GenAI & Creative Practices: Past, Present and Future – Day 1

Keynote Session 1 – Baptiste Caramiaux. ‘Steering the Unknown’

• What can an artist control and what can they not control during the creative practice?
• Vera Molnár as an example to opne the discussion
• Induction – a system to infer behaviour. Example of ocean scenes and how AI uses LLM to find a similar structure (an abstract space – model or manifold). Output to latent space (abstract dimension). Form the model, one can pick a point and have a generated version based on teh original input.
• The abstract (output) becomes a new form of expression
• Inductive practice: gesturing as a menas of creativiity based on LLMs
• Using LLMs to create movement in work – Anna Ridler, ‘Tulips’, crafting the output but also creating the model that the artwork was trained on. THis is an important aspect of ‘Induction’
• Mario Klingemann – moving image artist mixing modalities
• Terence Broad, exploring the concept of the gaze
• Forms of agency – new design material, new modalities, “induction enables many practices, generation is only one of them”
• Drifting Agencies – Ai expands the unknown much faster than our ability to steer it
• Sean Michaels – “Do you remember being born?”
• Documenting the undocumented? New narraives, new meanings – integrating AI with people to understand our histories. Could this push our agency (Lenny Martinez undertakgin PhD here)
• Disempowerment: Reducing agency, most of the practice is through platforms- we cannot change them, we cannot train them, we may be able to calibrate them. Normative behaviours, the ways in which stories are represented. Narratives how the scale impacts the ways in whcih we understand the creativity aspect of AI.
• Induction as a design material – this is the tool or the skill of GenAI, not simply the application of the existing models (bigger is not better).
• Creative AI is a power structure – think through moral, cultural and structural dimensions. W way to understand the agency and centralisation of labor. Narrative representation. Norms and identities, accountability, authorship.
• The Future is Collective – creative agency is through collective organisation (this is increasingly difficult in our current political and economic ecologies).

Parallel Session 1 – Deepfakes, Clones & Imposters

Creative Violence in the Age of AI: Deepfakes, Misogyny, and Community-Based Response in Mexico, Payal Arora,

Ana Miranda Mora

• Increased mysogonistic content, mostly porn, in AI spaces – this project partnered with Google to examine
• NCII – (Non-consensual intimate image) also known as Image-based Sexual Abuse (IBSA),
• Mapping the system, working on the language that is used
• Reputational harm, financial harm, mental health harms
• A social way to connect with these activities is through the language used, revenge or sextortion gains traction, whereas NCII does not
• Deepfake porn in Mexico is not used, it is digital violence and intensifies sexual harassment, 10 million women in the last year have experienced this
• 13 in depth interviews, survivor sample young women between 20 and 4, middle class, educated in urban areas,
• Face-swap technologies, AI-generated nudes
• Community based responses: Olimpia Red de apoyd digital.

Deepfakes under Strict Government Regulation: The Role of Social Influence and Perceived Playfulness on Chinese Centennials’ Intention to Use AI-Generated Media – John Maina Karanja

• Missing in action

Roundtable 2.2: Global Perspectives on GenAI & Cultural Production

Thomas Poell, Neha Bhatia, Godwin Simon, Lorena Caminhas, Tom Divon

Thomas Poell:

• The project has begun, but it is very much rooted in Western languages/concepts. The mission of today is to expand on this to better understand from a post-colonial perspective
• Much of the non-Western literature emerges from China, which is demonstrable of the ways in which the ‘theory follows the money’ – can we provincialise the field of cultural production

Godwin Simon

• Field production in Nigeria: how it works, & precarity
• Cultural competence in cultural production: ways to express through co-creation that reflects the authenticity of the industry

Neha Bhatia

• GenAI in Indian production and background: creativity, precarity and experiences of GenAI production
• Dispute around if GenAI is at all original – however, GenAI is trained and produces something novel
• What is creativity and how is it undertaken? (New, novel and somethimg surprising)
• The value of creativity is decreasing because it is rapid and fast-turnaround

Sin Melata

• GenAI is used as a tool within the cultural production process
• Ethical and moral issues emerge when the timeframe of the production cycle becomes tight
• ChatGPT being used as a writing tool within India

Lorena Caminhas

• Brazilian influencer industry
• Deeply shaped by platform logics – not a new thing, an extension of this type of cultural production
• It’s creativity and brand identity that creates authenticity
• The takeaway is that users can be seen as more authentic through the use of AI
• AI in Brazil is enabling greater creativity – this is presenting an opportunity to make a living in the content creation space, but this doesn’t change the impact of the influencer industry (brand deals, management etc.)

Tom Divon

• Recalling a pre conference from Brazil that brings to the fore things like regulation, labour, and restructing of systems
• Conditions under which creativity takes place, GenAI in the process, and then the political economy that enables/inhibits, this is where the decision making point sits
• Imaginaries as a way to understand the whole system together
• Resistance is often not through refusal, but rather renegotiating, positioning – these are systems that users cannot just move out of.

Thomas’ latest article here.

Q: data sovereignty: what about a third way and how that might integrate with the other countries in this project? Align with Vietnam

Thought process/book idea: Induction sovereignty – can cultural production be possible without induction sovereignty? If we are using GenAI as a production tool, we are already using a model that has been designed, trained and learned through a LLM training process. At best, it can only be calibrated to include localised nuances. Is this where PSM values become important?