AI's effects on climate change: Both good and bad
With more companies building supercomputers and infrastructure that requires a lot of compute power, AI may be doing more harm than good to the planet.
Severe wildfires, raging storms and other extreme weather conditions are all indications that the climate is changing and not for the better.
Earlier this month, the United Nations Intergovernmental Panel on Climate Change released its sixth report on the assessment of global climate conditions. The report looks at environments that are growing warmer, rising sea levels and species becoming extinct. The warning is clear: Something must be done to save the climate.
But some have attempted to use AI to combat climate change.
For example, Microsoft's AI for Earth is meant to create AI projects that fight global climate change. Also, Nvidia has revealed plans to build a supercomputer that predicts climate change.
However, the use of AI technology to counter climate change is in some ways contradictory because of the technology's tendency to consumer large amounts of energy.
In this Q&A, Neil Sahota, lead artificial intelligence advisor to the United Nations and CEO of AI vendor ACSILabs, discusses both how AI technology is helping to mitigate climate change, and the harmful effects of the technology. According to him, the key to using AI for climate change is innovative thinking.
How is AI technology helping to combat climate change?
Neil Sahota: There's a lot of things you can do in terms of modeling, digital twins, using AI with generative design to try to figure out new [tools] to try and remove carbon from the atmosphere. From a tool process side, AI has been helpful.
On the flip side, all this comes with a cost. AI requires a lot of computing power, a lot of data, a lot of infrastructures, and all that does contribute to climate emissions.
These big powerhouse servers and supercomputers that we need to enable the AI capabilities [involve] a lot of carbon that gets generated from the production. There is unfortunately some hazardous material waste in making some of these machines and then they have to be transported to the locations where these data centers are.
Now we're building these gigantic buildings to house all these computers and power AI and cloud services.
How can we mitigate the negative effects of AI on the global environment?
Sahota: People are looking at more eco-friendly ways of producing computers to restrict some of the electrical consumption and emissions and the use of fans.
Our bigger challenge is that we are so used to how computers work, or how we think they work, that we are not thinking more disruptively. We are not asking ourselves: Is there a completely different way to manufacture a computer? Should we power computers without fans or anything like that?
Neil SahotaLead artificial intelligence advisor, United Nations
It is really about finding [new] ways to actually manufacture and run servers; that's going to be the real mover for us. It is just a lot of people are not wired to think that way.
Even if we solve the hardware side, we still need to address the software side. We're not looking at optimal ways to code and run structures because memory storage CPU power is cheap. If we pay attention to that, that will also help.
I think [it] require more radical changes in how we develop hardware and software to reduce the impact from all our technology on the climate.
How do you get people to think innovatively about climate change and AI technology?
Sahota: It involves how we teach people about these tools and how to program. We are so focused on the technology and the capability enablement that we've kind of stripped out some of the infrastructure and impact concerns. And we need to reintroduce all these things and help people understand and be more consciously aware of [their impact] when we do something.
Should there be a pause with AI innovations?
Sahota: I am not calling for a pause. I am calling for a more proactive approach.
The real challenge we have is we don't make that investment in fine-tuning. We must try and plan things out a little bit more to better understand [where] ripple effects come into play.
It's just like what we see with all the cryptominers taking tons of cloud resources. All that kind of stuff takes a big environmental toll. Again, is this the most effective way to mine Bitcoin? No, we know that. But it works. That it works is not good enough.
Do you have any final thoughts on AI technology and climate change?
Sahota: We have a real opportunity here because as other technologies come up, talking about the metaverse and all that, it's going to exacerbate the problems.
This is not a problem that is going to get solved magically in two years.
Everything we do in our lives has some sort of impact. Human beings are good at dealing with that fast-moving immediate threat. That is the way we are wired.
It is the slow-moving long-term threats that we are not good at. And if we do not start doing something about climate change, we are toast.
Editor's note: This interview has been edited for clarity and conciseness.