Are you thinking about sustainability in AI?
Not a day goes by without a company, charity, government or some other gas-guzzling, forest-burning, ocean-destroying organisation making a grand claim about net-zero, carbon neutral or otherwise pledging to single-handedly save the planet by twenty-whenever. We’ll get the brag out of the way quickly – ContactEngine is already carbon negative – but, following the noise around COP26, the threat of irreversible climate change provides a good reason to discuss sustainability in AI more generally.
Cloud-based computing has rightly come under increased scrutiny in recent years for its energy use. Greenpeace estimates that by 2025 the tech sector could consume 20% of the world’s electricity, a huge rise from its current 7%, and one that will be largely driven by cloud computing. As it stands, a lot of this energy use doesn’t come from renewable sources, with nearly 4% of all CO2 emissions coming from data transfer and infrastructure – a figure 60% higher than aviation. Computing manufacturing also has terrible effects on the planet and the communities where rare metals are found (Google ‘lithium mine Chile’ to see just one example). Then there’s the ‘one Google search is enough to power a lightbulb for an hour’ claim, made by Sunday Times columnist Rod Liddle and echoed across various documentaries. While an overstatement, it does put cloud usage into a tangible environmental context. It might feel invisible but currently this innovation has very real effects.
But of course, the reality is much more nuanced. Cloud computing enables a broad range of efficiencies in both the public and private sector, such as enabling more people to work from home, so therefore travel less. Consolidated data centres can increase physical storage capacity exponentially while their energy increases very little, thanks to advanced cooling systems and the ability to reuse heat. Then of course there’s the fact that several of the largest cloud computing providers have received praise for their sustainability practices, like Google, which has been carbon neutral for 14 years.
AI has so much potential to build on this. To truly be a force for the greater good, AI investments must place sustainability at their core, equally important to the grand debates that have shaped the sector already in recent years, such as transparency, responsibility, and privacy.
Incorporating AI into sustainability
First, we need to be honest about the energy that training a single, deep learning, natural language processing model costs. The famous study by Strubell et al points to this costing approximately 600,000lbs of CO2 emissions – roughly the same as five cars’ total usage. But then it depends on what you train that programme to do. ContactEngine saved more than 23.5m tonnes of CO2 for our clients in 2021, largely by improving efficiencies in deliveries and home appointments and therefore reducing truck rolls. Apologies – that’s the second and last brag in this blog.
But it’s in finding these efficiencies that AI can save the world. I’ve discussed before about how AI can learn a person’s habits and manage power generation in their homes accordingly. Every AI decision needs to be taken with this in mind. What’s the cost? What’s the benefit? Are there any unintended consequences to be aware of? Should we do this?
PwC offers one of the more optimistic views of the potential of AI. Using AI for environmental applications has the potential to boost global GDP by 3.1-4.4% while reducing greenhouse gas emissions by 1.5-4%, they argue. Not a bad trade off. Across agriculture, water, energy and transport, AI for environmental solutions could contribute $5.2tn USD to the global economy by 2030.
We know intrinsically that AI and efficiencies are intrinsically linked and coupled with the move towards more sustainable energy sources, it’s clear that AI can be part of a solution for a much healthier planet. But getting there is not without its challenges. PwC’s models show up to 38.2 million jobs to be created worldwide thanks to AI, but that means a huge re-skilling must take place. By and large, there’s still insufficient focus on delivering responsible AI and the technology still faces distrust by the public.
There are numerous challenges that must be overcome if AI is to realise its potential. Firms that push its boundaries need to lead from the front. We need to focus more on responsible developments, we need to shout louder about our environmental credentials, and we must build trust with the public through transparency. We must publish our calculations when it comes to the environment, and we must ensure that our staff live these values through their innovation. AI can save the world, but only if we keep focused on the goal.