Over the past few years, deep learning has revolutionized the technology sector by providing significantly better results on classification or prediction problems that were previously impractical for machine learning. What are the benefit for game developers? What's the catch? This presentation is a practical introduction to deep neural networks, using millions of public player profiles and thousands of games ...
This tutorial with Spyridon Samothrakis will cover the techniques and tricks for applying Reinforcement Learning using Neural Networks, particularly to the problem of PlanetWars. You'll learn the different approaches for solving the problem, what needs to be done for it to work at all, and how best to tear your hair out when it doesn't work!
This tutorial will show you how to write bots for The Resistance that are both intelligent and chatty! In particular, you'll learn about building bots that are able to run searches to determine the best course of action, as well as how to use text-to-speech (TTS) and speech-to-text facilities in the framework. If you're curious about building AI for board/card games, find out more in this broadcast!
How are server-based AI systems built for modern free-to-play shooters? This interview with Dmytro Tsakhilov of Crytek Kiev, you'll hear about the implementation of WARFACE's enemy behaviors that populate the game's co-op mode. You'll also learn about the underlying systems built to run scalably on the game's servers.
The last time neural networks were popular it was the early 90s! What has changed since then and how come modern NN can solve the hardest problems now? This is an introduction presentation about neural network technology from the low-level representation to the training algorithms, with a particular focus on the recent advances in the field as well as game applications.
This interview with Omar Ahmad looks at the animation technology in mobile game Shark Eaters: Rise of the Dolphins. The game features a different system that animates skeletal rigs — inspired by neurology and learning of motor control. The result are smoothly animated fish and water mammals whose behavior partly emerges from the animation.
Join Insomiac's Jan Mueller for a look into the AI behind XBox One exclusive Sunset Overdrive. In particular, you'll hear about the challenges and solutions for building compelling boss AI, how to create a good dog companion, and general solutions for enemy AI in open-world where players traverse very fast!
This interview with Matthew Jack digs into the technical details behind the technology that powers Star Citizen, in particular for Arena Commander: a dog-fighting and flight simulation module for the game. This includes the maneuver splines and 3D velocity obstacles, as well as behavior trees and live inspectors. You'll see the debugging tools in action, as well as the wider challenges of building a living ...
Building an AI that can scale to large planets is no easy task, and doing so with a Neural Network adds its own additional challenges! In this interview with Mike Robbins, Software Engineer at Uber Entertainment, you'll learn lessons learned applying machine learning to this highly acclaimed RTS title — as well as dealing with strategy on multiple spherical planets.
This tutorial will focus on the new OpenCL specification called SYCL that allows developers to write parallel code in C++ as a single-source program. Using modern C++ programming techniques, kernels that run on the GPU (device) can be integrated into CPU code (host) easily using templates. This session will show some patterns to use in your code to make sure it's SYCL-compatible, and tools to use to help improve ...