Matrix Games Forums

It may correct my spelling, but it will never remove your anti-semitism.

http://www.rockpapershotgun.com/2012/11/30/interview-limit-theory-creator-josh-parnell/#more-133681

Interview over at RPS.

Guys keep these on topic or I will lock it up. There has been an unusually high level of friction recently and everyone needs to calm down. This is place to relax and have fun and talk about games. No politics please.

I backed the Elite:Dangerous kickstarter.

That would the final game needed to see me through to retirement (20 years or so :p)

ORIGINAL: wodin

http://www.rockpapershotgun.com/2012/11/30/interview-limit-theory-creator-josh-parnell/#more-133681

Interview over at RPS.

"dynamic, supply-and-demand economy in which numerous NPC merchants will compete to beat you to exploiting the price differentials"

Umm.. so it's going to be dynamic and living - where NPC activity will be going on while you're puttering about between planets or whatever. In an infinite sandbox? I wonder where he plans to get the infinite CPU time to do all the NPC actions and the infinite memory to hold all these infinite ports, ships, planets, asteroids, etc.

Oh, hold on - it's not infinite...

"On the contrary, the universe is dynamically generated on-the-fly as you push further into new systems. This means that, although the universe goes on forever, it’s not all stored on your hard drive at once."

So it's really only generating one system at a time - and when you go onto the next system, when you return, it's generated again. This is no Elite - any comparison to Elite withers and dies with that statement.

Kudos to his graphics API usage - but that's only the tip of the iceberg. It's only a tiny piece of making a game - much less a good one. NPC AI, combat AI, economic balancing as a result of NPCs - seriously, drawing the graphics on the screen is trivial compared to these.

ORIGINAL: Kayoz
"dynamic, supply-and-demand economy in which numerous NPC merchants will compete to beat you to exploiting the price differentials"

Umm.. so it's going to be dynamic and living - where NPC activity will be going on while you're puttering about between planets or whatever. In an infinite sandbox? I wonder where he plans to get the infinite CPU time to do all the NPC actions and the infinite memory to hold all these infinite ports, ships, planets, asteroids, etc.

Oh, hold on - it's not infinite...

"On the contrary, the universe is dynamically generated on-the-fly as you push further into new systems. This means that, although the universe goes on forever, it’s not all stored on your hard drive at once."

So it's really only generating one system at a time - and when you go onto the next system, when you return, it's generated again. This is no Elite - any comparison to Elite withers and dies with that statement.

Kudos to his graphics API usage - but that's only the tip of the iceberg. It's only a tiny piece of making a game - much less a good one. NPC AI, combat AI, economic balancing as a result of NPCs - seriously, drawing the graphics on the screen is trivial compared to these.

Hi Kayoz - I'm afraid it's going to take quite a while to dispel all of your commentary, but I'll try, nonetheless

Yes, you're correct, the entire universe obviously can't be simulated at once. The way the LT simulation works is that the local area around player assets is generated and simulated. Systems that are farther from player assets are simulated at a lower LOD than systems that are nearer - with the obvious maximum being at the system in which the player currently resides.

As far as generating systems goes - it sounds like you're describing a non-persistent system. But that's not the case in LT. When you generate a system, you can make change to it (i.e. building or destroying stations), and those changes persist! Furthermore, the systems aren't generated one at a time, they're generated in local clusters. This helps from the simulation perspective, as it enables "semi-global" system dynamics.

Finally, I'd like to encourage you (and others who might be tempted to do so) to avoid speculation on the "difficulty" of things if you haven't implemented them! Your statement that graphics is trivial compared to AI, for example, belies a lack of implementation knowledge. Real-time graphics is a field of hackery, of duct tape, and of generally trying to get a machine to do something that it doesn't want to do. AI, on the other hand, is a sandbox for beautiful, theoretical ideas and lovely abstractions. The hardware is perfectly happy to perform Q-learning in real-time or keep track of a state space model. In this sense, AI programming is significantly more natural than graphics programming. If we were to try to use the purest of theoretical ideas in graphics, we would have to perform path tracing, but of course GPUs aren't able to do so in real-time yet. So again, I encourage you not to guess at where the difficulty lies

Hope to respond to a few more of those comments in time!

ORIGINAL: Kayoz

20 year old kid with no experience, no track record, nothing more than an inflated ego and immense hubris.

That's fairly harsh. I do have a track record, I'm not sure if you bothered to check out the development blog, for example. And, I guess it's hard for me to say, but I don't think that I've got an immense hubris. I'm well aware that I'm still learning many things.

ORIGINAL: Kayoz
You're right, I don't know him. Nor does ANYONE in the software development industry, much less games development. Zero experience and zero projects to your credit does that.

You're right, I haven't played the game. Nor has ANYONE ELSE. There's nothing to play. There IS NO GAME. It's vapor-ware.

Did I touch a raw nerve? You seem to be rather defensive over my pointing out the facts of his age and complete lack of experience. You should be thankful that I didn't point out that he plans to do it all on his own - a project of that size isn't the sort of thing a wet-behind-the-ears inexperienced programmer should be tackling alone.

You keep saying "complete lack of experience," but it's definitely worth noting that "not having shipped a game" and "complete lack of experience" are not at all the same thing. Again, check my dev blog if you'd like a look at the amount of practical experience that I've obtained over the years.

I would also encourage you not to speculate about my industry connections, as, like Igor said, you really don't know me But that doesn't mean that "nobody" knows me...!?

Finally, I think vapor-ware is bit of an inappropriate statement to make about an engine that has been in development for a few months and continued to show rapid progress during that time.

By all means, I understand your skepticism Kayoz, but there's a difference between realistic skepticism and needless personal attacks. It would be nice if we could keep the commentary to the former

Nice work, Josh. Good luck with Limit Theory!

ORIGINAL: JoshParnell
Real-time graphics is a field of hackery, of duct tape, and of generally trying to get a machine to do something that it doesn't want to do.

Indeed, getting your brain wrapped around binary space partitioning can make your grey matter dribble out your ears. But it's all been done before. There's a vast amount of documentation and resources to help you through all the rough spots. With AI development however, there are precious few resources and of those, very few are applicable to what you are trying to accomplish.

This is what I mean by graphics being the "trivial" part of the project. There's a lot of work - hard and long work - but it's comparatively straightforward in comparison.

ORIGINAL: JoshParnell
AI, on the other hand, is a sandbox for beautiful, theoretical ideas and lovely abstractions. The hardware is perfectly happy to perform Q-learning in real-time or keep track of a state space model. In this sense, AI programming is significantly more natural than graphics programming. If we were to try to use the purest of theoretical ideas in graphics, we would have to perform path tracing, but of course GPUs aren't able to do so in real-time yet. So again, I encourage you not to guess at where the difficulty lies

You seem to be in disagreement with most game code-monkeys that I know of.

Here's an example
"The second challenge is where I throw down the gauntlet and challenge game AI to save the day. If increasing realism doesn't give next-gen games the requisite new feel, then something else must help achieve it" - Steve Rabin, AI Game Programming Wisdom

As you say, AI is full of "beautiful, theoretical ideas and lovely abstractions" - which is where the difficulty lies - turning an abstract idea into solid code is quite difficult. It's easy to see if the 3D models display correctly - it's quite another to check if the ships displayed are flying around in a sensible manner and reacting reasonably to the player's actions as well as the other NPCs and their environment. This, I contend, is much more difficult and time consuming than merely rendering them on the screen.

How many pretty games are there out there, with shiny graphics - lots. How many out there are actually challenging and make the player stay on their toes - precious few. The ratio of those - pretty to challenging ones - should be a warning sign as to how simple (as you seem to contend) AI programming is.

I can't find his (Rabin's) book on my shelf at the moment, but I contend that a good AI is key to a game's success - and is the most difficult part of game development. That you seem to think it's "more natural" to code is either an indication that you're a programming prodigy - or that you haven't really considered the scope of what you're trying to do.

ORIGINAL: JoshParnell
You keep saying "complete lack of experience," but it's definitely worth noting that "not having shipped a game" and "complete lack of experience" are not at all the same thing.

No professional experience.

If you go into a job interview and they ask you "what experience do you have" - they mean professional experience. Whether or not you've fixed your own car is of little interest or value in interviewing for a mechanic job - they want to know what professional experience you have. Regardless of your tech demos - they're not being done in a professional environment where you have programmers with decades of experience looking over your shoulder and picking apart your code. Nor the QA bods trying to break your code - and some of them are very good indeed at that. You haven't experienced this, and until you do, I contend that the "zero experience" label is justified.

I wish you the best of luck - I really do. But it seems that you've picked out an enormous mountain to move - all on your own and with no real experience at moving mountains. That isn't to say that you cannot succeed. It's possible - but in my experience - it's very unlikely. Time will tell. I'll be sure to bump this thread in a year to confirm or correct my prediction.

ORIGINAL: Kayoz
Indeed, getting your brain wrapped around binary space partitioning can make your grey matter dribble out your ears. But it's all been done before. There's a vast amount of documentation and resources to help you through all the rough spots. With AI development however, there are precious few resources and of those, very few are applicable to what you are trying to accomplish.

This is what I mean by graphics being the "trivial" part of the project. There's a lot of work - hard and long work - but it's comparatively straightforward in comparison.

That's fair, you're certainly right that, on the whole, there's more documentation on practical implementations of graphics techniques than AI techniques. Although, it's worth mentioning that the reverse is probably true of documention on theoretical ideas (but I won't claim that means anything). Still, if you're going to choose one part of the game to call "too difficult," I should think that it would be the procedural generation of everything. After all, there are no well-documented sources (of which I am aware) that describe how to build the kinds of nebulae that you see in my engine, or how to construct ships in the manner that I have done. I'm not saying either of those things is hugely difficult - just that one should ask why there's even less documentation on procedural techniques than on AI.

ORIGINAL: Kayoz
You seem to be in disagreement with most game code-monkeys that I know of.

Here's an example
"The second challenge is where I throw down the gauntlet and challenge game AI to save the day. If increasing realism doesn't give next-gen games the requisite new feel, then something else must help achieve it" - Steve Rabin, AI Game Programming Wisdom

As you say, AI is full of "beautiful, theoretical ideas and lovely abstractions" - which is where the difficulty lies - turning an abstract idea into solid code is quite difficult. It's easy to see if the 3D models display correctly - it's quite another to check if the ships displayed are flying around in a sensible manner and reacting reasonably to the player's actions as well as the other NPCs and their environment. This, I contend, is much more difficult and time consuming than merely rendering them on the screen.

This part I may have to disagree with; my point was that game AI is a field in which turning theory into real-time implementation is not prohibitively difficult! Think deterministic state space AI. Completely trivial, yet nonetheless powerful and certainly fast enough for real-time games.

As for the Limit Theory AI, it uses an even simpler model for high-level AI, which is nonetheless absurdly powerful. The core of LT AI is a planning engine, in which high-level goals are formulated as game state differentials, and then a set of heuristics is called upon to turn a game state differential into a sequence of concrete actions. For example, an obvious heuristic is that if a goal is to modify the location of oneself, the "solution" is to "travel" to the desired location. It's effortless to think of an AI model in this way, and equally effortless to implement. But I claim that the results are impressive and fast. I guess that's the part that I'll have to prove!

ORIGINAL: Kayoz
How many pretty games are there out there, with shiny graphics - lots. How many out there are actually challenging and make the player stay on their toes - precious few. The ratio of those - pretty to challenging ones - should be a warning sign as to how simple (as you seem to contend) AI programming is.

I can't find his (Rabin's) book on my shelf at the moment, but I contend that a good AI is key to a game's success - and is the most difficult part of game development. That you seem to think it's "more natural" to code is either an indication that you're a programming prodigy - or that you haven't really considered the scope of what you're trying to do.

I think my choice of genre is absolutely key to why I claim that AI is easy. Let's face it: in a space game, you have minimal things to worry about. Pathfinding is extremely easy, collision avoidance equally so...in general, there's just a lot less that one has to worry about. Thus, in a space game, most of the work falls into the high-level AI, i.e., the planning engine. And that part, as I already mentioned, is the place where one uses beautiful abstractions to make everything easy. Now, to be fair, I would NOT claim that AI is, in general, an easy thing for a game. In fact, I most definitely would not want to deal with AI in your average AAA game. But if you think about it for a while, I think you'll probably agree that space sim AI is a good bit simpler, and affords the programmer more time to worry about the "cool/fun" part of AI (high-level) and worry less about the low-level nuisances that pop up everywhere in other games.

And by the way, no worries, I have the book

ORIGINAL: Kayoz
No professional experience.

If you go into a job interview and they ask you "what experience do you have" - they mean professional experience. Whether or not you've fixed your own car is of little interest or value in interviewing for a mechanic job - they want to know what professional experience you have. Regardless of your tech demos - they're not being done in a professional environment where you have programmers with decades of experience looking over your shoulder and picking apart your code. Nor the QA bods trying to break your code - and some of them are very good indeed at that. You haven't experienced this, and until you do, I contend that the "zero experience" label is justified.

Ok, that's still not fair! I don't want to name drop, but if you go to my website you can download my CV and see that I have worked in the graphics industry. Sure, you can say that it's "not enough" professional experience, or whatever. All my work was code reviewed, and I did indeed get grilled by pros! But hey, credit where credit is due please.

ORIGINAL: Kayoz
I wish you the best of luck - I really do. But it seems that you've picked out an enormous mountain to move - all on your own and with no real experience at moving mountains. That isn't to say that you cannot succeed. It's possible - but in my experience - it's very unlikely. Time will tell. I'll be sure to bump this thread in a year to confirm or correct my prediction.

Thank you. Again, I understand your skepticism, and all I ask is that we can be civil about it. Obviously, the only way I can prove it to you is to make this game, and make it as good as I say I will. So yes, I hope we can speak again in a year-and-some

Finally, I'd like to encourage you (and others who might be tempted to do so) to avoid speculation on the "difficulty" of things if you haven't implemented them! Your statement that graphics is trivial compared to AI, for example, belies a lack of implementation knowledge. Real-time graphics is a field of hackery, of duct tape, and of generally trying to get a machine to do something that it doesn't want to do. AI, on the other hand, is a sandbox for beautiful, theoretical ideas and lovely abstractions. The hardware is perfectly happy to perform Q-learning in real-time or keep track of a state space model. In this sense, AI programming is significantly more natural than graphics programming. If we were to try to use the purest of theoretical ideas in graphics, we would have to perform path tracing, but of course GPUs aren't able to do so in real-time yet. So again, I encourage you not to guess at where the difficulty lies

Giving how easy is to code a decent AI, it's strange how many of the commercial, real-time games have problems with NPC / enemies bumping meaningless against obstacles they could easily avoid or move around.

Hi Josh.
I hope you can implement cockpits and TrackIR support. All true space combat/trade/exploration sims should have them. [;)]
Speak with the guys at NaturalPoint for a SDK TrackIR kit. I'm sure they will want to help you.
Best of luck with your project and make the dream come true. [:)]

ORIGINAL: HanBarca
Giving how easy is to code a decent AI, it's strange how many of the commercial, real-time games have problems with NPC / enemies bumping meaningless against obstacles they could easily avoid or move around.

Can't tell if sarcasm or not But yes, it surprises me as well. Then again, obstacle avoidance is one of those nastier low-level implementation details that can be tricky. Like I said above, the nice thing about a space game is that obstacle avoidance is all too easy!

Can't tell if sarcasm or not But yes, it surprises me as well. Then again, obstacle avoidance is one of those nastier low-level implementation details that can be tricky. Like I said above, the nice thing about a space game is that obstacle avoidance is all too easy!

Anyone remember some of the later versions of Elite which actually modeled planets as more than just a circle? It seemed that the AI was totally incapable of NOT smacking your ship straight into the planet that you wanted to land on. So - at least back then - mega-structures like planets were large enough to cause problems with path finding. Big, game breaking problems in fact.

Of course Elite was a long time ago and AI theory has had plenty of time to grow. Maybe those problems would be trivial to solve now?

ORIGINAL: JoshParnell

Can't tell if sarcasm or not But yes, it surprises me as well. Then again, obstacle avoidance is one of those nastier low-level implementation details that can be tricky. Like I said above, the nice thing about a space game is that obstacle avoidance is all too easy!

That response is almost distressingly scary. If path-finding is a big deal, Josh has some shocks in for him.

ORIGINAL: JoshParnell

ORIGINAL: Kayoz
No professional experience.

If you go into a job interview and they ask you "what experience do you have" - they mean professional experience. Whether or not you've fixed your own car is of little interest or value in interviewing for a mechanic job - they want to know what professional experience you have. Regardless of your tech demos - they're not being done in a professional environment where you have programmers with decades of experience looking over your shoulder and picking apart your code. Nor the QA bods trying to break your code - and some of them are very good indeed at that. You haven't experienced this, and until you do, I contend that the "zero experience" label is justified.

Ok, that's still not fair! I don't want to name drop, but if you go to my website you can download my CV and see that I have worked in the graphics industry. Sure, you can say that it's "not enough" professional experience, or whatever. All my work was code reviewed, and I did indeed get grilled by pros! But hey, credit where credit is due please.

Credit where credit is due? Sure.

Entire credit given: ZERO

No CV on website. NO professional experience. I stand by my evaluation of your experience (or lack thereof).

ORIGINAL: Kayoz
That response is almost distressingly scary. If path-finding is a big deal, Josh has some shocks in for him.

Did I say path-finding was a big deal for this game? No, not at all. In fact, it's already finished. A*...one of the most trivial graph algorithms known to man. And collision avoidance is finished as well, using a Monte Carlo point-repulsion model. Again, completely trivial. I have very few shocks in store, because I code these things all day and, as such, am aware of the implementation details. Again, I can't help but feel that you're getting your knowledge from books or quotes, not experience!

ORIGINAL: Kayoz
Credit where credit is due? Sure.

Entire credit given: ZERO

No CV on website. NO professional experience. I stand by my evaluation of your experience (or lack thereof).

Hmm, now I can't tell if you're being abrasive just for the fun of it? Or if you're perhaps trying to be clever in pointing out the difference between a resume and a CV? I'm sure you're aware that, as a newcomer to the forum, I'm not allowed to post links. But adding a dot-com to my username and doing a five-second scan of the page should land you on the proof.

Given the strength of your judgement, I'd be curious to hear of your personal experience, by the way...I'm guessing you've shipped many AAA titles?

Does this system build things? How are the build points/money transfered from planet to planet or is that not part of the system?

ORIGINAL: Mobius

Does this system build things? How are the build points/money transfered from planet to planet or is that not part of the system?

Sorry, I'm not entirely sure that I understand the question. Are you asking if the AI is capable of constructing assets?

If that is indeed the question, then the answer is as follows: money flows naturally between planets as a result of trade price fluctuations, and the fact that AI will naturally smooth out wealth/price differentials in order to make a profit. However, I should note that the planet itself is not the entity with money. Rather, the factions that control the locations on the planet with which you trade are the ones that control the money. When a faction has acquired enough wealth, through whatever means, and the HLAI that is serving as manager of the faction decides that it is in the best interest of the faction to build a new ship / space station, the faction will do so. Wealth will then naturally be transferred to the owner of whichever factory is commissioned to do the building. In many cases, the faction will own their own factory, in which case wealth will be transferred to traders who bring in the parts for building (the factory will post orders for parts and offer competitive prices accordingly, so as to move forward with construction).

Hope I didn't answer the wrong question!

ORIGINAL: JoshParnell
Did I say path-finding was a big deal for this game? No, not at all. In fact, it's already finished.

A* has been around for ages, yes - I'm quite well aware of that. There are as many variations of it as there are implementations, though - since that largely depends on how you do your map.

My observation was based on two statements:
"obstacle avoidance is one of those nastier low-level implementation details that can be tricky"
"space sim AI is a good bit simpler, and affords the programmer more time to worry about the "cool/fun" part of AI (high-level) and worry less about"

You seem to think that pathfinding is a big part of your game's AI - where it isn't. Look at any flight combat simulator - once the physics model is done to simulate their flying - the nasty bit of how they fly comes into question. Bank left, right, pull an Immelmann? In your case, will they play dodge-em around asteroids or try to hit your flanks while you focus on one of them? What goods the NPC traders are carrying and to where - That's where your "real" AI starts to come in.

Dude, games live or die on the quality of their AI - whether or not it challenges the player. If you seriously think you have the AI problem licked because you've managed to do A* - then you have some real surprises in store for you.

ORIGINAL: JoshParnell
But adding a dot-com to my username and doing a five-second scan of the page should land you on the proof.

Burying your resume somewhere on the 20th page of a forum post or some obscure Wordpress blog isn't being honest about things. You don't put it on your Kickstarter page, do you? You don't put it on your ltheory.com page, do you? If I look in the places normal people look, I can't find it. That's worrying.

My experience? When I come asking for $50k in donations, I'll happily release my details. The fact is that I'm not and you are. But you've got some $89k in pledges already. More than enough.

But wait - Elite: Dangerous is a project proposed by one of the very same people who made the original Elite - he invented the whole genre. Yet he - a seasoned professional who's actually done it - has budgeted $1.25 million. You claim to be able to do largely the same thing for... 4% (1/25) of the budget? And you seriously expect people to trust you at your word? If I propose to build someone a car that'll go like a Ferrari for 4% of the cost of a real Ferrari - I'd expect to be laughed at and openly mocked.

Besides, this is the Internet. Men are men, women are men, and children are FBI agents. Anything anyone says that can't be traced to a reliable source - is BS.

For the record - yes, I'm always abrasive. It's a gift.

ORIGINAL: JoshParnell
I can't help but feel that you're getting your knowledge from books or quotes, not experience!

Oh, that's precious. Coming from someone who has zero professional experience. You made my day. Biggest giggle all day. Abandon the game and go into comedy - it's clearly your calling.