As soon as we spotted it on Rightmove last year, my home-to-be acquired a nickname: the funny blue house.
asserted
home → spot → nickname
The shade — which seemed to defy description, but let’s call it “eye-catching” — was a quirk, we thought; a terrible decision made yonks ago that, while none too serious to us, was a source of mirth for friends and neighbours.
asserted
that → seem → friends
So when a relative staying with us over the summer offered to repaint it (horrified, clearly, by the indignity of being confronted with it daily), we were suddenly landed with a conundrum: which colour to choose
I had gone from AI refusenik to reluctant adopter
Green was the shade that first sprang to mind.
asserted
that → stay → mind
But after wandering through my east London neighbourhood looking for olive-based inspiration — and finding none — I sought out a place where my idea would be championed: ChatGPT.
asserted
idea → wander → place
I had gone from AI refusenik to reluctant adopter a few months earlier, when we were renovating our new house.
asserted
we → go → house
Unable to afford an interior designer and in the absence of any meaningful alternatives when needing to make a multitude of decisions each day, I turned to the chatbots.
asserted
I → afford → chatbots
Probably unsurprisingly, they did nothing for design inspiration or product sourcing — when I asked for suitable furnishings to match our walls and carpets I was served a stream of hideous products and the words “Scandi cool” and “mid-century chic” on loop.
asserted
I → do → loop
Trying to find ideas for a headboard
asserted
Trying → try → headboard
produced the same items that I had found when googling; when I toyed with suggestions for lampshades, I was given an infinite list of options that failed to narrow things down at all.
asserted
that → produce → things
No matter the prompts I gave on colour, texture and price, no chatbot-generated results were of any use.
asserted
results → give → use
B&Q vs ChatGPT
The visualisation apps, like IdealHouse and Homevisualizer AI, fared somewhat better.
asserted
apps → fare → IdealHouse
Granted, very few actually worked.
asserted
few → grant → ?
But on the occasions they did manage to generate an image, being able to drag and drop the bedside tables I’d been eyeing up into a picture of my new loft bedroom was quite useful for thinking about which colours would work— although I remain without bedside tables, so make of that what you will.
asserted
you → manage → that
After poring over numerous colour charts I started running them through ChatGPT (which rarely managed to render an image)
asserted
which → pore → image
Some stores, such as Ikea and DFS, offer visualisation tools where you can drag and drop furniture into a room; John Lewis provides renderings if you book an appointment with its Home Design team.
asserted
you → offer → team
After poring over numerous colour charts I started running them through ChatGPT (which rarely managed to render an image) and Google Gemini, watching the exterior of our house reimagined in Farrow & Ball’s Bancha, Olive and Ball Green.
asserted
exterior → pore → Bancha
The first — a mid-century inspired cross between a pea and an olive — was an easy winner.
asserted
first → inspire → pea
Yet there was one other problem: when I asked to see the house in that same shade on different occasions, each image was completely different from the last.
asserted
image → be → last
And so back analogue I went, traipsing off to B&Q for an armful of testers.
asserted
I → go → testers
They looked — as you might suspect — nothing like any of the AI images I’d been given, prompting a crisis of confidence.
uncertain
I → look → confidence
With days to go before an event at home and the aforementioned relative returning to the other side of the world, all we could do was pick our green — a far bolder shade than we’d initially imagined — and cross our fingers.
uncertain
we → go → fingers
Free advice, costly mistake
“AI is brilliant at giving you 20 different ideas.
asserted
AI → give → ideas
It is less good at knowing which idea is the right one,” says Catherine Seagrave, design director and co-founder of online interiors agency the Living House.
asserted
Seagrave → know → House
“It can help you get unstuck, find the words to describe your style and explore different directions.
asserted
you → help → directions
But it can also present a bad idea very confidently.”
asserted
it → present → idea
Given our house was already a hideous shade and the offer of free labour was time-sensitive, we were perhaps more willing than most to bite the bullet.
asserted
we → give → bullet
We are reasonably happy with the end result, although it certainly looks nothing like any of the renders.
asserted
it → look → renders
“But free advice can still lead to an expensive mistake,” says Seagrove.
asserted
Seagrove → lead → mistake
I would happily use AI for inspiration, but I wouldn’t let it make an expensive or permanent decision.
asserted
it → use → decision
Given the cost of hiring interior designers (as much as £450 per hour, according to London-based KDS Interiors) and the general lack of showrooms in which to see design ideas made flesh, it’s unsurprising that so many of us are turning to tech for help.
uncertain
many → give → help
Research published last month by furniture retailers Barker and Stonehouse found that some 61 per cent of UK adults have used AI for interior design inspiration, rising to 88 per cent among 25- to 34-year-olds.
asserted
cent → publish → olds
“It may replace a few hours of scrolling through Pinterest”
uncertain
It → replace → Pinterest
Given how fallible this technology is, however, mistakes are inevitable.
asserted
mistakes → give → ?
“One of our clients asked AI for a bathroom paint colour,” Seagrove says.
asserted
Seagrove → ask → colour
“It missed the lilac undertone in her tiles and suggested an orange peach that completely clashed.”
uncertain
that → miss → peach
It also fails to understand the liveability of a place.
asserted
It → fail → place
AI “can make a room look beautiful on a screen.
asserted
room → make → screen
However an interior designer has to actually make it work in real life — on a wet Wednesday morning, with children, pets, schoolbags...
asserted
it → have → children
AI can recommend a sofa but it doesn’t know whether it will fit through the front door, overwhelm the room or survive family life.”
asserted
it → recommend → life
The frequent mishaps caused by using this technology make it hard to imagine that chatbots will render IRL experts obsolete anytime soon.
asserted
experts → cause → technology
…and 2 more, not listed.
The Aporia
