There’s been a lot of debate about AI’s environmental impact, and rightfully so. The energy demands of large language models are more than concerning. Digging into the research, I did find some examples of AI being used responsibly to address environmental crises. Let me know what you guys think about these uses of AI.
AI for Ocean Cleanup:
The Ocean Cleanup project has used AI since 2021 to map plastic density in remote ocean areas. Their AI-powered system analyzes GPS-tagged images to optimize cleanup efforts. The results in 2023 showed they removed 77 tons of trash from California’s waters.
(They also have 2021 and 2022 reports on how much they removed from the ocean, you do have to dig a little through the reports to find the exact numbers though, just a warning)
Indigenous-Led “Green AI”:
A startup called “PolArctic” which is funded from what I could find by local Indigenous groups such as the Nunavut Fishery Association and Qikiqtaaluk Corporation. They deployed a low-energy AI model trained with Indigenous knowledge to:
Map sustainable fishing zones
Increase fish biomass through regenerative practices
Support infrastructure planning without harming ecosystems
This is a decentralized, community-driven AI, wouldn’t this exactly be what a solarpunk initiative/solution would look like?
https://www.polarcticllc.com/ice#l-case-studies-case-study-2
Sustainable Farming in Africa:
AI is making small-scale farming in Kenya more efficient by:
Optimizing water/fertilizer use to reduce waste
Predicting crop yields to improve food security Unlike Silicon Valley’s profit-driven models, these tools are designed for equitable impact.
https://kenyaai.ke/research/applications-of-ai-in-agriculture-in-kenya/
(Check out the references in this article, it holds a treasure trove of information on this topic)
The Rise of “Green AI”:
The main AI models that are being pushed by Silicon Valley and many others definitely have a high carbon footprint, but I did find a lot of organization that are trying to find solutions:
HuggingFace prioritizes renewable energy and open-source models.
Selecting efficient AI architectures can reduce energy use by up to 90%
The UN’s AI-powered platform (WESR) analyzes environmental data to guide policy.
https://www.sciencedirect.com/science/article/pii/S0925231224008671
I just feel like after looking at all the research the problem might not be AI itself, but maybe the people who are developing it, and who have control of it? Of course how it is used as well, but it more seems like a tool not good or bad but just how people choose to deal with it. I’m curious to know what you guys think about all of this?
https://www.polarcticllc.com/ice#l-case-studies-case-study-2
You must log in or # to comment.
When people complain about energy hungry AI, they’re not talking about AI being used for research purposes. They’re talking about generative models that create text and images that are being pushed onto people when nobody asked for them. Trained on stolen work. Possibly being queried a billion times a day. Enshittifying everything for profit.
I don’t see many complain about specialized models being trained for specialized purposes. Noise cancellation. 3D pose estimation from a picture or video. Helping detect illnesses. Predicting dangerous weather. Folding proteins. There’s many great applications that I can’t think of on top of my head.
Sure, we should use traditional algorithms (I hope I used the right term) where possible and important. They would be less computationally expensive to run and you can actually reason about the implementation, where AI is just a black box that we hope gives us the right answer. But in some cases, it’s not unreasonable.
