Artificial Intelligence, Architecture and Aunty
Updated: Apr 26, 2019
AI in architecture today is focused on the wrong parts of the process .
AI will augment and to some extent replace clients.
AI and big-tech is the best hope to avert environmental catastrophe.
Architects' opportunity will be to create a new architecture reflecting an AI-managed world.
Artificial Intelligence: Machine systems of computation that autonomously gather and analyse data necessary to create a desired outcome, the parameters of which they may set themselves.
Greater, and more learned minds than mine have grappled with the definition of AI, since the 1950s, and it is not a settled matter, so feel free to disagree with the above. My definition is more demanding than the limited expectations that seem to be the consensus now, wherein, mimicking human perception and organisation qualifies as AI.
Although complex tasks, such as facial recognition, are clever, they are not intelligent. Reading this does not make you human. The constant chatter of dispute and agreement in your mind, as you do so, does that. Intelligence may not require quite such a high level of mental activity as humanity, but without free-association, I do not think it can exist.
For the purposes of this blog, I limit my definition of intelligence to the sort imperative in any design process: predicting human response. We are provoked to feelings by objects, spaces, enclosures, vistas, pathways, instinctively. Predicting these reactions requires empathy, the ability to correctly guess the freely-associative phenomenological response of others. I contend that no system has yet attained a wandering mind for itself, never mind empathised someone else's, and is therefore yet to attain intelligence in the context of design.
However, it may be that I am falling the same old trap of anthropomorphic values. AI is, or will be, dimensionally different from human thought. It may never understand us, or us it. Perhaps the empathetic response I describe above will be reversed: AI might never have a emotional, conscious mind of its own, but will simulate others perfectly.
AI and Architecture - The Story so Far
Discussion of what passes as AI now in architecture tends to be framed in production methods: an aid to design and construct, augmenting existing processes. Testing and analysis of proposals can be carried out in great detail, essentially no more than extreme data crunching. When today's AI is allowed to design, the outputs tend toward extremely complex forms, expressing the computing power of the process, rather than a resolution of a brief.
In some instances, it is almost as if the design process has been aborted at an early stage, with experimental sketches frozen and declared proposals. Some proposals are baroque, a development since the "surface" architecture of the late 1990s. It is hard to discern fundamental architectural devices such as progression, or hierarchy, which have existed for millenia, and are shared by classicists and modernists alike. Ironically, none of them display intelligence. It is as if mid-twentieth century design, which was focussed on spatial clarity, on using less to describe, enclose and release so much, exemplified by Phillip Johnson,
Pierre Koenig, Alvar Aalto et al, never happened - that the most intelligent architectural decisions ever made, in abstracting, framing and holding the beauty of humanity and nature in muted clarity were some sort of dead end. There’s no accounting for taste, and perhaps no computing for it either.
In any case, the connectivity and hyper-vigilance of AI, its global reach and lack of boundaries, makes discussion of it in discrete terms irrelevant. It must be considered as simultaneously parts of a system and the overarching system itself, serving innumerable human clients. Here our client is nothing less than the entirety of human settlement, with its myriad of social variance. For simplicity, I write here in the limited context of a western, capitalist, consumer society, in a generally non-critical way. There are of course many other cultures to be considered, and I invite any comment that broadens the discussion.
AI - Leave it with me.
In the near future, we, in privileged parts of the world, will enjoy homes that know us, where we are, how we are, at all times. Spaces will be cleaned, refrigerators filled, clothes washed, by an AI system that will teach itself through Machine Learning, continually improving itself, autonomously.
Our AI will be a constant companion in all devices, personal to us. Say good-bye to Siri or Alexa, and welcome your new best friend, with a personality of your choice. Mine will be called Aunty: representing someone who has known me all my life and trusted implicitly, and won’t surprise me. Who will yours be? Nigella lawson? Obama? Trump? Be careful what you wish for.
Consumerism will remain ingrained, and will become assisted: As AI proves itself to be transparent and free from human capriciousness, people will willingly give up choice, even of things that they now use to define themselves.
An example: AI will choose our holidays. We will be collected by an autonomous vehicle, delivered to a plane, which will take us to a perfect destination. The stress of picking the wrong place will be a thing of the past; risks of flight delays, shoddy hotels, mosquitoes, and crime, will be eliminated, as will airport terminals. AI will figure out the necessary parameters for itself. AI, with its infinite reach, won’t trust a travel review site. It will ask its AI colleague at a potential destination for all the necessary data regarding human satisfaction. Will that system lie? Will we see AI systems suing each other? Possibly.
On the way to our holiday, there will be spaces to wait, but the chaotic halls of airport terminals and customs clearance will be no more. Instead, there will be an expectation of pleasure at every stage, with lounges (for want of a better term) tailored to your tastes. Satisfaction, at all stages of a holiday, will become an expected norm. Consumers will demand it, when it becomes possible and known. Remember what Henry Ford said: “ If I had asked people what they wanted, they would have said: “a faster horse””.
People have yet to experience the total relaxation of entrusting their most precious time, and that of their dearest, to a machine. But once it happens successfully, they will not do it any other way. Eliminating the risk of a bad holiday is.. priceless. The effect this will have on choices of home, school, work and partners will be profound. As Jamie Susskind explains in Future Politics, to not consult AI in some professional situations will be in itself negligent.
AI to save the Planet.
The shift in decision making, from human to AI, must not be frittered on holidays and games, but should of course be deployed against the existential crisis of our era: environmental catastrophe. Those who trust in tech solutions to everything may or may not be correct, but AI can not be ignored. We can't rely on taxes, nudges or legislation to change human behaviour, and the damage we do. They are all too slow. Parliament in the 19th c. had time to anticipate and plan laws, but the pace of societal change now has left government, and the environment, in crisis: Prosecution for emissions, pollutants, bee-killing, loss of diversity, takes years. When laws are good, the justice system that deploys them is slow and unreliable, just as the biosphere we (if not cockroaches) need is running out of time.
But we do have systems powerful enough to make global change, ones that have outpaced, outflanked and out-thought Government: Big tech. Armed with AI, Big-tech must use its immense power to the greatest good. Will AI set parameters to avoid environmental destruction? It will, if self-preservation is allowed it, and if that is contingent upon humanity’s survival. It may not be (Skynet!) - so let’s not leave it to chance.
These firms will soon be the most effective medium by which we limit environmental damage, via protocols such as cybernetics, and code-law: The operational parameters that program code sets within devices, for example the way an online form will not allow you to submit it without filling in certain fields. This sort of restriction will be embedded in physical devices, for example microwaves that will not heat sour milk, or drones that will not fly near airports.
Saving our biosphere will require a vast geographical reach: reforestation: re-wilding: robust protection of rural spaces. Code-law embedded in our vehicles, devices, even clothes and shoes, could stop us entering certain areas, or at least make it very uncomfortable. When Big-tech is empowered to prevent local messes such as the trash heap climbers left on Everest, expect some radical solutions that bypass traditional law-making.
To achieve this, the nature of the products we consume and the markets that transfer them must change - a task so intricate and vast, only a global AI system will be able to have a significant affect. Code law will be ubiquitous: entertainment devices, homes and transport, will not function in illegal ways. Cars will refuse to park on double yellow lines, buildings will not allow energy consumption beyond a certain point, or will automatically charge occupants as mitigation.
Our habits and preferences will be changed profoundly in turn, with a simultaneous broadening of overall variety (of which you will only see small selection, as AI will assigned to you the things it thinks best for you), and a narrowing of individual choice.
Diverted from wasteful, environmentally damaging activities, people will focus disposable income on experience: sports, family, friends, clubs, festivals, shots(!), as the opportunities for expression through status (high carbon and material consumption) are removed.
We see already a trend to consume fewer physical possessions, of individually high value, replaced frequently: an increasing churn of tech. Our dissatisfaction with last year’s model is apparent on eBay, where perfectly functional goods are sold for very little, purely because they are a year or two old. Even the time spent selling a computer, TV or phone is questionable, given the paltry fiscal return. This is because our demand for ever improving performance has outstripped our desire for assets. The consequence of this now is worsening environmental damage. AI taking control of supply and demand may be the only way we can stop the slow strangulation of our world. We should not forget Earth itself will continue with or without us.
Apple’s recent switch to higher priced iPhones, replaced every four years, profitable through apps, is an early version of this: less material consumption, growth emphasis switched to data and entertainment. Whether Apple have done this in response to the market, or caused it themselves, is debatable, but in environmental terms, it appears to be a positive effect, in line with the changes described here.
AI in the Built Environment
How will this transformation of our social and economic world specifically affect architecture? What does any of it matter to architects in terms of place making, of meaning, of the lived experience of buildings?
Example: consider the impact of ownership of vehicles. The removal of parked vehicles from urban environments, in favour of shared autonomous vehicles, (Govt taskforce, the Centre for Connected and Autonomous Vehicles favours shared vehicles, presumably charged at source), will be profound. DVLA (UK) estimates 26M cars are currently occupying 18.7k ha. As models of ownership disappear we will not only find ourselves with more external space to work with, the language of our streets will change.
In the public realm, the most obvious change will be in security. As AI tracks and monitors all-comers, the need for physical barriers will disappear. As mentioned above, facilities such as airports may not even have perimeters. AI will detect risk well in advance, for efficiency and security, and code-law will prevent the misuse of vehicles, drones etc. It will not be possible to launch many forms of terrorist attack. Surely a change for the better?
At a domestic level, as mentioned above, consumption will be increasingly focused. As private property decreases in scale, the pressure on the environmental quality of homes will be greater. Housing (and healthcare) will become sensor dependent – our homes will record habitation patterns, bio-monitoring will track our moods, our sleep, probably sharing this data with other neighbourhood, city and national data analysis systems. Physically, all buildings will routinely measure insolation, energy use, internal and external environmental conditions. This will affect design in at least two ways:
1. Continual, personal briefing will dictate architectural qualities better than any client brief does now, because, as people trust AI, peer influence will lessen. Qualities of light, enclosure, sound will be generated specifically rather than generally. People will be bolder, prompted and backed up by the one thing they really trust: AI. Determined to express their individuality, as choices are increasingly removed from other realms. The architect's job will be both easier and harder, as people will give more reign to departures from traditional form, and expect increasingly bespoke design. “AI told me I like a small dark music room..”
2. Expectation will appear for homes, offices, galleries, to predict desired environmental qualities, day to day. We will return to homes that already have the lighting and music set to what we need, whether it is to hide from the world, or be energised for an evening out. And if that is wrong, we will be displeased.
These precisely tailored qualities will be increasingly important in off-setting dissatisfaction with ever decreasing space in homes and workplaces. This may be the biggest opportunity, and responsibility, for architects. What propositions can we make that will continue to create wonderful places to live and work? How can we respond when traditional ideals of house and home, of private gardens, and leafy suburbs become irrelevant or unattainable? (As they already are for countless generation X and millenials).
The Affect on the Profession of Architecture
Some or all of these changes will happen, and architects have a choice: Either be part of the discourse, directing change, or be reactive. In my view it must be the former, for if we are not, we will be under threat. Lawyers are already: studies are very mixed, but overall, AI is able to predict supreme court judgements correctly, in advance of oral testimony, with about 80% accuracy. Human experts achieve roughly 60%. We will have a legal system running on AI, with a very few cases going to human appeal judges, because it will be better, quicker, and more reliable. Similarly, medical diagnosis is more accurate using AI, and it easy to se why: a tired, worried human doctor can miss things AI won’t.
If there is to be a cull of lawyers and doctors, architects should be worried, we are next on the list. We have a get out of jail card: empathy intuition, taste. As reasonable as architects have to be, these irrational qualities may keep us in work, or at least some of us. Anyone designing car parks probably should be worried. And we must keep our eyes open; ensure we see the changes coming and make our input informed and relevant.
As mentioned above, Government is stuck, and it is hard to see how statutory control of the Built Environment can respond adequately. Perhaps we should not worry too much about our region or nation - the most successful architects of the next few decades will (still) thrive in a global context, embracing the biggest of changes.
The deep, structural changes that AI will bring in the coming decades are being discussed by many, learned writers, and this blog can only touch upon some of them. For properly considered analysis and speculation, a brief reading list should include: Future Politics (Susskind); Radical Markets (Posner & Weyl); Homo Deus (Harari), to all of whom I owe a great debt.
Ivan Jordan RIBA