Towards a green AI; the strategies needed to put a check on the AI industry’s carbon emissions
When the science fiction writers of the mid-20th century wrote of artificial intelligence and the apocalyptic dangers contained therein, few speculated a problem as banal as the one we now face.
This two-part problem, a syllogism, comes from the facts that AI is vastly useful. And that it is carbon-hungry on a scale few seem to grasp.
In short, AI may be the death of us not because of any Terminator-foreseen master plan, but rather because it falls into that category of the deadly and irresistible, shared by such world-ending forces as cigarettes, refined sugars and diesel engines.
But before we move on to the effective strategies for curtailing AI’s carbon output – we must address the argument from luddism.
Why can’t AI simply be legislated into a less harmful form?
This isn’t beyond the realm of possibility.
The EU’s AI Act goes some way to addressing environmental concerns. And elsewhere, other governing bodies are drawing up similar legislation aimed at reducing AI’s potential harms.
However, there are compelling arguments for why this 1) won’t and 2) shouldn’t happen.
One only has to look at the decades-brewing interventions in tobacco, oil, ocean waste, refined sugar, or any number of environmental or public health hot topics to see that such interventions are often delayed until long after a great amount of damage has been done.
Add to this the huge growth of AI and you have the argument for why, on balance, a meaningful legislative intervention in AI won’t happen.
Here’s why it shouldn’t.
The Mayo Clinic recently announced the dawn of a new era in AI-powered cancer research.
This while researchers at Princeton leverage the technology to decode the basic building blocks of the universe.
Around the world, people are using AI to revolutionise, make more efficient, innovate and create.
And, in ironic touch, AI may just be the answer to the climate crisis we’re so worried it will exacerbate. In short, AI has the potential to be the answer to life, the universe and everything. As well as itself.
So, if we can’t and shouldn’t put the cat in the bag – how do we keep it from scratching the furniture?
Towards a greener AI
The below are three key ways we can transform AI’s carbon output.
- Start thinking realistically about the virtual causes of AI’s carbon emissions
At Turn it Off, this is what we do.
As cloud increasingly becomes the cornerstone of AI (and why wouldn’t it?) the carbon-generating power of cloud computing becomes ever more important.
The first place one might think to look are the vastly wasteful data centres still used by many in the public cloud space. But we’ll get to those later.
First, there is the often-heard call for more energy-efficient forms of virtualisation: microservices, serverless, etc.
In most cases, these efficiencies are passed on at the hardware level, resulting in less carbon output.
However, we also need to get creative.
Our own intelligent SaaS solution is focused solely on turning off cloud environments and resources when they’re not in use. A specialisation not replicated in any other tool.
Turn it Off allows engineers and business users to create schedules, rules and leverage AI to turn off AWS and Azure cloud environments when they’re not in use and back on when they are. We also offer carbon and cost savings reporting so you can see the real-world impact of our platform.
While many place emphasis on overall cloud management and efficiency, we’re focused on reducing waste with immediate impact. It’s in the name.
By building our tool explicitly as a waste reducer (both at the brand and function level) we aim to reduce carbon in real terms and promote awareness of the often-overlooked impact of data centres and cloud computing on the environment. It’s free to start, and based on % of savings thereafter.
2. Stay informed and resist greenwashing
In the public cloud space, many have made lofty commitments to carbon neutrality.
In some cases, this is fuelled by genuine technological innovation or direct action.
Next-generation cooling technologies like liquid immersion cooling, evaporative cooling systems and geothermal cooling; the integration of renewable energy sources either through direct purchase or onsite renewables; custom silicon like that employed by Amazon’s Graviton processors and the integration of 80 Plus ratings materials like gold and titanium.
All these are good things. But we shouldn’t be lulled into a false sense of security.
For one example, many companies (including public cloud providers) are achieving their neutrality ratings via carbon credits.
These are purchasable tokens that pledge a certain carbon offset to factor into one’s own output (and CSR reporting). However, there are serious questions as to these credits’ authenticity.
Alongside real intervention, a greenwashing cottage industry is booming. It’s up to us all to sort the fact from fiction when it comes to investments, partnerships and purchases.
3. Support AI as a planet-saving technology
And now we come full circle.
The fact is, AI as it currently stands is a major threat to the environment. But it’s also one of our best chances for a paradigm shift.
Our own product is one small example of AI leading change in reducing carbon. But there are many more.
Artificial intelligence has already made huge advances in the way we organise networks – energy grids, supply chains, water networks and many more have seen vast increases in efficiency under AI management.
It’s used to make vehicles less wasteful and even change the habits of people for the greener.
Last but not least, artificial intelligence is now at the forefront of the energy sources that may break our reliance on fossil fuels for good.
Despite the danger it poses, AI could still be the greatest technological revolution we’ve yet seen, and the greatest tool we have at our disposal. As Arnie would say, ‘come with me if you want to live.’