AI is Killing Our Environment
AI is an umbrella term for a group of technologies that processes information and can generate responses by mimicking human thinking. AI has been around since the 1950s, but the technology's capabilities were limited. Recent technological advancements relating to AI’s computing power and data have quickly made AI explode in popularity and use.
How is AI Affecting the Environment?
While AI can fuel innovation and offer impactful benefits, such as increasing worker productivity and advancing scientific research, the carbon footprint that comes along with its development, maintenance, and disposal is becoming increasingly concerning.
The creation of AI models and computing processes requires large amounts of energy and electricity. Furthermore, when AI models are released to the public, it enables millions of people to use the technology daily which drives up its energy and electricity consumption. The increased amount of AI usage also creates a larger need for model maintenance and performance improvements which elevates energy and electricity use even more.
The frequent increase in energy consumption associated with AI directly contributes to greenhouse gas emissions which in turn contributes to the growth of the climate change crisis.
With the increase of AI model usage comes the increase of high-level computing performance hardware which demands a great deal of water in order to cool down the systems used for training, maintenance, and computing. This can put a strain on local water supplies and potentially disrupt local ecosystems.
According to an estimate, AI-related infrastructure can consume six times more water than the country of Denmark.
Data centers also play a major part in the overall environmental impact of AI. Data centers are temperature-controlled buildings that house all of the computing infrastructure involved in running AI models and are powered by coal or natural gas plants.
The computers stacked at these data centers work to rapidly generate responses at a rate that can take up to 10 times more energy to complete than a google search. Complex questions can even produce up to six times more carbon emissions than questions with more concise answers. Additionally, the microchips used to power AI need rare Earth elements that are often mined in extremely environmentally degrading ways.
Data centers also produce mass amounts of electronic waste. The e-waste associated with these data centers often contains mercury, lead, and other hazardous substances.
Unfortunately, putting a number on AI’s environmental impact is complicated. In order to get a good read of its impact you would have to look at each type of AI model and examine what resources it uses for each type of task it can perform. Many AI companies choose not to disclose any information on their energy and water consumption or optimization techniques, which makes it even harder to measure the environmental impact.
AI energy consumption can also vary depending on how close a user’s proximity is to local AI energy grid and hardware and based on how accurate a model is. “Smarter” models that are more accurate and function more complexly can produce up to 50 times more carbon emissions than simpler models even when answering the same prompt.
Is It Possible for AI to Help the Environment?
While AI has negative impacts, there are ways it can benefit the environment. AI is able to detect patterns in data and pull from historical knowledge to accurately predict future outcomes based on the anomalies or similarities from the data patterns it analyzes.
This process can enhance efficiencies with resource use and energy distribution, monitor the environment and biodiversity in different scenarios, and help guide governments to make more environmentally conscious decisions.
What Can We Do To Limit The Environmental Impact of AI?
Each AI model is made with different efficiency levels based on the level of tasks they are equipped to handle. When choosing an AI company and model to work with, be sure to research which ones have a lesser impact on the environment. Task-specific models tend to be smaller, more efficient, and work well for a regular person’s day-to-day AI uses.
When searching for AI companies, choose ones that are actively working on finding solutions to their environmental impacts and offer transparency. The company Mistral AI has started conducting comprehensive studies that analyze the environmental impacts their AI models produce and is just one example of what other companies should strive to follow.
AI also expends more energy the longer and more complex its answers are, when using AI try to be as straight forward as possible and ask the model to limit the length of the answer to be as short as possible.
In the end, most people don’t need to be using AI in their daily lives. Instead of relying on AI for everything, try going back to basic sources like regular google searches, phone calculators, online encyclopedias, and so on to answer your questions.
