3: Ecosystem

As Chapters 1 and 2 lay out, developments in AI are reconfiguring organisational and artistic practices, while, at the same time, pointing to new spaces of opportunity to shape the interactions and expectations attached to the notion of public AI. What is currently unfolding may, in retrospect, appear as a ‘pilot’ phase, meaning that there is an urgency to set the foundations on which robust Art x Advanced Technologies (AxAT) strategies can be developed, and to start acting on these strategies.

Since 2020, Future Art Ecosystems (FAE) has been advocating for dedicated infrastructural development of the AxAT ecosystem. The research and insights that have informed strategic briefings to date, the ongoing projects by Serpentine Arts Technologies’ R&D Labs and its expanding network of collaborators, as well as production of new artistic commissions, have led to the recognition of key areas where ecosystemic development and investment are required. Strengthening of these areas is a prerequisite for the AxAT ecosystem to leverage its agency in negotiating the publicness of AI. Below is a review of strategic priorities in relation to these areas of focus and to the AxAT ecosystem’s engagement with the AI stack. This review is followed by a set of recommendations for cultural, civic, technological and policy-making actors for orienting the AxAT ecosystem to advance public AI.

Advocating for independent, in-house, and public sector-led production models as a key driver for AxAT practices

For the cultural sector, ensuring that technical literacy is a strategic priority will allow organisations to develop advanced production capabilities that make sense for their missions. Investment into capacity development and inter-organisational training programmes are two vehicles via which private and public funders can support the cultural sector in this process. Critically, however, funding should not be attached to the integration of specific systems; training curricula should be steered by independent or civic organisations.

Further, funders and cultural institutions who support artists working with advanced technologies need to attach key performance indicators (KPIs), and, ensuingly, resources, to robust and accountable production pipelines as seriously as they are currently being attached to outputs (i.e. artworks, exhibitions, and visibility). These production pipelines can be developed as general organisational capabilities for the wider public as opposed to being project-specific.¹1

Devising new benchmarks and systems for deeper and longer-term collaborations between organisations across cultural, technological and civic ecosystems

The scale of challenges and opportunities presented by all advanced technologies, and AI specifically, means that impactful intervention necessitates a plurality of specialisations across cultural, civic, legal, technical, and policy domains, in order to foster an environment where longer-term partnerships between and across contexts and sectors should be developed. Individual cultural organisations with the relevant capabilities should be encouraged to allocate capacity to engage in this specific type of partnership development, including setting up additional operational mechanisms (e.g. subsidiaries with missions that are legible to a distinct set of supporters). The cultural field is experienced and well-placed to act as a convening space; however, it requires a more dedicated approach to harnessing this capability.

Further, this type of activity can pave the way for the development of cross-sectoral protocols and policies for the adoption of AI systems at an operational level. Within the cultural sector itself, the state should champion projects that allow the sector to study and understand itself as a whole in relation to wider societal dynamics. Towards a National Collection, supported by UKRI’s Arts and Humanities Research grant, is an example of this phenomenon.²2 Research and innovation funding will be required for groups of cultural actors to organise around cross-cutting issues relating to new sectoral protocols for expanded data policies, the adoption of AI tools, IP standards, and frameworks for engaging with private technology providers.³3 ⁴4

Prototyping new models to achieve generative and equitable value distribution that supports producers and their communities

The current model of corporate, philanthropic, and public funding for the cultural sector sets up a framework where cultural organisations are seen to be at the receiving end of a value exchange.

In order to shift this model, or to develop parallel ones, risks need to be taken. This means supporting AxAT (as well as non-AxAT) artists whose practices are experimenting with new formats of investment in and distribution of their work. This could also go beyond backing individual artists, providing a platform for audiences and other communities to assert their agency by contributing and interacting with institutions in novel ways.⁵5 Organisations with the advanced production capabilities and the experience required to support AxAT artists are well-placed to develop and share these ownership and distribution prototypes across the AxAT ecosystem, but, more significantly, to share them across the wider cultural sector and beyond. Additionally, supplementary supportive measures need to be taken (e.g. wider cross-sectorial partnerships, the creation of new entities, wider funding) to allow for the cultural sector to shape new markets in this way.

Moving beyond footfall and media visibility as the dominant metrics of success, and devising new measurement systems for communicating the value of AxAT in society

Development of new AxAT categories of metrics and approaches is contingent on progress within the three areas discussed previously. For example, advancing production capabilities and ownership models that deliver thick public claims on resources such as data, modelling, and compute, and forming mission-driven coalitions with partners, will by default require a different set of metrics to assess the project than audience and media engagement through footfall and clicks. Long-term or cross-sectoral impact metrics would better capture the impact of such projects.

Asserting agency and strategic intention within the evolving AI stack not only requires the infrastructural foundations detailed above, but also a vision for the role that AxAT (and some parts of the cultural sector) should play as an intermediating space between technological and societal transformations. One of the unique features of AxAT, in contrast to many other art ecosystems, is how operational experimentation lies at the heart of the AxAT production process, both for artists and organisations. The development of AxAT projects straddles technical, legal, operational, and creative processes. They lead not only to the emergence of a new artwork (or other form of public output) but to various insights emerging from the entanglement between these processes and associated cross-sectorial communities. What would it mean to commit to these prototyping affordances of the AxAT ecosystem for the advancement of public AI?

Functionally, what AxAT has to offer is the development of practices and potential new organisational forms for the data and model layers of the AI stack (and, to a lesser degree, the compute layer) as public resources. This is usually delivered in the context of projects that involve the typical cohort of AxAT stakeholders: artists, arts institutions, technologists, technology companies, public bodies, audiences, specialist communities (e.g. researchers, experts from other fields and universities), and funders. The three vectors listed below offer only a handful of potential sandboxing experiments that the authors of FAE have tied to the creative R&D focus of Serpentine Arts Technologies projects. There is ample remit within these vectors for a multiplicity of approaches and stakeholders.⁶6

Speedrunning and developing early operational frameworks for data stewardship, data bargaining, data valuation, and stakeholder coordination of data

While the contested scraping of the open internet has been a norm in AI development until the present, new data markets, provenance standards, data brokers, and newly formulated relations to data subjects, who collectively bargain for the value of their networked data, are likely to emerge. These marketplaces will likely be largely automated but will require new platforms, vendors, pricing and validation mechanisms, and stewardship protocols. This presents an opportunity to build a new landscape with thick public resource distribution. AxAT projects can become laboratories for testing all the components of a data market with a variety of stakeholders, determining ownership, governance, advocacy and pricing mechanisms of different datasets, and how they are informed by the data relations of the cultural context. Working with research initiatives in university, policy, and industry settings can offer an opportunity to bridge these insights with policy and design work that will inform future data markets.

Testing out networked IP, recombinant IP, and creative licensing as a means of evolving and/or departing from the inherited copyrights-focused frameworks for protecting IP within the cultural context

We are in a historical moment where individual creators, legacy institutions and media (e.g. The Natural History Museum, The New York Times), entertainment corporations (e.g. Disney), and some platforms (e.g. Reddit), find themselves, however briefly, within a relatable struggle to assert their rights in an uncertain climate relating to IP ownership and the governance regime for training AI models. In a world where infinite media can be generated without specialist technical know-how, users will want to find ways of accessing and remixing media at a new depth and scale. Users will generate new content inside existing worlds, or build their own with derived assets, or some combination of both. The move towards a highly personalised media landscape means that IP holders of the current media landscape (artists, institutions, and conglomerates) may need to experiment with different reconfigurations of ownership. For example, one potential reconfiguration could be motivated by creating more flexible licensing frameworks to ensure that users can personalise and fork characters, lore and worlds, and reintegrate new recombinant media into their social online interactions. More generally, new online media dynamics will necessitate participation mechanisms that protect users from extractive AI training practices, whilst still allowing for circulation as a norm for online interactions.

The focus here is on the exploration of constructing the legal-technical layer for new media interactions through the development of networked and/or recombinant IP (i.e. IP that is sensitive to recombinant media creation as a new normal). New IP categories and their technical implementation can build on data governance experiments in the art and civic contexts, and extend to the licensing of small-scale models where a narrow remit means they can be more precise, experimental, and less resource intensive. The proliferation of such trusted models will underpin new economies and services that public organisations are well placed to provide.⁷7 These new frameworks can take the networked nature of artistic production, creative inspiration, and audience interaction dynamics into account to map out relational, collective, and fluid ways of assigning different levels of contribution, attribution, ownership and rights. AxAT organisations, artists and projects could combine efforts with groups who can leverage these learnings to lobby and advocate at policy and developer community levels.⁸8

Supporting the development of blockchain x AI digital economies for artists and new AxAT organisations

Virtual production and blockchain integration for the creative economy are two (potentially overlapping) spaces where the AxAT ecosystem has the opportunity to shape the integration of AI systems. While experiments in new IP paradigms and public data market mechanisms will be critical for setting some of the terms for a space that is being completely transformed by AI, how this intersects with virtual production and blockchain technologies will then redefine the roles and rights of ‘content creators’.

Certain AI tools will soon be proficient at creating 3D virtual assets and self-programming virtual worlds. Coupled with open source interoperability mechanisms such as the Universal Scene Description (USD) file format, a major transformation of production pipelines for various media industries and artists is likely underway, with the potential to disrupt the huge film, TV, advertising, and online marketing labour markets.⁹9 ¹⁰10 Users, and thus artists, will soon be in the position, at least for a short period of time, where they can access exactly the same open source, and often interoperable, toolkits as those used by industry as products are tested and improved by the tool and platform developers. This will shift the models for how virtual worlds, animation, 3D, and any other modes of CGI production are structured, ultimately lowering the costs for production (though potentially only temporarily) with less requirement for niche technical specialisms.

As new operational and business models will start to emerge at the scale of media and entertainment industries, the art field’s capacity to be positioned alongside industry players will be contingent on a robust AxAT ecosystem that can incubate new skills and production pipelines, and lobby for how new economic and distribution models will benefit a broad cross-section of creative sectors and society.¹¹11

A similar dynamic may unfold as the market for the integration of blockchain and AI technologies starts to emerge.¹²12 Blockchain’s core affordance of providing a decentralised immutable ledger for various information flows such as decisions and transactions, could address some of the systemic risks and challenges that AI poses for tracking of provenance, attribution, and verifiability. The art context served as an early prototyping space for blockchain based digital markets, as well as niche cultural experiments around governance, smart contracts, mechanism design, and decentralised autonomous organisations (DAOs). The confluence of these factors makes AxAT a fertile context for trialling the ways in which blockchain’s and AI’s technological capabilities can mutually support each other.¹³13 For example, prioritising how the provenance of recombinant media worlds can be traced using blockchain technology could become a game-changer for artists and users more generally, allowing for more layered and nuanced compensation frameworks to emerge for recombinatory contributions to new media worlds (e.g. datasets, models, and model artefacts).¹⁴14

Utilising cultural reputation, technical literacy, insight, and strategic understanding of the technology sector to negotiate on behalf of the cultural sector

Access to deeper levels of AI systems than what is offered by increasingly consumer-facing AI products and services built on closed foundation models will be critical for artists to work with these systems as creative media, and for the cultural sector to lobby on behalf of creatives and the sector. Meanwhile, for artists and institutions who want to train their own models, access to compute, or partnerships with compute providers, will be essential. In order to ensure that compute privileges don’t only reach those who are able to negotiate for them, a campaign for ‘public cultural compute’ should include leading AxAT organisations and actors, including setting up a public cultural compute bank.

Plural and concerted ecosystemic action today means that the AxAT ecosystem can articulate de facto precedents that either serve as experiments, or help to shape forthcoming legislation and cultural norms around AI. Outside of the EU's AI Act, which is expected to enter into force in 2024, few jurisdictions have taken a comprehensive approach to regulating AI. The UK has set up a number of AI-related bodies, but has yet to legislate.¹⁵15 As these bodies concretise policy, there is an opportunity for the cultural sector to make use of the AxAT ecosystem in order to test inter-organisational policies and standards, as well as to leverage the ‘Brussels Effect,’ where necessary, in order to embed measures that allow public AI to blossom.¹⁶16