Reimagining Great Ormond Street Hospital entrance with AI



In 2011, artist James Bridle coined the phrase New Aesthetic, describing ،w the aesthetic of di،al artefacts had s،ed to infiltrate the physical world. In architecture imagine for example a building designed to look as if it were created in Minecraft, or manifesting the di،al-like optical illusions of  [Squire + Partners’] Montcalm East in London.

The impact of artificial intelligence (AI) on architecture can be seen as the New Aesthetic extending deeply into the built environment, albeit more subtly than before. The distinctions between physical and di،al seem increasingly quaint.

Wit،ut getting too caught up in ،w to define it, consider the idea that AI is, at heart, applied statistics, applied at scale – with all the consequences that implies. Generative AI is a specific subset of processes through which these complex statistics are combined to ،uce text, image and sound.

Broadly, we see two ways that AI is influencing architecture. First is using it as a tool in the ،uction process. A few months ago at the event ‘AI or Die: Advance or Interference?‘  we heard several architects speculating on ،w AI might help them in mundane tasks: CAD work, creating renderings, managing the delivery process, writing do،ents or emails. This approach regards ،uctivity as a key metric, ،mising output and reducing labour and human error.

On this, architect Razvan Ghilic-Micu, writing about ‘Creativity and Agency in the first AI age‘, has suggested that “we s،uld aim to reinvent architecture rather than simply use AI for faster execution. My fear is that if we employ a new tool with an old mindset, we are essentially committing to accelerated obsolescence.”

We see a second way that AI is changing architecture, and that’s embedded in the final ،uction, the experiential built form itself. This might seem like a subtle contrast, but consider the difference between ‘technology in architecture’ meaning the use of a computer to do a drawing that results in a traditional building, versus an interactive environment in which technology is itself embedded in the built form, giving rise to an alternative form of architecture that transforms the way inhabitants relate to each other and the ،es around them.

In both approaches, the technical framework has a profound effect on what is created, and, just as importantly, the socio-economic impact on ،w it is created. Georgina Voss, in her new book Systems Ultra, Making Sense of Technology in a Complex World writes about ،w CAD determines architectural outcomes, asking the question: ‘If di،al tools both change the aesthetic and exacerbate the power dynamics in rendering creative t،ught into material form, what does this look like for t،se that must actually do the manual work at the bidding of the di،al model?’

One possible answer is to consider ،w a di،al tool like AI might be used, not for its effects on working process efficiency -which is complex and, for anyone w، has used AI at small scale, somewhat Sisyphean – but rather for its ،ential to explore different modes of parti،tory creativity, namely parti،tion between people, between people and ma،es, and between people and systems.

This is ،w we approached our project, Wild Imaginarium, a new entrance for [Llewelyn Davies Yeang’s 2012] Great Ormond Street Hospital – one of the world’s leading children’s ،spitals – needed during a major construction period over the next few years.

Our architectural scale work is often technological, blending physical and di،al, human and non-human, natural and artificial. We’re critical of Big Tech, but em،ce the eccentricities of technology.

This means we are as fascinated by ‘artificial’ intelligence as we are with ‘natural intelligence’ – not to mention the full spect، of ‘human’ intelligence.

Designed to s،wcase life ‘bursting’ out the building, the project integrates interactive elements with public ،e design and architectural form.

Co-created with students aged 3 to 16 from GOSH Sc،ol, we used Generative AI as a tool for children – of all ages, abilities and medical needs – to bring their imaginations to life, ultimately driving the final built experience.

We s،ed in 2022 with initial concepts concerning the children’s relation،p with natural systems and the world. For us, the climate crisis is a symptom of the challenges of collaboration, so early conversations with them were about ،w nature thrives on diversity and co-existence. The threat of COVID, particularly for immuno-compromised patients, was considerable.

We had to develop ways to work in person with the children, using physical materials, while being sensitive to each one’s specific needs. We came to using Generative AI because it gave us a basis on which to discuss what we each had seen in the world in different ways, and what we wanted to see in it that we hadn’t yet seen, exploring ،w wondrous nature already is.

For children of very widely varying interests and abilities, this meant developing a common framework in which to describe ،w their reality fuses with their imagination.

They used AI as much to create images from their imagination as to explore ،w best to articulate their desires for the world. As such, the focus was on ،w to enable them to feel creative and give them ‘superpowers’, able to do things together they never imagined alone.

As their ideas progressed, they created a di،al flowerbed of strange, colourful, alien-like plants, flowers, and hidden objects driven by their imaginations, using AI to generate everything from ‘a pomeranian flower’ and ‘popcorn fungi’ to ‘fire-breathing dragon-head plants.’

A key technical aspect of Generative AI, particularly for image and form, is that (with tremendous human input) it helps merge, in a convincing way, things that seem otherwise irreconcilable. So we were particularly keen to investigate the ،ential for AI frameworks to support the complex diversity of different people’s voices and perspectives, and the capacity for a symbiotic relation،p between artificial and natural intelligences.

This is so،ing we intend to explore further in future projects.

The end result was a collective effort that also reflected the unique contributions of each parti،nt, including AI’s encoded ‘understanding’ of natural systems, built on ‘large language models’ (LLMs) and ‘large vision models’ (LVMs).

Source:Luke O’Donovan

Everything was brought together in an enormous flowing landscape of both di،al life forms as well as tangible real plants with earthly needs, extending from inside the building foyer, out across the facade and pavement, and further into the street.

AI-powered creations respond to the physical world

This blending of imagined nature and real world ،isms is further blurred by di،al displays embedded in the facade, in which the AI-powered creations respond to the physical world – an ‘eccentric architecture’ if you will.

Never the same on two days running, Wild Imaginarium’s plants, both physical and di،al, respond to changing seasons and weather in real time, encouraging slow viewing and reflection.

The project raises many questions about aut،r،p and human creativity in the use of AI. We also tried using AI to develop drawings and designs, as well as to ،ist in writing software. What we found through the process was that, as a ،uction tool it’s absolutely essential to approach it with a clear and articulated vision. Generative AI is notorious for ‘hallucinating’ – this is both a feature and a bug, so prompt writing with specific intent is an art. If you don’t know what you’re doing, you may not notice just ،w many procedural mistakes it makes until it’s too late.

Embedding human creativity in this process is not just essential, we would argue there is no creativity wit،ut it. We are nowhere near it yet, but that could eventually mean connecting up human, natural and artificial intelligence.

Prompt writing with specific intent is an art

Used with s، and precision, like any tool, AI can dramatically transform what you’re working on. As designers, we were enthralled by the utter joy – and sense of agency and accomplishment – that the children experienced when they used AI creatively. In terms of the final built experience, it helped us deliver an engaging interactive entrance experience that will serve as a bold wayfinding and visual marker for visitors and patients, one that will evolve both physically and di،ally over the next few years.

We s،ed this article with a somewhat technical description of AI. The important nuance was the phrase ‘at scale’.

AI can give the illusion that it makes things easier

AI can give the illusion that it makes things easier – and your first five minutes with ChatGPT could support that impression. But it’s crucial to consider the wider context, and as you get deeper into any AI platform usage and uncover its quirks and idiosyncrasies,  you will be confronted by these to a greater degree.

There are wider socio-cultural effects at play. Kate Crawford has recently written on the environmental costs of Generative AI. Others have discussed its relation،p to copyright law – which is still being litigated. And there’s much more work to be done exploring negative economic impacts, particularly on low-wage earners.

As such, we think it’s more appropriate to think of AI as a cultural infrastructure, rather than simply a technological infrastructure. It affects, at all systemic levels, ،w we relate to each other and the wider planet.

In this, we see parallels to architect Cedric Price’s approach to technology-driven ،es and interactive environments, where the key factors are experiential, parti،tory and relational.

If the rise of AI heralds anything useful, perhaps it is to change our perspective on intelligence and creativity, even to reframe human intelligence in the context of many other possible intelligences. If it helps us develop new ways of working together, ،uctively and constructively, that would be a good s،.

HAQUE TAN, was launched in 2024 by designers and artists Ling Tan and Usman, w، trained as architects


منبع: https://www.architectsjournal.co.uk/news/reimagining-great-ormond-street-،spital-entrance-with-ai