Real-time Decisions (Day #1)

Join experts from Microsoft Studios, Creative Assembly, Lionhead and the Universities of Essex and Dortmund for a day-long course on state-of-the-art techniques in decision making from Monte-Carlo Tree Search to Deep Reinforcement Learning or neural networks, why and how they can be applied in interactive applications.

Now in its 7th edition, nucl.ai brings together hundreds of developers from creative industries and interactive media to resolve the biggest technical and design problems with Artificial Intelligence. One of the event's tracks focuses specifically on “Real-time Decisions” for an entire day. (Day #1: Monday, July 20th)

19 Day Tickets Left! Attending Companies & Studios

  • Ubisoft Montreal
  • CD Projekt Red
  • Creative Assembly
  • Bohemia Interactive
  • Epic Games
  • INRIA Research Lab
  • University of Muenster
  • CCP Games
  • Square Enix
  • Microsoft Game Studios
  • University of Essex
  • Unity Technologies
  • Lionhead Studios
  • Iron Galaxy Studios
  • MASA Group
  • ...

I had a great time attending the AI Conference and met so many amazing and nice People there. Once again thank you for giving us students the Oportunity to visit this great Conference!

Jonas Jentsch

Gamedesign Student

Mediadesign Hochschule, GERMANY

The most informative and conference I've been too. Great value and great people. I'll be back.

Chris J. Rock

Wooga, GERMANY

This year's Game/AI Conference was the best one I've gone to so far. The programme was a great balance of high- and low-level, tech and design, as well as academia and industry. The organization was flawless and the location was great.

Jurie Horneman

Independent game designer / programmer

Awesome conferences, where professionals are sharing their knowledge, and their passion. It's quickly going to be the reference in the AI community.

Xavier Dolci

Lead AI Programmer

Ubisoft, FRANCE

This conference was a real deal — cutting edge presentations and experienced speakers. My take away list from this one is impressive! I feel this event can really push Game AI forward!.

Mieszko Zielinski

AI Programmer

People Can Fly, POLAND

I thought the conference was a great success. I learned a lot from the speakers at the conference, as well as interacting with the attendees.

Frederik De Caster

AI Programmer

Creative Assembly, ENGLAND

Thank you for this amazing conference! Great speakers, interesting subjects and nice people. Ten years that I’m looking for a good conference on AI in video games, I’m really happy to have found it.

Gabriel Robert

Senior AI programmer

Ubisoft, PARIS

Well done for putting on such a great conference, the program was varied, interesting and inspiring. Definitely hoping to return next year.

John Lusty

Lead Coder

Ninja Theory, CAMBRIDGE

The quality of the talks was consistently high and the event well organised. I thoroughly enjoyed the opportunity to discuss AI with so many people passionate about the subject and hope to return next year.

Vicky Smalley

Senior AI Programmer

Sony Computer Entertainment Europe

Thanks for organising a thoroughly enjoyable conference. It's nice to meet so many people with a common interest in AI. It was great to finally meet a number of people who I've spoken to via email over the past 2 years...

Neil Armstrong

Lead AI Programmer

Blitz Games Studios, ENGLAND

Thanks for hosting such an excellent conference. I found it great to meet such a broad cross section from the AI development community. It has certainly given me some inspiration for my upcoming projects. I will definitely try to come back next year!

David Partouche

Senior AI Engineer

CCP Games, ENGLAND

Congrats once again on pulling the conference off: it was by far the most inspiring work-related event I've ever been to. I hope it will become a regular event.

Remco Straatman

Lead AI Programmer

Guerrilla Games, AMSTERDAM



Day Tickets
(1 day)

All Access
(3 days)

Early Bird (limited)

All these tickets have gone already...

Sold Out!

Early Worm (timed)

Available in batches of 25 tickets until price increases.

from € 256

from € 540

Notes

  • The main amphitheatre (including some talks in this track) are accessible with regular Conference tickets, starting at €119.
  • Discounts are available for small teams and for those between jobs; email events(at)aigamedev.com for a coupon code.

Image Credit: Forza Motorsport 2

How do you build your simulation to respond to user input at interactive rates, especially if it needs to make intelligent decisions? Artificial intelligence techniques have evolved significantly over the past few years, allowing you to make choices about complex problem domains with any-time algorithms. If you're creating intelligent runtime agents for your games, simulations or applications, this track will help you make the most of recent advances.

Keywords

  • utility systems
  • monte-carlo tree search
  • deep neural networks
  • tactics and strategy

Format

  • Keynote
  • Presentations
  • Panel Discussion
  • Expert Q&A

Track Schedule (Day #1: Monday, July 20th)

Download Real-time Decisions schedule in .ics format.

Track Schedule (Day #1: Monday, July 20th)

Day #1: Monday, July 20th

Crowds and Ambient Life

Real-time Decisions

Virtual / Augmented Reality

Jeffrey Schlimmer, Turn 10 Studios

11:00 Drivatar and Machine Learning Racing Skills in the Forza Series (Amphitheatre)

Gwaredd Mountain, Lionhead

12:00 Tactical Planning and Real-time MCTS in Fable Legends (Amphitheatre)

Chris Poulin, Patterns and Predictions

15:15 Deep Reinforcement Learning In Practice for ATARI Games (Masterclass)

Spyridon Samothrakis, University of Essex

12:45 Imitating Human Play from DOTA2 Game Logs (Workshop) (Laboratories)

Diego Perez, Gwaredd Mountain, Piotr Andruszkiewicz, Simon Lucas

16:45 Open Discussion on Challenges in Modern AI (Laboratories)

Download Real-time Decisions schedule in .ics format.

Keynotes and Presentations

Drivatar and Machine Learning Racing Skills in the Forza Series

The Forza Drivatar technology powering games in the Forza Horizon and Forza Motorsport series is based on machine learning. It enables gameplay beyond robotic drivers and ghost replays by learning player- and style-specific Drivatars that can compete with friends. This presentation will go behind the scenes in the development of Forza Drivatar, explaining the approach from a technical perspective as well as the design and integration with the gameplay.

Jeffrey Schlimmer, Turn 10 Studios

Jeffrey Schlimmer is a principal AI programmer at Turn 10 and has a PhD in artificial intelligence. Turn 10 Studios is the creator of the Forza Motorsport franchise. Founded in 2001 as a first-party Microsoft studio, Turn 10 released the original Forza Motorsport for Xbox in 2005. The studio has...

Read full biography »

Jeffrey Schlimmer, Turn 10 Studios

Jeffrey Schlimmer is a principal AI programmer at Turn 10 and has a PhD in artificial intelligence. Turn 10 Studios is the creator of the Forza Motorsport franchise. Founded in 2001 as a first-party Microsoft studio, Turn 10 released the original Forza Motorsport for Xbox in 2005. The studio has...

Read full biography »

Optimizing MCTS Performance for Tactical Coordination in TOTAL WAR: ATILLA

This presentation focuses on an advanced AI search technique called Monte-Carlo Tree Search, and its improvements for the latest iteration of the TOTAL WAR franchise. Introduced previously in ROME 2, the tactical coordination system saw some major improvements to maximize the number of iterations executed without deteriorating performance. Piotr will dig into techniques that helped improve performance, for example in the structure of the tree itself as well as some higher-level optimisations.

Piotr Andruszkiewicz, Creative Assembly

AI programmer, world conquest enthusiast. Worked on AI for the campaign portion of several Total War titles focusing on diplomacy, pathfinding, tactical planning and optimization.

Read full biography »

Piotr Andruszkiewicz, Creative Assembly

AI programmer, world conquest enthusiast. Worked on AI for the campaign portion of several Total War titles focusing on diplomacy, pathfinding, tactical planning and optimization.

Read full biography »

Tactical Planning and Real-time MCTS in Fable Legends

Fable Legends is a 4v1 asymetric multiplayer game, in which a team of 4 heroes compete against 1 villain. This poses some interesting design and programming challenges. Tactical decisions (e.g. revive a team mate or go for the objective) are not necessarily straight forward when there are a wide range of characters, abilities, constantly changing designs and different game modes. The aim was to create a robust tactical AI that could adapt to any situation and create interesting challenges for players. One that could come up with strategies on its own and surprise us as players. Rather than go a more traditional route (utility scoring, planners/HTN) we implemented a simulation based approach, which involved modelling the game play and using MCTS to search the potential plan space. Overall this worked well, has interesting characteristics and has a number of potential opportunities not readily available with other techniques (e.g. feeding in telemetry data, learning player behaviour, statistical analysis). Being able to update the model and have the system automatically find new strategies is very advantageous. This talk would be a deep dive of the implementation, challenges, performance and result as well as future plans.

Gwaredd Mountain, Lionhead

Gwaredd has been making games from longer than he cares to remember. His career has spanned almost every conceivable programming role from senior management to deep in the trenches, across many platforms and for companies large and small, such as Microsoft, Criterion and Climax. Despite this he still likes making...

Read full biography »

Gwaredd Mountain, Lionhead

Gwaredd has been making games from longer than he cares to remember. His career has spanned almost every conceivable programming role from senior management to deep in the trenches, across many platforms and for companies large and small, such as Microsoft, Criterion and Climax. Despite this he still likes making...

Read full biography »

Innovations in Search-based AI from MCTS to Evolutionary Algorithms

MCTS has revolutionised AI for classic board games such as Go, and more recently has been applied with some excellent results to controlling agents in real-time video games. However, there are many cases where vanilla MCTS fails to produce satisfactory results for a number of reasons, including massive branching factors, limited horizon depth, limited roll-out budget and flat reward landscape. While these problems are not insurmountable, they may involve significant expertise and effort to overcome. On the other hand, there are a number of alternative and similarly general approaches such as neural networks (trained using evolution or temporal difference learning) and rolling horizon evolutionary algorithms that have their own strengths and weaknesses. In this talk I’ll outline the main factors to consider when choosing these methods, and also describe how they can be combined to provide even better solutions. I’ll also discuss initial results from our General Video Game AI server (http://gvgai.net), and how this could create an efficient marketplace for agent AI.

Simon Lucas, University of Essex

Simon Lucas is a Professor of Computer Science at the University of Essex where he is currently head of the School of Computer Science and Electronic Engineering and leads the Game Intelligence Group. Simon is the founding editor-in-chief of the IEEE Transactions on Computational Intelligence and Games, and co-founded the...

Read full biography »

Simon Lucas, University of Essex

Simon Lucas is a Professor of Computer Science at the University of Essex where he is currently head of the School of Computer Science and Electronic Engineering and leads the Game Intelligence Group. Simon is the founding editor-in-chief of the IEEE Transactions on Computational Intelligence and Games, and co-founded the...

Read full biography »

Deep Reinforcement Learning In Practice for ATARI Games

The DeepMind (now Google) team had recently proposed their Deep Reinforcement Learning solution 'DQN' as a solution to human level game playing, using a variant of Q-Learning. With credit to this prior project, our research group has also decided to develop a solution which would be 'at' or 'better' performance in playing video games. Our initial goals were for our agent to similarly reach or exceed human levels of performance, and then to create an open source library for broader use by the research community. These first goals met, we have named this library 'DRL'. And now our current work is to build a dedicated network around agents of this type, eventually making this available for the general public via an existing machine learning API, called Predictus®. During this talk, we'll briefly discuss the science of Deep Reinforcement Learning, the open source DRL library and the commercial API for game development Predictus.

Chris Poulin, Patterns and Predictions

An accomplished Data Scientist in applications of Cognitive Computing technology, Poulin is the Principal Partner of Patterns and Predictions. While he was recently Director of The Durkheim Project, a non-profit big data collaboration with the U.S. VA and Facebook, Inc. Where Fast Company said "This may be the most vital...

Read full biography »

Chris Poulin, Patterns and Predictions

An accomplished Data Scientist in applications of Cognitive Computing technology, Poulin is the Principal Partner of Patterns and Predictions. While he was recently Director of The Durkheim Project, a non-profit big data collaboration with the U.S. VA and Facebook, Inc. Where Fast Company said "This may be the most vital...

Read full biography »

Multi-Objective and Fuzzy Techniques for Balancing and Improving RTS AI

Players of strategy games (as MOBA and RTS) often express their desire for more "intelligent" opponents or bots. Recent game AI developments and the outcome of the last Starcraft AI (bot) tournaments suggest that this can be achieved by Fuzzy Logic approaches that adapt the basic strategy to the current situation. Additionally, another topic of that gets more and more important in the light of user or dynamically generated content and online updates is balancing. Recent works show that a multi-objective approach can be helfpul here because balancing must often respect multiple, contradicting goals. In this talk, we will discuss how these techniques are able to provide a more intelligent and adaptive behaviour in real time strategy games.

Mike Preuss, University of Muenster

Mike Preuss is Research Associate at ERCIS, University of Muenster, Germany. He obtained a PhD in Evolutionary Computation from the Chair of Algorithm Engineering at TU Dortmund in 2013. His research interests focus on further developing and applying multimodal and multiobjective optimization in the context of game AI, namely for...

Read full biography »

Mike Preuss, University of Muenster

Mike Preuss is Research Associate at ERCIS, University of Muenster, Germany. He obtained a PhD in Evolutionary Computation from the Chair of Algorithm Engineering at TU Dortmund in 2013. His research interests focus on further developing and applying multimodal and multiobjective optimization in the context of game AI, namely for...

Read full biography »

Imitating Human Play from DOTA2 Game Logs (Workshop)

Can we use machine learning to imitate human-style play? Can you infer how "average" players behave in given scenarios? To what extend can knowledge mined be incorporated into a game? Provided that logs of user actions are available, one can use supervised learning methods to imitate these expert actions. We will show how this can be achieved, either by creating explicit preferences between actions or labelling states according to some reward scheme. We will examine both methods in the context of DOTA 2, and show how one can train neural networks to learn how to imitate expert user actions, with concrete examples in Python.

Spyridon Samothrakis, University of Essex

Spyros Samothrakis is currently a Senior Research Officer at the University of Essex. He holds a PhD in Computer Science with a focus on Computational Intelligence and Games from the University of Essex, an MSc in Intelligent Systems from the University of Sussex and a BSc in Computer Science from...

Read full biography »

Spyridon Samothrakis, University of Essex

Spyros Samothrakis is currently a Senior Research Officer at the University of Essex. He holds a PhD in Computer Science with a focus on Computational Intelligence and Games from the University of Essex, an MSc in Intelligent Systems from the University of Sussex and a BSc in Computer Science from...

Read full biography »

Open Discussion on Challenges in Modern AI

After many years of research in games, a big boom in academia has changed the way Game AI is done, first after the emergence of Monte Carlo Tree Search, and more recently with Deep Neural Network architectures. Naturally, the video-games industry has taken advantage of these advances featuring MCTS in several titles, and it is reasonable to wonder if DDN can burst into video-games as well. Are we at the birth of a new era for Game AI? Modern AI can help us building better and faster Real-Time decision systems. Follow us into this panel where we will discuss the current open challenges of these techniques, the future directions in the field, and how the interaction between industry and academia can boost and spread modern AI.

Diego Perez , Senior Research Officer at University of Essex

Gwaredd Mountain , AI Programmer at Lionhead

Piotr Andruszkiewicz , AI Programmer at Creative Assembly

Simon Lucas , Head of School of CS at University of Essex

Speakers & Organizers

Jeffrey Schlimmer

Piotr Andruszkiewicz

Gwaredd Mountain

Simon Lucas

Chris Poulin

Mike Preuss

Spyridon Samothrakis

Diego Perez

Browse Conference Tracks

Character Animation Technology

Real-time Decisions

Agent Behavior

Procedural Content Generation

Virtual / Augmented Reality

Crowds and Ambient Life

Advanced Analytics & Data Science

Systemic Design, AI Directors

Cognitive Bots & Language