Generative AI as an object of research and method?
A workshop on using new text-to-image AI models in SSH research
Tuesday, 21st of February: 1400-1700
Faculty of Theology hall, Fabianinkatu 24, room 524 (in person)
Please note! The workshop has been rescheduled from December and there is room for a few more participants – sign up by 12.2. Now also including ChatGPT.
New text-to-image AI models such as Dall-E-2, MidJourney and Stable Diffusion are rapidly changing public conversation around the cultural impacts of AI. Behind the power of these new AI models is a new approach to training neural networks, which uses images and their captions in parallel. By capturing both visual and textual information contained in hundreds of millions of such text-image pairs, these models can now generate realistic images simply based on textual descriptions. The emergence of this new type of “synthetic media” has reignited critical questions about the cultural and ethical impact of AI on contemporary society. The training sets used to produce these AI models often include toxic and stereotypical content, and questions related to bias, and what content should be removed from these models, are currently a heated discussion in open-source communities.
Social sciences and humanities’ research methods looking at the cultural impacts of AI have often approached AI models from the outside; that is, as objects of inquiry that have focused on cultural significance of AI in different contexts. Less attention has been focused on how new AI models themselves could also be potentially repurposed from the inside as potential new tools and methodologies for research.
ABOUT THE WORKSHOP?
This workshop, facilitated by Anton Berg and Matti Pohjonen from the Helsinki Institute for Social Sciences and Humanities (HSSH), explores some of the opportunities and challenges of using these new text-to-image models for research. The emergence of a new generation of synthetic media tools such as web apps that facilitate the use of these AI models, however, now offers new opportunities for researchers to explore the models and the training data used in them.
The workshop provides an introduction to researchers interested in using these tools in their own research and some of the methodological questions raised by this. It first introduces popular text-to-AI models currently in use and the AI systems behind them. It then outlines different research projects where these models have been used to explore questions of representation and bias in AI systems. Finally, it provides hands-on examples of how to use these models, what platforms and tools are available, and what are some of the technical issues required to run them in a more systematic way. This includes an ecosystem of easy-to-use apps and web pages, as well as Google Colab notebooks and GitHub repositories for the more technically literate.
The workshop is targeted toward researchers interested in understanding these new text-to-AI image models as new cultural phenomena but also how these models themselves could be repurposed for more rigorous social sciences and humanities’ research interested in the cultural impacts of AI.
HOW TO PARTICIPATE?
To sign up for the workshop please send an email to email@example.com. Please use “Generative AI as object of research and method?” in the subject-line of your email. Sign up by 12.2.!
As we have limited spaces available, if you are interested in attending the hands-on (in person) part of the workshop, please also indicate the type of research you are doing and the reasons why you are interested in participating in the in-person workshop. The selected participants will be notified.
ABOUT THE PRESENTERS
Anton Berg is a doctoral researcher at HSSH with a background in digital humanities, religious studies, and cognitive science. Berg focuses on how modern technologies, such as computer vision and image recognition, generate and affect the formation of our social imaginaries. He is also affiliated with the Mediatized Religious Populism project (MERELPO), funded by the Academy of Finland, and the interdisciplinary Helsinki Social Computing Group.
Matti Pohjonen is a University Researcher at the Methodological Unit at HSSH. He recently formed a synthetic ethnography network for researchers interested in exploring new theoretical and methodological questions related to generative AI. He has also been exploring the use of generative AI in a Finnish Cultural Foundation-funded science-art project dealing with representations of the Covid-19 pandemic.