DeepMind trains AI to draft rough film scripts

DeepMind has created an AI tool, Dramatron, which generates character and location descriptions, plot points, and dialogues for film scripts.

Introducing Dramatron, a new tool for writers to co-write theatre and film scripts with a language model. 🎭

Dramatron can interactively co-create new stories complete with title, characters, location descriptions and dialogue.

Try it yourself now: https://t.co/86zalyXBWQ pic.twitter.com/qWkUH4NoWF

— DeepMind (@DeepMind) December 9, 2022

\n

According to the developers, writers will be able to compile, edit, and revise Dramatron texts into the script they want.

DeepMind researchers asked 15 experts from the film industry to assess the algorithm’s performance. They said they would not use Dramatron to create a full play. In places, the AI’s outputs could be boilerplate, the experts added.

However, the experts noted potential value of the tool for \”generating creative ideas.\” They said Dramatron could help build a world or explore alternative approaches by altering elements of the plot or characters.

✏️ We interviewed 15 industry experts including playwrights, screenwriters and actors who produced work using Dramatron.

Canadian company @TheatreSports edited co-written theatre scripts and performed them on stage in Plays By Bots to positive reviews. https://t.co/FlGzIdCuqX pic.twitter.com/6gqnB8e1L9

— DeepMind (@DeepMind) December 9, 2022

\n

As part of the experiment, the Canadian company Rapid Fire Theatre staged four plays that used \”heavily edited and rewritten scripts\” created with Dramatron.

According to DeepMind’s representatives, in the performances, experienced actors with improvisational skills gave meaning to the algorithm’s plays through performance and interpretation.

Overall, the productions received positive reviews.

The developers noted that Dramatron can output fragments of text that were in the training data used to train the language model. Using a script created by the algorithm could lead to accusations of plagiarism, they added.

\u201cOne possible way to mitigate the consequences is to search for substrings in the output data with a human co-author to help detect plagiarism,\u201d the company said.

\n

Earlier in December, DeepMind published results of testing the AlphaCode code-writing algorithm. The developers said the model had reached the level of a \”junior programmer\”.

In November, the lab developed AI agents that interact with people in a natural way and learn from them.

In October, DeepMind, with the help of artificial intelligence, accelerated matrix multiplication, solving a fundamental problem in computer science.

Subscribe to ForkLog AI news on Telegram: ForkLog AI — all the news from the world of AI!