Why AI is the last thing climate needs
How do you feel about artificial intelligence or ‘AI’ as it’s commonly known?
Whether you love it or loathe it, it seems to be creeping steadily into our everyday lives, almost unnoticed.
There is no doubt that it can make things easier, such as “helping” with advanced internet searches or saving time by writing simple documents - I’m sure it could have written something a bit like this column in seconds, had I asked it to (which I didn’t).
But there are much bigger issues in play: such as the coming weaponisation of (AI-fuelled) disinformation, the erosion of creativity, the risk of inaccuracies (where AI ‘hallucinates’), an invasion of privacy, a huge coming growth of AI fake friends (including girlfriends) and (most worryingly of all) AI being misaligned from human values.
For people with environmental concerns, which, of course, should be all of us, there are other things to think about, such as the massive amount of energy, water and resources it uses.
A request made through ChatGPT, a well-known AI-based virtual assistant, uses about 10 times more energy than a regular Google search, for example.
While global data is sparse, there are estimates that in the tech hub of Ireland, the rise of AI could see data centres account for more than 35pc of the country’s energy use by next year.
For me, AI has almost become the new flying.
‘AI is the last thing climate needs’
I think of conversations I’ve had with people for so many years now...
Instead of wanting to know whether (or how often) we should be flying, especially to destinations accessible by train, they now ask whether (or how much) we should be using AI because of the cost to the planet.
My response?
As I explained recently in a packed debate held at The Forum in Norwich, ‘Steering Humanity through AI & Climate Risk’, AI is the last thing climate needs.
It is very hard to stop using AI, though, while everyone else is using it; you risk suffering from a ‘competitive disadvantage’ if you do.
That is why we ought to support campaigns to enable collective society/worldwide pauses of or even turnings-off of AI…
For: While some who support AI naively believe it will be the great saviour of our time, helping, for example, to monitor deforestation or to be able to predict natural disasters like the increasing number of (hard to predict) wildfires, it really won’t save us.
No; that’s up to us.
And, so far, we’re not doing a very good job of even attempting to make a go of it.
In fact, in the last year, we seem to be going backwards with many governments, not limited to the US and our own, back-pedalling on policies aimed to, at the very least, slow climate breakdown.
While AI is being allowed to let rip.
So, where do we go from here?
First, we need to face up to where we are, and then we need to start adapting to what is here and the far worse material conditions that are to come, together.
We need to create a relocalised future, a #SAFER future, for everyone.
We need to build our collective resilience, including crucially community resilience.
For the tragic reality is that we are going to face more and more climate disasters, some in effect fuelled by AI, and we may start to face more direct AI-induced disasters too.
If, for example, AI is used by foreign powers to sabotage our critical systems here.
The good thing about the building of community resilience is that it is multi-purpose.
It builds our preparedness for climate-, AI- or war-induced vulnerability, all at the same time.
Readers of my columns will be aware already of my calls for the building of climate resilience in our part of the world.
But what I am adding today is the following less-obvious truth: that if you help build community resilience in any way at all - e.g.
through running a local soccer team, or supporting your local WI - then, just by virtue of doing so, you are building resilience against climate- or AI- disasters that will be coming.
And that is an encouraging thought, in the darkness of this time.
For we are going to be needing each other.
And what we really want is, exactly: each other.
If you want a picture of the future, imagine being in a group of friends and neighbours, sitting around fires (maybe singing, or drinking, certainly chatting) — forever.
Whether or not you are doing so because the electrical power is out…
Thanks for the assistance on this column to my wonderful human researcher, Tara Greaves.




I'm a bit late to this -but this discussion is urgently needed. i feel the biggest risk is AI's inherent propensity towards planetary scale and the rapid growth of its infrastructure; we are essentially creating an entirely new massively heavy industry - and there willl never be enough energy and resources to meet its demand.
openAI is planning on 250GW of data centres - which is about the same as India.
In Sth Korea they are building an AI "mega-cluster" that will eventually require 16GW of electricity - a 6th of South Korea's entire grid, its water demand will be half that of Seoul.
As always I appreciate your writing, and relate strongly to all you say. But I feel compelled to mention one researcher (Andy Masley) who recently did a systematic deep dive into the hard data on AI and energy/water use, and concludes that the alarms over environmental impact from AI are at best overblown and at worst, an epic failure of science communication. For instance, he roundly debunks as myth the common claim cited at the beginning of your post, that ChatGPT uses 10x as much energy as Google Search.
To be clear, I'm not convinced he's correct (and intuitively, want to be skeptical). But it's not obvious he might *not* be correct either: his research appears exhaustive, he seems to take most objections into account, and I honestly don't know what to think anymore; this whole topic has become very confusing. Meanwhile, he's gotten lots of uptake on Substack and is definitely influencing how people think about this; Hannah Ritchie has chimed in to similar effect citing his research, for example. So at least for the environmental piece, it could be important to engage with his research at some point. Really curious what your own take might be. Links below if you're curious.
And it goes without saying, none of this speaks to the many *other* reasons to be alarmed about AI - but it would certainly be helpful to have clarity and consensus on the real environmental impacts, given these conflicting narratives.
https://andymasley.substack.com/p/ai-and-the-environment
https://andymasley.substack.com/p/empire-of-ai-is-wildly-misleading
https://open.substack.com/pub/andymasley/p/individual-ai-use-is-not-bad-for