AI Dawning
Picture by Igor Omilaev
Article By Helen Lovell Wayne, MS
https://www.instagram.com/agreenerftr/?hl=en
https://www.linkedin.com/in/helen-lovell-wayne/
Every few years, a technology emerges that supposedly changes how society operates. Sometimes it catches on, and other times it becomes a footnote in history. What is alarming about the newest technological wonder taking over the Western world is its environmental consequences. AI models, especially deep learning models, have grown substantially in both size and complexity over the last decade. This change has put a strain on the resources required for both development and the four phases of its operation. These AI models require an immense amount of energy to operate, resulting in an exponential increase in carbon emissions. Taken together, AI has the power to harm or even eliminate all human life on Earth (A. Sharma 2024).
Currently, A1 is utilized in numerous sectors of our economy, including but not limited to: medicine, farming, education, and finance ( Wu, C et al 2022). Managers of these industries use AI algorithms to reduce operational costs by scheduling jobs more effectively (A. Varghese 2022). But the problem with this rationale is that it assumes AI has all the data to make informed decisions. That is often not the case. In practice, it uses three times the energy as other technologies, and in many cases it produces inaccurate answers (Choudry, A., S. Vivastava, and V. Sharn 2026). Some of these answers have serious consequences.
💧AI CAUSES ENVIRONMENTAL PROBLEMS💨
The scope of the AI environmental problem is far-reaching. It uses power in every phase of it’s existence. Researchers estimate that training a single large AI model such as GPT-3 releases over 284 tonnes of CO2 (A. Zhuk 2023). To put this into perspective, an AI model can release as much carbon dioxide as the lifetime emissions of five average-sized gas-powered cars A. Zhuk 2023). That is just one phase of one model. AI models have 4 distinct phases. All these phases utilize large amounts of energy. The acquisition and transportation of that energy also cause environmental harm.
If the environmental harm caused by AI were simply limited to energy usage, the problem might be lessened by switching to renewable energy sources to provide AI’s power needs. Unfortunately, energy is just the tip of the environmental iceberg. This technology also reduces biodiversity, contaminates the water (with heavy metals and microplastics), air, and soil. It also creates plastics (for more information on plastics, read The Plastic Alchemist) and generates e-waste (A. Zhuk 2023). Plus, as previously mentioned, it releases enormous amounts of carbon dioxide.
🧊 AI ATTEMPTS TO MINIMIZE DAMAGE 💻
Additionally, the above-mentioned e-waste contaminates the earth’s atmosphere and soil with toxins. The AI machines contain elements and compounds that can physically damage all animal life. Examples include: lead, mercury, and flame retardants. The cooling off of AI machines with water can cause these elements to seep into the earth, which can migrate to nearby waterways (A. Zhuk 2023). This inevitably poisons the entire food chain.
To be fair, some measures are currently being taken to reduce AI’s overall environmental impact, but at this moment, they are not sufficient. Engineers and specialists are currently working to reduce the energy used by AI. This is accomplished by increasing server efficiency, optimizing cooling systems, and designing data centers to be energy-efficient (A. Zhuk 2023). The motive is usually cost savings rather than environmental concerns. However, it is a step in the right direction.
💰 AI DATA PROBLEMS 🌎
Ironically, AI is being utilized to “protect” the environment from the very same issues that it is causing. For instance, it is being used to improve pollution monitoring and control. Sophisticated sensors, statistical analysis, and autonomous response systems accomplish this. In theory, this offers real-time and scalability solutions (Q. Wang, Sun, T. and Li, R. 2023). In other words, instead of having human beings take samples (which can be expensive), AI will simply predict where all the water and pollution are occurring. Then AI will suggest improvement measures. Pollution results in the premature death of 9 million people every year in premature deaths a year (Choudry, A., S. Vivastava, and V. Sharn 2026). In theory, AI could help prevent some of those from occurring.
Notwithstanding its benefits, AI modeling presents numerous problems with accuracy. These systems depend significantly on gathering an extensive amount of data. For environmental issues, these topics include but are not limited to: topographical, social, and economic information. All this data is crucial to predict precise forecasts and modelling. Gathering this data creates issues with privacy, ownership, and governance. Consequently, environmental datasets frequently exhibit fragmentation, bias, or inaccessibility, hence constraining the ability to scale AI applications (Choudry, A., S. Vivastava, and V. Sharn 2026). AI models developed on inadequate data are prone to perpetuating systemic biases, resulting in both unjust and inaccurate outcomes.
🤢 AI CONCLUDING REMARKS 🚫
As a society, we have created a new technology with AI. It appears to have some benefits, but at the present time, the costs appear to outweigh the benefits. Especially considering that the data is not completely accurate. The average person can limit their use of AI to what is completely necessary. Then use well-established techniques for everything else. In this situation, it appears that we have taken two steps back, one step forward.
