“Using a machine as a collaborator, an algorithm as a sketching tool, or buildings as canvases – these were always fundamental ideas in my imagination, inspired by science fiction,” says Anadol. Born in Istanbul and based in Los Angeles, Anadol studied at UCLA before making his first forays into the nascent field of “data painting” in 2008. Since then, his immersive video and installation works – now created using machine learning tools and often presented at epic, audience-subsuming scale – have been exhibited at New York’s MoMA; the Venice Biennale of Architecture; SXSW; and recently adorned the enormous outer face of the Las Vegas Sphere. His newest work, Glacier Dreams, turned his imagination into reality.

The piece, a room-sized installation encompassing light, sound and scent, debuted at this year’s Art Dubai. Confronting viewers with both the beauty of glacial processes and the dangers of ecological damage, the work uses machine learning to create a medley of ever-morphing snowscapes. Eventually, the vast ice structures depicted begin to melt and come apart, collapsing in a pointillist wash of blue pixels. Though the shapeshifting environments shown are impossible, a result of Anadol’s careful tinkering, their link to reality is clear: the basis of the installation is a massive corpus of visual data, much of which was personally collected by Anadol and his team during research trips to the real-life volcanic glaciers of Iceland.

Pushing the boundaries at Art Basel 2023

Commissioned as part of Julius Baer’s NEXT initiative curated by Hans Ulrich Obrist, Artistic Director Serpentine, Glacier Dreams works as a traditional video piece, too; at Art Basel 2023, with Anadol’s usual appetite for the spectacular, it was projected onto the exterior of the Theatre Basel. 

“We used more than 100 million images in Glacier Dreams, taken from across Iceland, Antarctica and Greenland,” Anadol says. Generative artificial intelligence tools were put to work on the amassed material; Anadol explains that he and his team eschewed ready-made models in the preparation of the piece, preferring instead to train their own digital learners to perform the tasks required. While such work can be a challenging, “uncomfortable zone” for creators, Anadol says that support from Julius Baer allowed him to dig more deeply into the unique and fast-moving field of engineering bespoke models.

“Training our own AI models from scratch allowed us to push the boundaries,” Anadol says. “AI algorithms are designed to predict and mimic reality – they are trying to create ‘real’ things. In my practice, I’m trying to find the fantasy. The dreams, the hallucinations, which are not the fundamental objective of AI research.”

Machine learning, predictive technologies and the financial services industry

Supporting and collaborating with technologically-minded artists like Anadol is a natural fit for Julius Baer, as the financial industry at large undergoes a major shift. While automated or algorithmic trading has been a common practice among banks and hedge funds for decades, recent years have seen financial companies using predictive and generative tools to augment a broad array of day-to-day operations, with many firms prioritising the development of machine learning-related applications. 

“For a private bank, a key issue is making sure that you always provide the best advice to your client at the right point in time,” says Nicolas de Skowronski, Julius Baer’s Head of Wealth Management Solutions. A fan of Isaac Asimov, he is sanguine about the benefits predictive technologies can provide for customers.

“AI can assist our relationship managers, providing a fingerprint of the client – their characteristics, trading behaviour, risk appetite, asset allocation – and we can screen [these traits] against opportunities, helping the relationship manager form their advice,” de Skowronski says.

Ethics, transparency, and the way forward

Though the art and business worlds approach AI and related technologies from differing starting points, practitioners in both realms encounter their fair share of challenges in using such applications. Across contexts, the deployment of algorithms is often met with criticism: artists may be accused of dishonesty or laziness, while financial companies can run afoul of emerging regulation and incur reputational risks in cases where new tech is considered opaque. As early generative tools aimed at consumers – like OpenAI’s much-discussed ChatGPT – continue to build steam, public debate over machine ethics and job security has swelled.

“There are a lot of fears about AI, and they are justified,” says de Skowronski; he recalls a tricky recent parenting moment he had to navigate, where one of his children asked why they should study for an upcoming exam when ChatGPT could simply provide the answers needed. At Julius Baer, de Skowronski continues, key concerns among staff are the importance of data privacy; the avoidance of model bias, and the problem of explainability, where firms need to understand why and how a given model arrived at a particular result.

For Anadol, transparency is a similarly central issue. “In our exhibitions, we always expose our processes,” he says. “We are very open and honest about our use of AI – and every AI artist, I think, has a responsibility to share how [their techniques] work. We share where our data comes from, which algorithms we use and how we fine-tune the AI. This is all as open and exposed as possible, and it helps the audience connect to the artwork.”  

Entering the age of human-machine co-creation

As committed sci-fi enthusiasts, both Anadol and de Skowronski are happy to speculate on the sorts of futures AI could produce. The “holy grail,” de Skowronski says, would be the arrival of the machine superintelligences currently confined to fiction; the ‘artificial general intelligence’ (AGI) able to meet or surpass the cognitive capacity of humans. Anadol, meanwhile, is most excited by what he describes as the vast artistic promise of complex models and their ability to work alongside human creators.

“I believe we will soon explore not only text to image, but text to sound, text to scent and beyond,” he says. “We’re entering a new phase of human-machine co-creation.”