DeepMind announces AI 'Dramatron' that co-writes scripts for movies and plays
Google's sister company,
Dramatron
https://deepmind.github.io/dramatron/
GitHub - deepmind/dramatron: Dramatron uses large language models to generate coherent scripts and screenplays.
https://github.com/deepmind/dramatron
[2209.14958] Co-Writing Screenplays and Theater Scripts with Language Models: An Evaluation by Industry Professionals
https://arxiv.org/abs/2209.14958
According to DeepMind, Dramatron can generate long, cohesive texts with existing large-scale pre-trained language models. By entering a simple summary or idea of the story, you can create and edit the script by interactively generating the title, description of the characters, the plot of each scene and the point of scene change, the stage setting, and some lines. It seems that I will present a hint to human writers.
Want to find out more? The team will be presenting this research at #NeurIPS2022 :
—DeepMind (@DeepMind) December 9, 2022
???? December 9
⌚ 3pm CST https://t.co/ulHOnbHoEh @ml4cdworkshop pic.twitter.com/zTTbKQcDnS
DeepMind is a source of inspiration and exploration for writers, and is designed as a writing tool. I think it's a tool. Also, Dramatron's hierarchical story generation structure does not correspond to the writing process of all writers.
To evaluate Dramatron's usability and features, DeepMind reported that 15 screenwriters and playwrights held a two-hour user study session and co-wrote a script with Dramatron. A playwright is said to have performed four plays after editing a script written jointly with Dramatron.
✏️ We interviewed 15 industry experts including playwrights, screenwriters and actors who produced work using Dramatron.
—DeepMind (@DeepMind) December 9, 2022
Canadian company @TheatreSports edited co-written theater scripts and performed them on stage in Plays By Bots to positive reviews. https://t.co/FlGzIdCuqX pic.twitter.com/6gqnB8e1L9
Please note that Dramatron may generate offensive text due to prejudices and stereotypes contained in the material. Therefore, mitigation measures have been proposed to estimate the 'toxicity' of the output text using ' Perspective ', which automatically detects hate speech on the Internet by machine learning, and filter it.
Also, since the elements of the learned data may be output to text as they are, using Dramatron's script as it is may be plagiarism. As one mitigation measure, DeepMind proposes that human authors who are co-authoring work search for and edit the plagiarized parts from the output results.
Related Posts:
