The True Cost of AI
The thought goes – we’re destroying the planet, the true cost of AI is potentially frightening… and indeed, we are. I read a recent article about Arctic adventurers quitting their dream jobs in the Polar Regions because Tourism is so rampant in these sublime places. Boatloads of people are literally cheering when they see the Ice Shelf ‘calving’. They seem not to understand that every bit of ice that melts could be a disaster for mankind. A sort of sleepwalking to our own doom.
AI is the same, we all use it, no one wanted it, or asked for it, but it’s a part of everyday life, and it’s fostered onto us through our banks, governments, institutions and so on. Sure, we’ve all had a go at ChatGPT. You either love it or hate it, embrace it or get left behind. It makes no difference the bulk of the world now uses it.
The great debate in the media is whether we’re heading for our doom, and not just because computers will ‘take-over-the-world’, but more simply because of the incredible energy and resource requirements of our silicon chip friends.
So, how do you work out the True Cost of AI?
Financial & Environmental
It’s almost time to invent a new form of currency; no one else seems to have tried yet. But what if everything we did was measured in terms of Environmental Impact? We could clearly split the costs into two clear segments. Financial and Environmental.
Take a tomato, you buy 6 on-the-vine tomatoes for £2.49 from your local supermarket in June, and we’re told the weight. We’re often also told the source, i.e. where it is grown, the Isle of Wight, for example. The farm is dedicated to growing Tomatoes in greenhouses. In the summer growing season, we could suggest the Environmental Impact is say 2% or some other measure.
Now consider, same tomatoes, same price probably, same supermarket, but it’s December and the tomatoes has come from sunny Spain, some farmer has built greenhouses and can produce tomatoes all year round, but this time the environmental impact going be 8%.
The concept can be rolled out to any product, any service, anything that has a financial cost. We buy electricity from a supplier, who sells it at 25p per kw, but it’s known that the supplier uses oil to produce their electricity, 30% environmental impact. Another supplier sells at 30p per kw, but mostly sources from renewables; they have a 10% environmental impact rating. And so on.
Environmental Impact Scores
I guess someone has tried to do this already, we have carbon credits and the like, but it’s a bit of a lame effort, and more appropriately, it’s very very controlled by business. In capitalism, there is no case for anything other than financial reward, but if people could work out their Environmental Impact, would they be swayed?
The data currently suggests not by much, Tourism in the Antartic is set to increase from 120,000 people per year to over 400,000 within the next 2 years. The use of AI is exploding in a direction that is unprecendented and I don’t think anyone is about to give up their cars.
Humans have an odd herd mentality that seems to move in the opposite direction to common sense. Who switches off the lights any more?
But this only reinforces the need for a national requirement for an Environmental Impact score for everything we do.
The AI Environmental Cost
Assuming that we could work out a score, what would your website cost? One assumes it would be high. Data suggests that a single promotion for coding a website is equivalent to 130 prompts for a regular ChatGPT request. Let’s put that another way, if you are looking for information about Waterloo or how long it takes a train to stop, sure the 300-500 words ChatGPT generates in response to your question will cost something, but if you want to build a new page on your website that is SEO friendly and lists a new product or service that people can enquire about – that’s expensive.
The cost is huge… They say, true or not, that every word on AI costs about 10 litres of water; the Datacentres being built now will consume more electricity than we produce at this point in time. The USA uses more electricity in their current datacentres than the UK uses for 60million people. The numbers are huge.
Efficiency is key
The light at the end of the tunnel is efficiency. The core issue at the heart of AI is how much more efficient it can be. If it produces bloat-free, well-written code in 20 lines, rather than 40, that’s a 50% saving straight off. It’s not only a 50% saving in the coding time, but it’s a 50% saving for as long as that code is used forever.
Computers work in 1’s and 0’s; it’s as simple as that. Every character within some code has to be processed in 1’s and 0’s every time it is processed. If we can reduce the code by half, we reduce the processing by half.
The tech world has assumed for years that they can just add more code to any App or Programme, from Microsoft right through the DIY’er, but that is not true, and right now Microsoft is reducing its code base, removing obsolete features, functions and unused code from our computers. Sure, they replace it with new stuff. But you can see where I am coming from.
Our greatest hope, therefore, for the future is the reduction is processing needed for the tasks we want to achieve.
The True Cost of AI
To work this out, we have to be practical. The true cost of using AI to build websites, as we now recommend, has to be considered from a financial aspect and an Environmental one.
The Financial side is easy, pick a plan, £20 a month, £1500 to build a website, you name it, it’s easy to work out.
But environmentally, you need to hold your ‘developer’ to account, ensure the tools they use are the most efficient, responsible and have the least environmental impact.
To get this going, you have to ask. And you should want answers.
At NCompass here, just like when we offer various options from a financial perspective, we can do the same from an environmental perspective as well. Try us!