1 / 15

The Challenges and Follies of Building a Generic AI engine

The Challenges and Follies of Building a Generic AI engine. Dr. Paul Kruszewski, CTO © 2000–2004 BGT BioGraphic Technologies Inc. Overview. AI.implant: Post-mortem the first 4 years

RoyLauris
Download Presentation

The Challenges and Follies of Building a Generic AI engine

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. The Challenges and Follies of Building a Generic AI engine Dr. Paul Kruszewski, CTO © 2000–2004 BGT BioGraphic Technologies Inc.

  2. Overview • AI.implant: Post-mortem the first 4 years • I thought I was going to talk over game could teach film but it is really about what we learned from all of you • The vision • What went right • What went wrong • Pleasant surprises • The vision’ (Take 2)

  3. The mindset • Background • Procedurally modelling of branching patterns (“the tree guy”) • My Virtual Model • Just quit my job as CTO to start-up a company • Convergence between film and game • SIGGRAPH vs GDC • Crowd simulation vs game AI • I felt they were solving the same problem but didn't know it

  4. The vision • Build an AI generic game engine • Different markets • Special Effects (SFX) : digital extras • Video games : AI middleware • Simulation: digital soldiers • Different users • Animators • Level editors • Programmers • Roll out / Adoption • Film • Cut scenes • Game engines • Simulation

  5. The vision • Build an AI generic game engine (cont.) • Architecture • Character based • If it moves in an intelligent way, it’s a character (e.g., human, bird, fish, etc.) • Otherwise it’s physics • Real-time • C++ SDK • Integrated as visual plug-ins into art packages and components into game application • Data-driven

  6. What went right • We survived as a company • Dot com crash • Many (AI) middleware companies came and went

  7. What went right (cont.) • Overall architecture was correct • The character / vehicle paradigm • Humans • Fish • Flaming Cows • Tanks • Space ships • Spiders • Weird bipeds • Tools went well • Integration into Maya/max was key to every sale in the entertainment space

  8. What went right (cont.) • Customers used it in all three markets • For film and video • For game cinematics • For a PS2/Xbox game (PSI-Ops) • For simulators

  9. What went wrong • $ • VC crash; money came more slowly and more expensively • Game industry shakedown • developers went broke after they bought the product but before they paid us • Price slashing in animation • All of this naturally affected our execution

  10. What went wrong • Technology • We initially built something that no one liked • Game people said that it was only good for special FX • SPX people thought of it as a game engine • We built the wrong things first • Animation • An AI system was no good if you couldn't • control the underlying animation • render things out • Game • People really cared only about the path finding • Didn't want level editors/animators to author things • Build decision trees instead of FSM • Focus was on flocking and decision making • Pathing and animation control turned out to be the hot stuff

  11. What went wrong • Technology (cont.) • Too ambitious • Too split on the two markets with two different pipelines • Too many tools • Wide but not deep functionality • Consoles are hard • Bleeding edge tech causes you to bleed • Culture • Learning the 3 cultures took longer than we thought • speaking game with a heavy film accent • Really underestimated the resistance to middleware • You can't argue on cost-it really has to be better • It is a very new field so we had to make up a lot of stuff including language

  12. Interesting twists • Military simulation • We really expected the military to be ahead of us • KMW uses our system to drive tanks and control all humans • Decision trees • Wrong choice for game (should have used FSM) but animators love them • Did a lot of things that weren’t AI (we really thought somebody handled them already) • Surface solving • Animation control • Essentially turned Maya into a game engine

  13. Moving forward • Challenges in game AI • Complex worlds • New graphics cards allow huge worlds • Physics creates dynamic worlds • Automatic tools are no longer an option • Navigation meshes are becoming the core data structure • Complex characters • Integration of all human systems to form intelligent skeletons • AI driven NLA • IK/FK • Ragdoll • Volume • No more ghost town (Craig Reynolds) • Parallelism

  14. Moving forward (cont.) • Industries • Game will continue to be the most dynamic environment for innovation • Animation (SFX). It is really becoming a question of playing the game with the record button on • Military is most open to these techniques • Convergence is happening, I was just out by 4 years • Film, games and simulations use the same database (Blackhawk down) • Most users want a drag and drop system particularly in sim and film

  15. Thanks • Dan Fu for inviting me • My team back home for building such difficult software • All the customers who have helped us along the way

More Related