AI Workshop "AI as Work"

EMW (European media sciences) and the Harun Farocki Institute are organising an international AI workshop in Potsdam. The workshop will focus on the relationship between AI and work in order to gain more insights into this technology, which is more than just a technical development.

The current dissemination of various AI platforms and services is particularly dependent on human labour. The workshop will focus on two forms of labour:

• on the one hand, the work of ‘data workers’, who are distributed around the world and usually process data for machine learning under precarious working conditions
• on the other hand, the workshop deals with the unpaid work of the production of texts and images that are automatically collected and processed by AI companies and thus become the basis for later ‘data work’.

The workshop will feature representatives from theory and practice (from the US and Europe) to shed light on and discuss this relationship between AI and work – and also to raise legal questions about working conditions and copyright. The combination of both aspects of work with and for AI should help to develop a better understanding of this technology and to discuss it together. Therefore, the focus of the workshop is on discussions between speakers and all workshop participants.

Admission is free. Details about the content and guests will follow below. You can sign up for the workshop messaging AI-work@emw-potsdam.de.

A workshop on the role of labour in artificial intelligence with Ariana Dongus, Krystal Kauffman, Nicolas Malevé and Tianling Yang

The rapidly increasing availability and use of generative AI since late 2022 – epitomised by platforms such as ChatGPT, Stable Diffusion, etc. – raises a number of questions. These include not only the widely known (optimistic or worrying) predictions about the beneficial use of AI in research, the alleged ‘end of truth’ or an expected ‘end of work’. Rather, such predictions are directly related to developments that affect current work – for and with AI. In contrast to the idea of purely machine-based systems, human labour is an essential factor and a prerequisite for practically every manifestation of contemporary AI: the human labour of data workers (who label huge amounts of data for the ‘learning’ machines) as well as the creation of textual and visual content that is captured and crawled on the Internet (to form the basis for human-driven learning as data).

For both forms of human activity, which are integrated into a supposedly artificial and autonomous technology, the question of legitimacy arises. The years of exploitation of human (pre-)work for automated processes of artificial intelligence does not legitimise such extractive practices. Rather, they require critical interventions from the legal and scientific communities, which relate to labour The circumstances of AI production and its conditions must necessarily be brought to the fore as clearly as possible.

The working conditions of data workers scattered around the world, poorly paid and psychologically burdened by labelling and filtering out violent and disturbing content, must be questioned and improved. Copyright on the data used (not only, but especially for the work of artists) must also be addressed. It is always about masses – about enormous amounts of data being processed by the huge number of data workers, currently estimated by the World Bank to be between 150 and 430 million. So both aspects highlight the importance of the production and use of data for this new form of technology. They help with the urgent task today of understanding how artificial intelligence works.

Initiatives such as the ‘Content Moderators Manifesto’ of 2023, the collective research project ‘Data Workers’ Inquiry’ launched in the summer of 2024, and the ongoing lawsuits by authors, screenwriters and publishers against Open AI and Microsoft are important steps. They provoke further questions, such as what can become data at all, where new data comes from, who has the interpretive sovereignty over the data and how, and what role the users of platforms like ChatGPT, Stable Diffusion and many others play in this.

The workshop ‘AI as Work. Conditions of Contemporary Image and Knowledge Production’ aims to bring together researchers, practitioners and activists Ariana Dongus, Krystal Kauffman, Nicolas Malevé and Tianling Yang. Precisely because the infrastructures and concrete functionalities of AI are often hidden, abstract and not easy to understand, the focus of this meeting is on discussion with all participants.

The workshop is part of the project ‘Terms and Conditions. The Legal Form of Images’ and a cooperation between the European Media Studies (EMS) degree programme at the University of Applied Sciences Potsdam and the University of Potsdam, the Harun Farocki Institute, Berlin, and the Academy of Fine Arts, Leipzig.