AI uses electricity and water, but what would we use to do the same jobs without it? After nearly 50 years of changing technology in estate agency, I think the environmental comparison may be more complicated than it first appears.
AI: To be or not to be, that is the question.
I keep reading about the environmental cost of artificial intelligence. AI uses electricity. Data centres can use significant amounts of water for cooling. These are perfectly reasonable things to question.
But there is another question that I don't see asked quite so often.
If I didn't use AI to do a particular job, what would I use instead?
I don't know the answer to the environmental argument. I'm certainly not qualified to calculate the relative amounts of energy, water and other resources involved. But after nearly 50 years in estate agency, I have seen quite a few changes in the way we work, and I do know that the alternatives weren't environmentally free.
Everything wasn't better before
When I started work, we used manual typewriters. They didn't use electricity. Then we moved to electric typewriters. Suddenly we were using electricity simply to type a letter. But they made the job easier and more efficient.
Eventually computers arrived, followed by email, the internet, smartphones, cloud storage and all the other technology that we now take for granted.
Property details used to be printed. At one stage I can remember us spending a considerable amount of money every month just on toner for colour laser printers.
But toner was only part of it. There was the paper, the printer, the envelope and the stamp. Somebody had to put the property details into the envelope and take them to the post box. They then had to be collected, taken to a sorting office, transported to another sorting office and eventually delivered to the potential buyer.
Today somebody can look at the same property online from almost anywhere in the world. That isn't environmentally free. Servers and telecommunications networks need electricity and physical infrastructure. But neither was everything they replaced.
Even accessing information used to involve travelling
I can remember having to go into the office on a Saturday morning simply because that was where the telephone and the property database were. If somebody called, somebody needed to be there to answer the phone. If I wanted information from the database, I needed to be sitting in front of the computer that contained it.
Today I can access the information I need from almost anywhere. That means data centres, networks and devices are working somewhere in the background. But it can also mean that I haven't driven to the office simply to access information.
The same applies on a much larger scale to business meetings. A video conference uses electricity and data infrastructure. But if the alternative is somebody flying to another country for a two-hour meeting, surely the environmental comparison needs to include the flight, the journeys to and from the airports and possibly a hotel.
Photography is another example
We used to photograph properties on film. You took the photographs and sent the film away to be developed. Proofs came back. Occasionally something had gone wrong and the film came back blank, in which case you went back to the property and started again.
Film had to be manufactured. Developing involved chemicals. Film and photographs had to be transported backwards and forwards. Eventually the photographs were reproduced in printed property details.
Digital photography changed all of that.
Now AI is changing another small part of the process for me. If a property photograph is too dark, I can put it into an AI tool and say:
“This photograph is too dark. Please make it lighter without changing the details or the subject.”
Previously I might have opened Lightroom or Photoshop and worked out which controls and sliders I needed to adjust. Both methods use a computer and electricity. AI also involves processing in a data centre. Which method has the smaller environmental footprint?
I honestly don't know. And that is rather my point.
What would this room look like in sage green?
AI can also do something that previously would have required considerably more effort.
I can photograph a room and ask:
“What would this look like if I painted the walls sage green?”
Within moments I can see it.
Strangely enough, when I did the photograph for the head of this blog article, I looked at the room with the sage green walls and thought, “Hmm, I think I'd rather have that with a colour-blocked sage green ceiling as well.”
So I asked AI to show me. I could probably have produced the same three images using Photoshop, but with my level of Photoshop expertise I could easily have spent half a day doing it, and I'm fairly sure there would still have been things wrong with them.
Does a few moments of AI processing use more or less energy than me running a computer and Photoshop for half a day to achieve the same result? I honestly don't know. But surely that is the comparison we need to make.
That doesn't mean I should advertise a property using altered photographs as though they were real. We don't. But as a practical tool for answering a question or considering an idea, it can be extremely useful.
It might even prevent somebody painting a whole room sage green before discovering that they hate it.
I use AI because it is useful
I use AI for research. I use it to check spelling and grammar because my typing isn't particularly good. I use it for dictation because it works remarkably well. I use it to help with repetitive tasks that would otherwise take me a considerable amount of time.
We also use it for practical jobs such as improving the legibility of an old floor plan or making sensible corrections to photographs.
I don't use AI because I think everything should involve AI. If another tool did a particular job better, I would use the other tool. For me, it is simply a tool.
Compared with what?
Nobody is arguing that AI is environmentally friendly. Data centres consume electricity. Some cooling systems consume water. There are legitimate questions about where those resources come from, how efficiently they are used and how the environmental impact can be reduced.
We should be asking those questions. But I think we also need to be making comparisons with the alternatives.
If AI performs a task that would otherwise have required another computer program, several hours of somebody's time, printed material, physical storage, postage, a car journey or even a flight, those things belong on the other side of the calculation. Sometimes the alternative may turn out to be environmentally better. Sometimes AI may be better. Sometimes there may be so many variables that nobody can give a sensible general answer.
Different solutions suit different people and different circumstances. There isn't necessarily one right answer for everybody.
But comparing the environmental cost of AI with an imaginary alternative that consumes no resources at all doesn't tell us very much. It's good we're talking about it because it puts pressure on the industry to make AI and data centres more environmentally friendly.
Everything we do has a cost somewhere. The interesting question will always be whether there is a better way of doing it.
