ask Sawyer what hes planning with weapon and armor variety. I found Dark Souls doing particularly good on this, with multiple, nearly perfectly balanced sets of end-game armors. Low variation in base weapon damage also made us consider other factors when building our character, such as stat scaling, unique abilities, weight, speed, moveset and elemental effects. This made for a huge variation of offensive styles to be available.

[MusedMoose]

This was a great read. I mostly play games like this for the story, lore, and characters, so I hadn't thought much about AI programming and all that goes with it, but this was really interesting. One of the things I really enjoyed about the BG games was getting into a tough fight and having to figure out how to beat it through something other than brute force. Now I know who to blame when that happens in P:E. And flails! Yay!

 

Also: I'm originally from southern California, so huzzah for Wahoo's Fish Tacos! And the office that goes out to coffee together works better together; I've experienced this firsthand. Would that you guys were hiring writers; I'd apply in a second even though I know my chances are slim.

 

Thanks for your time, Mr. Weatherly. Looking forward to more interviews of this sort.

Edited November 26, 2012 by MusedMoose

[Macs]

Great! I love this kind of developer who-is-who. Paradox did it in one of their Christmas calender specials and even though they are generally very close to the community, it was nice to have some faces and facts to put to the nicknames.

[Steve Weatherly]

Thanks for the answers!

 

So, you drink whiskey. Me too! What are you favorites?

Mine are Islay whiskys, preferably Lagavulin and Laphroaig. That stuff brings you straight to misty highlands and craggy cliff-sides when you drink it.

 

Do you know if there's whiskey or some cool booze in Project Eternity?

 

Is it easy to, I dunno, "implement" A.I.s in Unity? I mean, not the logic (because that is as challenging as you make it, as far as I know), but actually assigning an A.I. to a character. For instance, would it be possible to actually have a battle in which A.I. plays a huge role and the A.I. represents the actual challenge (and not the level/power of the enemy)?

 

Have you guys thought about naming Project Eternity yet? Has there been any talk around the office about it?

 

Also; watching someone programming and developing stuff is NOT boring. Markus Persson (Notch - Minecraft) did this stream and it was very nice to watch. And that was Java. I believe Doube Fine will attempt a similar behind-the-scenes/"watch us code" stream as well.

 

-I mentioned this above but the best whiskey I've had is The Balvenie Scotch whiskey, if you can drink it neat you know it's good.

 

-Haha, I'm sure there will be alcohol in there. I will demand a fine Orlan whiskey be included.

 

-Unity doesn't have any sort of built in AI system. I've created one for us and tried to make it stupid easy to give that AI to characters as they are created. It's about 3 mouse clicks right now. :D

 

-I don't think the final name has been seriously considered yet, I think think we're going to wait until the story is finalized and the lore more fleshed out.

[Steve Weatherly]

-The game systems are DEFINITELY easy to prototype with Unity, in that regard it is the best engine I've worked with. It's not the best thing ever though, whiskey is.

 

I assume this means that it's easier/faster to prototype with Unity than Onyx, so I'm going to base my question on that: some game developers, to my knowledge, have prototyped stuff on Unity and then moved to their own engines/licensed engines to work on the game past the prototype phase.. do you see Obsidian adopting this practice, or would you rather Onyx to be made more prototype-friendly? Considering Obsidian is small in terms of AAA development (but too big to be considered small or even mid-size), I'd think prototyping things fast would be a huge boon, but not being a developer (or even particularly knowledgeable in terms of tech) there might be a lot of aspects I'm not considering.

 

I can't imagine us ever prototyping things in Unity and then converting them to Onyx. Because the two engines are functionally very different it would mean doing everything twice, I don't know of very many people that like to do the same job twice. As we work with Onyx we work on maturing the pipelines to make building things faster and easier, generally if some process is cumbersome we try to fix the problem.

[Steve Weatherly]

Thank you for the interesting responses, Steve.

 

If I may, a general question I'd like to ask you is regarding AI: how much consideration do you think will be given to group AI by intelligent enemies? Will trained enemy soldiers attempt to fight in formation? Will they coordinate attacks so that, say, suppressing fire will be targeted at the party's PCs while the heavy fighters take on the PC front line? Likewise, will they work to protect their spell casters? Will they exploit bottlenecks and other terrain advantages? I'd be interested to hear your thoughts on this.

 

Meanwhile, I hope your liver survives the PE experience.

 

I don't know how much of this we're going to do, I like all of those ideas and I think it would be fun to build an AI that can do them. We'll just have to see which of those behaviors is a "fit" for the game. I think we'll almost certainly have smarter AI that will try to do things like protect a spell caster or other weak allies. That kind of things isn't too difficult to implement and makes for a more interesting challenge.

[rjshae]

Thank you for the interesting responses, Steve.

 

If I may, a general question I'd like to ask you is regarding AI: how much consideration do you think will be given to group AI by intelligent enemies? Will trained enemy soldiers attempt to fight in formation? Will they coordinate attacks so that, say, suppressing fire will be targeted at the party's PCs while the heavy fighters take on the PC front line? Likewise, will they work to protect their spell casters? Will they exploit bottlenecks and other terrain advantages? I'd be interested to hear your thoughts on this.

 

Meanwhile, I hope your liver survives the PE experience.

 

I don't know how much of this we're going to do, I like all of those ideas and I think it would be fun to build an AI that can do them. We'll just have to see which of those behaviors is a "fit" for the game. I think we'll almost certainly have smarter AI that will try to do things like protect a spell caster or other weak allies. That kind of things isn't too difficult to implement and makes for a more interesting challenge.

Thank you, David. I can understand that some of this type of AI might not be the best fit for a tactical game of this type. Still, some type of group tactical behavior would be nice to see, even if it is just cooperative behavior between neighbors in the mob, or targeting specific enemies. Whatever you decide upon, I'm sure it will be enjoyable. Thanks again.

[Osvir]

False statement(?): A trigger is kind of, putting a "thought" (externally) into the AI "Go that way and use animation 43 to attack".

 

Question: Is it possible for you to create an environment in the game during development where you can play 1on1 against each other (Enemy versus main party/classes)? Is it possible to create/record a "Live" AI based on how you actually play? This is not any flirting about multiplayer or LAN, I'm curious as to how an AI is crafted and if there are more than one way to do it.

[Hassat Hunter]

Liked the update, look forward to seeing more of the team.

[lolaldanee]

all this talking about drugs makes me sick

great update none the less - it is not often that you get to know and your questions ansered by the AI programer

[Parmenides]

I don't know how much of this we're going to do, I like all of those ideas and I think it would be fun to build an AI that can do them. We'll just have to see which of those behaviors is a "fit" for the game. I think we'll almost certainly have smarter AI that will try to do things like protect a spell caster or other weak allies. That kind of things isn't too difficult to implement and makes for a more interesting challenge.

 

It's odd how AI features that aren't hard to implement aren't in many games. Most people want everything to be easy?

[IndiraLightfoot]

I love these "Meet the developers"-updates!

 

I'm looking forward to meeting all the Obsids this way, and hopefully you'll throw in a presentation of George Ziets as well. That would make me and a lot of peeps here very happy indeed!

[Zed]

Thanks for the answers, Steve :brofist:

[Hormalakh]

I don't know how much of this we're going to do, I like all of those ideas and I think it would be fun to build an AI that can do them. We'll just have to see which of those behaviors is a "fit" for the game. I think we'll almost certainly have smarter AI that will try to do things like protect a spell caster or other weak allies. That kind of things isn't too difficult to implement and makes for a more interesting challenge.

 

Tim Cain talks about fuzzy logic in AI programming for Fallout 1 (if he were to do it again) and I'm sure he's got a whole array of AI ideas that you could implement. I hope you're working together with that guru on this very difficult and complex problem.

[anek]

Because the two engines are functionally very different it would mean doing everything twice, I don't know of very many people that like to do the same job twice.

 

I heard that some software companies like to let their developers create a completely functioning full-featured prototype for a new software product, and then (no matter how well it works) discard it and let the developers design and develop it again completely from scratch, without being allowed to reuse any code from the prototype, but learning from the design mistakes and problems that they encountered with it. This is supposed to result in more robust software design and save the company money in the long run (less cost for fixing bugs and providing customer support).

 

But maybe that's just an Internet myth...

[Falkon Swiftblade]

hey I had another question. Some companies have a day a week or once a month that let the staff do anything they want no matter what as long as it's related to a feature in game. Do you guys do anything like that? If not you really should as time goes on

[Morkid]

Cheers from an up-and-coming AI programmer!!

 

There's one thing I would really love to see in this game in terms of AI: Behaviour that scales according to intelligence (or whatever you call it in your new system). A creature with 5 intelligence would have to bevave differently in combat than a creature with 17 intelligence. Probably a more brutish, straightforward assault as opposed to a cunning tactical approach. That would also make intelligence-altering spells more meaningful in combat.

 

Thank you for answering!

[Kaldurenik]

Cant wait for some nice mage battles!

 

Oh i also hope we will have good path finding ai for both enemies and companions

[Steve Weatherly]

hey I had another question. Some companies have a day a week or once a month that let the staff do anything they want no matter what as long as it's related to a feature in game. Do you guys do anything like that? If not you really should as time goes on

 

We don't have a time specifically devoted to that no. But as long as time isn't taken away from planned features, we're usually free to explore those kinds of things (especially in pre-production).

[Scorch]

On AI stuff, have you guys discussed whether you're going with more structured aggro/threat management like in MMOs and DA, or more freeform like in the IE games?

[Hassat Hunter]

Oh, heck, I hope they don't take over the horrible MMO "Aggro" idea...

 

I despise it about as much as Tank/DPS/Healer. Let's break that downward spiral, please.

It didn't help DA at all... rogues should be rogues... not damage dealers :/

[PrimeJunta]

I would very much like to see better AI, with enemies trying to actively outflank you and such.

 

I recognize that it would probably be too much of a break with cRPG tradition, but I'd like to see AI features of squad-based real-time tactical games introduced. Like suppressive fire, units seeking cover autonomously if under fire, even (dare I say it?) morale. The latter would be a natural tie-in to abilities and skills and would make things like willpower more meaningful.

[50dimensions]

that guy is my uncle, noice