Design - Generative AI with Large Language Models (LLM) for Game Narrative and Level Design (Tommy Thompson)
Formation créée le 19/06/2024. Dernière mise à jour le 29/10/2024.
Version du programme : 1
Programme de la formation
The recent emergence of Generative AI has led to a lot of open questions as to how it can be used in game development. Is it practical to use in games? How can it be adopted? What are the skills developers need to get started? In this class we focus on how large language models (LLMs) popularised by GPT can be used in a variety of use cases, ranging from traditional game design ideation, to narrative design, dialogue generation, and even NPC control and level generation. This class will prime attendees with how to get the best out of these tools, and safe and practical ways to use them in a game development pipeline. 1) Generative AI: An Introduction a- Fundamentals of Artificial Intelligence .IP issues and how to protect your data b- AI for Content Generation: A Short History .Procedural Content Generation .PCG-ML .Natural Language Processing .NLP before LLMs c- Generative AI: Explained .Foundational Theory .Generative AI Use Cases 2) Large Language Models: Theory and Practice a- What is a LLM? b- Understanding transformer architectures. .Relevance of parameters counts etc. .Common Techniques and Approaches c- Prompting and Prompt Engineering d- Training LLMs for Practical Use .Challenges of training LLMs .Pre-training for domain adaptation. e- Practical Examples: .Prompt Engineering .Training Mini Models .Pros and Cons 3) Generative AI for Game Design a- Overview .Mechanic Design .Narrative Design .Controlling Players .Controlling NPCs .Generative AI in Level Design .Building Commentary Systems b- Prompt Engineering for Game Design c- Technical Requirements of Each 4) Generative AI for Narrative a- Overview b- Model Practices c- Prompt Practices d- Practical Task: Building a Character with AI 5) Generative AI for Mechanic Design a- Overview b- Prompt Practices c- Running Code in GPT d- Practical Task: GPT Mechanic Design 6) Using LLMs as Players & NPCs a- Technical Overview b- Design Considerations c- Practical Example: AI that ‘Plays’ Using GPT d- Practical Example: NPCs using LLMs 7) Building LLM Powered Level Generators a- Technical Overview b- Design Considerations c- Practical Example: Building Levels Using LLMs 8) Summary / Open Discussion
Objectifs de la formation
- Develop a broad understanding of the fundamentals of generative AI, and how it compares and contrasts with traditional artificial intelligence and deep learning
- Utilize the basics of prompt engineering practices used to interact with large language models and why it is important
- Employ core techniques and technologies in a variety of game production problem areas, ranging from non-player characters to player modeling and analytics, texture upscaling and procedural content generation
Profil des bénéficiaires
- Game Designers
- Level Designers
- Programmers
- Video Game experience
- Master Class in English
Contenu de la formation
-
Generative AI: An Introduction
- Fundamentals of Artificial Intelligence
- AI for Content Generation: A Short History
- Generative AI: Explained
-
Large Language Models: Theory and Practice
- What is a LLM?
- Understanding transformer architectures
- Prompting and Prompt Engineering
- Training LLMs for Practical Use
- Practical Examples
-
Generative AI for Game Design
- Overview
- Prompt Engineering for Game Design
- Technical Requirements of Each
-
Generative AI for Narrative
- Overview
- Model Practices
- Prompt Practices
- Practical Task: Building a Character with AI
-
Generative AI for Mechanic Design
- Overview
- Prompt Practices
- Running Code in GPT
- Practical Task: GPT Mechanic Design
-
Using LLMs as Players & NPCs
- Technical Overview
- Design Considerations
- Practical Example: AI that ‘Plays’ Using GPT
- Practical Example: NPCs using LLMs
-
Building LLM Powered Level Generators
- Technical Overview
- Design Consideration
- Practical Example: Building Levels Using LLMs
-
Summary / Open Discussion
Dr Tommy Thompson has worked as a professional software engineer since graduating from his PhD in artificial intelligence for games in 2010. After a brief stint working in the investment banking sector, he spent the next 10 years working in academia in the UK, fostering his love and enthusiasm for AI in games. Starting as a lecturer in computer science at the University of Derby in 2012, Tommy subsequently led the BSc in Computer Games Programming, during which time both the course and Tommy himself were nominated for and won several accolades from both the UK games industry and student body alike. After a period teaching at Anglia Ruskin University’s games courses, he ended his academic career after starting work at King’s College London in 2020 and taught artificial intelligence in the Department of Informatics. During this period, Tommy published over 50 peer-reviewed conference publications in the areas of machine learning for games, AI for non-player character design, and procedural content generation. In addition, Tommy presented at and helped organise numerous academic and industry events around AI for games. Outside of academia, Tommy is largely known as the voice of the YouTube channel ‘AI and Games’, Launched in 2014, AI and Games is an educational series that aims to provide insight into how AI is used in games, and how academic research is changing the start of the art. The channel has accrued over 10 million views since launch and achieved recognition both among the gaming press, and the wider games industry. In 2022, Tommy became a full-time consultant and games software engineer specialising in artificial intelligence for games. Since 2017 he has worked with companies on developing, refining, and communicating efforts in AI for video games. His clients have ranged in size and scale, from the likes of Intel and Ubisoft to indie studios such as Four Circle Interactive, Spilt Milk Studios, and Jaw Drop Games in the UK and the likes of Digital Lode in Australia. Outside of his own business operations, he is on the advisory board of the AI Summit at the Game Developers Conference (GDC).