Shyam Sudhakaran on MarioGPT

Published March 19, 2023


2 min read

Image of Mika Senghaas

Mika Senghaas

Hi all 👋! Do you remember MarioGPT? 🍄 A few weeks ago, our reading group had a delightful time exploring its contents. In case you missed it, MarioGPT is a fine-tuned GPT-2 model with the impressive ability to generate Super Mario Bros levels when provided with text input. This marks a milestone in the field of controlled game-level generation through AI, as it is one of the first generative AI models in the text-to-level domain. If you want to dive deeper into its workings, make sure to check out our MarioGPT Blog Post and the Original Paper.

Animation of MarioGPT Level Generation

Last week, we were privileged to have a virtual meeting with the main author of the paper, Shyam Sudhakaran (LinkedIn), who joined us from the United States. Shyam presented his perspective of the paper as an author, and we engaged in an enlightening discussion about the research’s motivation, international research collaboration, technical details and future work that builds on top of the paper.

It was a remarkable experience for all of us to see a paper that we have all read be explained by the main author. Huge thanks to Shyam for joining us and enriching our understanding of his work. In the future, we hope to talk to many more authors of papers, as it gives insights into the thought process behind research that cannot be gained by only reading about it.

Online Meeting with Shyam Sudhakaran

Shyam presenting MarioGPT in front of AITU 🙌

📣 Stay in touch

That’s it for this week. We hope you enjoyed reading this post. 😊 To stay updated about our activities, make sure you give us a follow on LinkedIn. Any questions or ideas for talks, collaboration, etc.? Drop us a message at