OpenAI GPT-2 writes alternate endings for Game of Thrones

Chintan Trivedi
Towards Data Science
12 min readMay 23, 2019

--

I trained the GPT-2 language model on GRRM’s book series “A Song of Ice and Fire” and let it complete the HBO show’s storyline. Can it do better than HBO’s season 8 train-wreck?

Game of Thrones season 8 storyline has left its fandom divided with millions of fans disappointed by its rushed and unsatisfactory ending. Disclaimer: Opinions expressed in this article are solely my own.

Game of Thrones season 8 storyline has left its fandom divided with millions of fans (including myself) disappointed by its rushed and unsatisfactory ending. Many people are calling for a rewrite of the last season for the TV show while others are coming up with their own version of the ending to get a satisfactory closure. Rest are waiting for GRRM to finish his books hoping he will provide a more compelling build-up to the characters’ unconvincing actions, but this might take many years. Thus, I wanted to let an AI learn GRRM’s writing style and make it finish the ending of the show.

In this article, I’ll explain how we can achieve this by using the text generation language model called GPT-2, introduced by researchers at Open AI in early 2019. Then, I’ll share the model’s predicted outcome for the storylines of three major villains of this season: the Night King, Cersei Lannister and Daenerys Targaryen.

Generative Pre-Trained Transformer 2…

--

--

Published in Towards Data Science

Your home for data science and AI. The world’s leading publication for data science, data analytics, data engineering, machine learning, and artificial intelligence professionals.

Written by Chintan Trivedi

AI, ML for Digital Games Researcher. Founder at DG AI Research Lab, India. Visit our publication homepage medium.com/deepgamingai for weekly AI & Games content!

Responses (2)

What are your thoughts?