20 May 202111:30 - 15:30Online event

Description

pyschograph_edited.png

Ghost fictions: Automatic Writing in the Age of Machine Learning 
Computational techniques for generating ‘natural’ language through statistical models created using huge datasets in conjunction with neural networks have advanced rapidly in recent years. These ‘Large Language Models’, such as GPT-3 and BERT, can be used for a wide range of tasks with little or no modification, including writing short stories, answering philosophical questions and apparently proposing potential medical treatments — although GPT-3 did have some difficulty with the question “how many eyes does a horse have?”
 
This CDH Open workshop will delve into the production of such ‘synthetic’ or ‘ghost-written’ texts combining insights from speculative fiction, computer science and digital humanities, through a combination of demonstrations, discussion and hands-on experimentation with automated text generation. 
 
No knowledge of programming is required. Participants who wish to take part in the practical experiments will need a Google account
Image credit: Image modified using the iOS Prisma app, which is based on an open source implementation of the neural network style transfer algorithm from Gatys et al (2015) https://arxiv.org/abs/1508.06576 
 
Contributors:

Dr Anne Alexander – Director of Learning, Cambridge Digital Humanities, University of Cambridge
Professor Caroline Bassett –  Professor of Digital Humanities and Director of Cambridge Digital Humanities, University of Cambridge
Professor Alan Blackwell –  Professor of Interdisciplinary Design, Computer Laboratory, University of Cambridge
Dr Joseph Walton – Research Fellow in Digital Humanities/Critical and Cultural Theory (Media and Film) University of Sussex

 

Live session 1: 11:30am – 1:00pm
Lunch break / experimentation lab: 1:00 – 2:30pm
Live session 2: 2:30 – 3:30pm
 
This event is free, booking is now open on Eventbrite

Cambridge Digital Humanities

Tel: +44 1223 766886
Email enquiries@crassh.cam.ac.uk