AI, our planet and that rare thing: a balanced view

Generative AI is like many useful innovations we quickly normalised – like using a Google search to find a website when you know the URL, or booking cheap flights. Just because we can’t immediately ‘see’ the environmental impact, doesn’t mean it’s treading lightly on our planet, and it’s very easy – too easy – not to think about it. Time to make your life coach proud and lean in to the difficult conversation.

With great power comes great responsibility

Life and supply chains are now impossibly, globally, complicated and largely invisible from the punter's end of the pipe. It has never been harder to make ethical choices because we can’t reasonably keep track of every unintended consequence.

You know you need to shop local, buy fair trade and choose organic, but you also need to do your Big Shop online in that gap between Teams meetings. Can you really check the origin story of everything you buy, get your job done and live within your means? The ability to visit that new eco-friendly zero waste shop and then the local organic butcher, baker and candlestick maker is, let’s be honest, only an option for the privileged few. Don’t beat yourself up; we’re in this cost-of-living crisis together.

Articles and podcasts that are either relentlessly gung-ho or hopelessly doom-laden about AI adoption offer little in the way of balance and even less practical advice. So what’s really happening, and what can we do?

Well, this is awkward

Training a single large AI model, like the ones our engines rely on, can produce more than 300 tons of carbon dioxide emissions, equivalent to about 300 round-trip flights between New York and San Francisco. A ChatGPT prompt and response uses nearly 10 times as much electricity as a single Google search. Don’t look away: that’s the reality.

Calling it ‘the cloud’ was a master stroke of PR. What we’re really talking about is our rapidly expanding use of not-at-all nebulous data centres to power the AI boom. They consume a lot of electricity and rely on a continuous supply of fresh water to keep them cool. But that’s not the whole story: AI also has the potential to help mitigate climate change and offset other resource-intensive activity, if used responsibly. So, how can we find the right balance?

It’s not me, it’s you… and me

Before we get into solutions, let's take a step back and have a word with ourselves about our own daily behaviours - all of which pre-date the arrival of ChatGPT in 2022. How often do we leave our computers on overnight? Were we ever giving a lot of thought to the energy consumed by our endless Google searches? Probably not, until we started scrutinising AI because the media pastime of demonising it is both easy and fun. But we can’t ignore our own contribution to energy waste and if AI power consumption is our wake up call then there's good news: if we’re part of the problem, we can be part of the solution.

What can individuals do?

1. Use with caution: Be mindful of when you truly need AI tools and right-size them to the job. Do you really need AI-powered assistants like Copilot to set reminders when you can create them in your calendar?
   
2. Think before you ask: Refine the prompts and queries you enter into AI tools before you enter them to ensure they're as fit-for-purpose as possible the first time, rather than asking essentially the same question a dozen or more times because you got the phrasing slightly wrong or missed out a key bit of context that meant the AI didn't give you quite the response you were looking for.
3. Choose better AI services: Opt for AI tools from companies using data centres running on renewable energy – or at least have an active plan to get there. A little research can go a long way.
4. Adjust your settings: Use lower-resolution or less computationally intensive settings for AI applications. Do you need a high-res AI-generated image for a quick social media post? Probably not. Go low res. Let Excel sum up that column on the spreadsheet.
5. Extend your device’s life: Maximise the lifespan of your existing devices instead of upgrading. This not only reduces electronic waste but also conserves the resources needed to produce new gadgets. Turn it off when you should, do updates when you’re told to.
6. Put your waste in the right bin: Recycle your AI-capable devices. You don’t need all those old phones and laptops: take your local council’s advice on how and where to recycle them or pack them off to a re-seller. Resource cost per device can come right down.
7. Support eco-friendly initiatives: Choose products and services from companies that are honest about their efforts to reduce their environmental footprint. It’s as easy as asking them to tell you about it. You won’t need to be an AI expert to know if they’re bullsh***ing you; you know nonsense when you hear it.
8. Use AI for good: Use tools and applications that help monitor and reduce your personal environmental impact, like smart meters.
9. Influence, advocate, be an activist: Fear not, we’re not telling you to up your Insta game: Support policies and companies that prioritise the development of environmentally responsible AI technologies. Be that person in your own organisation.

“No man is an island.” What can organisations do?

1. Choose sustainable cloud providers: Major players like Google, Microsoft Azure, and Amazon Web Services have all pledged to use 100% renewable energy. Look under the bonnet and check what deadline they set themselves and whether they’re on track. If you are shopping for alternatives, see how they compare. The potential benefits probably warrant a perplexity.ai query.
2. Choose your data centre locations with care: The carbon intensity of electricity grids varies. Choose data centres in regions with cleaner energy mixes to substantially reduce emissions. Consider locating data centres in regions with abundant renewable energy sources, like hydroelectric power in Scandinavia.
3. Make the most of your resources: Use cloud management tools to adjust the size of your services to fit your needs, shut down any unused digital services, organise tasks efficiently, and train your teams to write smart, energy-saving prompts and code.
4. Stay cool: Data centres in cooler climates generally require less water for cooling. Some providers use innovative technologies like liquid immersion; others have replaced freshwater with recycled. Look out for examples like Google's use of seawater cooling in Finland.
5. Adopt sustainable IT practices: You don’t have to leave it to busy people in your hard-pressed teams: extend hardware lifecycles and recycle e-waste properly, at an organisational level. Choose energy-efficient equipment.
6. Monitor and measure environmental impact: Use tools provided by cloud vendors or third parties to track energy usage and emissions. Set internal sustainability goals for cloud and AI operations. Tell everyone what they are, then tell, honestly, them how it's going.
7. Support renewable energy schemes: Some cloud providers allow customers to purchase renewable energy credits or participate in sustainability programmes. Don’t ignore them or tell yourself they’re just for the knit-your-own-knickers crowd. They’re designed to help everyone. Literally.

Leading by example

It’s always tempting to be snarky, but Google, Microsoft, and Amazon Web Services have all made real efforts to reduce their environmental impact. Google uses machine learning to optimise cooling systems, reducing energy consumption by 40%. Microsoft has pledged to become carbon neutral by 2030 and has developed AI solutions to help businesses optimise for sustainability. AWS aims to power its operations with 100% renewable energy by 2025 and become water positive by 2030. And that’s in the full knowledge of how our demand for data centres is expanding.

Engine occupies a very different place in the tech market: entirely home-based, small, building simple, affordable tools mainly for SMEs and non-profit organisations. But we are facing the same challenges, and scale shouldn't let us off the hook. We are continually looking for ways to minimise resource consumption in our supply chain and take our own advice. Crucially, we make sure everything we build is useful. Enquire within if you want to know more…

Finding the balance

Every rule has justifiable exceptions: learning to use AI will involve using it to make some things that aren’t mission critical. It’s like using a flight simulator before you pilot a jumbo full of passengers. Safe places to practice are a good thing if you have your eye on the longer term outcomes. Your choice of data centre also has to meet your security standards; for us, they have to be UK-based for our UK customers’ data security, and that limits our options. Recognise and realise the benefits of those choices too. Off-set them against the resource that has to be invested to realise them.

We aren't going to insult your intelligence - or shoot our business model in the foot - by telling you to avoid AI, but there’s lots you can do. Start today. You probably have a net zero strategy but, unless you are an AI company, it probably needs an AI-themed re-write. Decide your AI policy, map out your journey, manage the risks but recognise and exploit the opportunities too – responsibly. AI has the potential to contribute positively to environmental protection and society at large. But it will, as ever, come down to how we take responsibility and hold ourselves to account, so be mindful of your behaviours and make informed choices.

‍

‍Engine builds private AI tools that let you keep things simple and honest; private GPTs trained on content you curate and your brand guidelines, giving you answers with citations that your teams can verify. We charge a one-off, affordable build cost and a single, fixed monthly cost for your whole organisation. Email us at info@engine-ai.co.uk.