Tuesday, September 11, 2012

Talk Points

I just found out the audio for my talk at the Mont Royal Game Society was very poorly recorded. It seems the only way you will hear this presentation is you invite me for beer. Anyway for what it is worth here are the slides. They were designed to go with the talk so there is a lot missing.


31 comments:

  1. What an awesome set of slides! Pity we can't hear the talk.

    ReplyDelete
  2. I was looking forward to hearing the talk, oh well. In light of that could we get a post detailing some of the more important parts?

    ReplyDelete
  3. Thank you for posting the slides! It is definitely a disappointment that the audio was poorly recorded. The slides inspired me a ton already, and there are many holes in my knowledge of what you're doing that I wish could be filled by this talk! Do you plan on giving the talk at another time, or a similar talk?

    ReplyDelete
  4. Very nice work. Reading the slides I've got the conclusion that you mix pre-rendered elements like "column_top.3ds" as primitives of your L-systems. Is there any good (public domain) object library?

    Thanks.

    ReplyDelete
  5. Thank you very much! These are really appreciated!! :)

    ReplyDelete
  6. Great slides! I wish the audio was available! Nonetheless, the slides by themselves were a great eye-opener. Thank you!

    ReplyDelete
  7. God !
    Even though the audio is bad quality, i'd still love to hear it. But thank you for the slides! They do sum up a lot of what you said here since the beginning.
    Anyway, I can't wait to see what you'll do with your tool, it's really an amazing piece of work !

    ReplyDelete
  8. By using logic, and pre-acquired knowledge, I can figure out some of the things =3, and its a nice recap of all the stuff you did. It also makes a few things more clear.

    So... did the last slide have as last step baking? O.o

    ReplyDelete
  9. Well, I hereby invite you to Leipzig/Germany. I will pay for food and drinks - so come and get some beer.

    Damned, I think we will have to pool some money to get a flight to Europe for you. And apparently there are much people here who would appreciate this.

    Keep on the good work - and thanks for sharing your ideas.

    ReplyDelete
  10. Maybe you could redo something like this as a Siggraph 2013 talk?

    ReplyDelete
  11. Although I love the slides and your work in general. I am somewhat offended (not really) by you saying that procedural generation of drama is impossible (or at least that you say there is no way)

    One interesting field of AI is the research being done in adaptive dynamic narrative. There are a plethora of ways you could enrich your "states fighting" to include stories of kings and warlords, old feuds and allegiances. NPC's could use basic speech generation to make remarks about big events and certain leaders. You could generate missions based on this. For instance, an NPC from one state could ask you to assassinate an important leader or figure from another state.

    True, you would have to do some work in enriching the possible stories, but it would be amazing to find that a city that you previously visited, but where you didn't perform an important story-quest has been razed. Or that you could lead an uprise to overthrow a leader of a town where the population is unhappy.

    Diseases could spread and figures that battle along you could die. People could become angry at you when rumours are spreading that you visited another town.

    These are things you can generate, which could become very rich in content if you have a complex enough system.

    Also: could you write up some notes for each slide to explain what the general notion was?

    ReplyDelete
    Replies
    1. Aha! You have every reason to be offended.

      This is one thing I did mention in the talk. Generating history is not difficult, but history does not equal drama.

      Just to reduce it to a very simple example. History is when the Judaic/Christian Bible goes on forever on who fathered who. It is quite boring.

      Even if you take it to more interesting verbs, like assassination, it remains history. Who killed some other guy and why.

      What makes Hamlet dramatic is not your uncle murdering your father and banging your mother. These are just facts. The play is appealing because it is constructed in a way that binds to your human fiber.

      So while facts can be generated with success, they are boring. I think if you rely on this for your game, you are making a boring game. Of course humans make boring games and stories too, but I yet have to find a computer generated story that makes me laugh or cry.

      We are talking AI singularity here. It has not happened and when it does, better run for the caves.

      Delete
    2. Dwarf Fortress has some great "drama" generation. But the laughs mostly come from the absurdity, and the tears from the emotion that the player instils in the characters.

      Delete
    3. Yes, I agree. In general dramatic circumstances may arise from any simulation, especially when the player is invested in his/her creation. Seeing your city burn to the ground in SimCity can be quite moving.

      But these are exceptions in a large number of bland and silly scenarios. This cannot be considered a system.

      Delete
    4. I feel like there should be SOME way to make drama and stories procedurally generated, but that we just havent found that way yet... I understand why you say no way, however, people in the field of AI (since thats where it fits, doesnt it?) might be discouraged (Or actually encouraged, just to prove you wrong XD), which might actually mean it will take longer for this way to be developed.

      But I dont blame you =D.

      Delete
    5. I mean it is not possible right now. Computers beat us at Chess for a while, now Jeopardy...

      It is a matter of time. Then war will break. We will win. Computers will be forbidden. There will be tales of the time we used to live in Procedural Worlds.

      Delete
    6. Or we just, you know, BECOME the computers =3.

      Delete
    7. This is a very interesting problem. I think it could be solved, at least in the sense that it gives results that are "good enough" but not true simulation. I propose mapping internal emotional states and opinions about "things" and other NPCs of NPCs and perhaps doing some optimization such as inherited opinions from group belonging.

      The big problem I think is getting believable dialogue, because without AI capable of analysing a lot of literature, one would have to have a huge team of writers inputting data and doing emotional state mapping examples to create a large enough database of suitable strings to exchange for the basic NLP ones. ex: Brutal character "I will kill you" => "I will break every bone in your body"

      Delete
  12. First post on your blog and must start by complimenting you on your quest. I'm truly amazed by the quality of your work.

    I've been reading most of your blog post on here so your presentation slide made some sense anyway but if it wouldn't be to much hassle for you maybe you could rerecord the sound and combine it with your slide. You could just use a screen recorder like Camtasia.

    And for what it's worth, I for one would really like to see your Voxel Studio published and that it would export in a format that would be easy to integrate into a game engine like Unity.

    And lastly will your generated world be interactive? With that I mean would you be able to use a shovel to dig a ditch or cut down a tree. As far as I understand it you convert everything into polygons and that makes it hard to interact with I guess?

    Sorry for all the questions and suggestions but I'm just extremely interested. And even though I agree with you that people get bored easily, in no way is your blog boring - keep it up!

    ReplyDelete
    Replies
    1. Thanks.

      There is another talk coming in November. Let's see if we have better luck. To be honest it feels weird recording this talk alone. I rather not do it.

      I have two versions of the Voxel Farm engine, one if for static worlds like Skyrim's, the other allows dynamic modifications like Minecraft. They both use polyongs, they do not create any hard problems when it comes to carving or adding stuff.

      Hopefully all this will published soon, with ties at least to Unity.

      Delete
    2. About the editable one, I take it it will be possible to dig under ground, creating overhang?

      Delete
    3. Yes, you can dig/carve anything you see. But I am keeping this branch of the tech under a lid for now. I will make announcements later. This is all I will say.

      Delete
    4. Awesome, sounds really nice!
      Hope to hear more about it.

      Delete
    5. Ties to Unity sounds INCREDIBLE! I can't wait!

      Delete
  13. Just a curiosity, what tool you use to create the slides??

    What kind of pen??

    ReplyDelete
    Replies
    1. I used the "Paper" app for iPad and a crappy $14.99 pen I got at BestBuy. I don't remember the brand, one of my twins ate the tip and it was thrown away (the pen, not the twin...)

      Delete
  14. Would love to have a link to the audio, even if it's poorly recorded. Perhaps someone could perform some processing on it, and create a transcript?

    ReplyDelete
    Replies
    1. I'm sorry, it is really unworkable. Also the first 10 min of the talk were cut.

      Delete
    2. Hey, a transcript would be great, but even better, get invited somewhere else and do it again :D

      Ooh, oh! Get invited to TED :D Or make a local TED conference on PC graphics.

      Delete
  15. cant decide which is better raytracing polygons or or what ever you are doing!

    ReplyDelete