In the lead up to COP 26 I’ve noticed an increasing number of references to how Artificial Intelligence is going to save the planet and help us deal with climate change. But whenever I dig down into those claims, I don’t find much evidence to support them.
The UK Government’s National AI strategy, launched in September contains references to how AI can ‘play a role in innovating solutions to climate change’, and while it contains scant details of how this can be done, it does cite a report by Microsoft and PWC which makes the bold claim that AI could help deliver a reduction in global CO2 emissions of up to 4% by 2030. While at the same time delivering a 4.4% uplift in global GDP.
There is no doubt that AI has a part to play in delivering climate solutions like enabling more renewable energy to be plugged into the electricity grid. But there is a danger that techno-fixes like AI are being promoted as future solutions to climate change to distract us from the real (and less popular) solution which is to rapidly phase out fossil fuels, in order to stop pumping more carbon into the atmosphere.
The PWC/Microsoft study makes grand assertions such as that AI can help decouple economic growth from GHG emissions, while at the same time listing a number of somewhat questionable ‘solutions’ to climate change that AI can deliver such as autonomous vehicles; increasing food production via the automation of agricultural tasks and better weather forecasting. Whilst at the same time, the report skates over the reality that the machine learning that powers AI is incredibly energy intensive, and rapidly becoming a major source of carbon emissions in itself.
The report predicts that Europe, North America and East Asia will get the most out of AI given their existing strengths in data and tech. Latin America and Sub- Saharan Africa are predicted to benefit the least. Which answers my next question: who will own this new technology and who will profit from it?
Arguably it will be in the hands of many of the players who have caused climate change in the first place, with limited benefit for those who are experiencing the impacts, many of them in the global south. In fact the study admits that without policy changes to ensure all regions are ready to capture the benefits of AI, economic and climate inequality will be exacerbated, rather than improved.
There is also an admission that the net result of this increased AI will be job losses, but that these will be compensated for by job gains in other sectors of the economy, requiring both businesses and governments to support and retrain displaced workers. Who will be displaced from their jobs, in which parts of the world, and whether meaningful and well paid alternative jobs are available is not clear.
What we do know is that AI is already entrenching existing inequalities in society, by baking bias into automated decisions, where there is no human in the loop anymore. Algorithms are already deployed to: decide where to deploy police; decide whether or not I get insurance or a bank loan; decide how long my prison sentence might be; and whether I’m likely to be the kind of person who cheats the benefit system.
There is no transparency in these decision making processes, and no right to redress, with harmful bias most likely to affect women; ethnic minorities; people from immigrant communities and people from lower socio-economic backgrounds.
In tackling climate change the last thing we need is a tech-fix which could further disadvantage those already suffering the most from its effects. We know what we need to do to avert a planetary disaster, which is political action to limit greenhouse gas emissions. The promotion of AI to help solve climate change could be a dangerous distraction from that path.
By: Emma Gibson ( WLiAI steering member)