The era of artificial intelligence (AI) is here, and it’s changing our everyday lives. We use AI for everything from cooking recipes to workout plans, songwriting, and even programming. It’s even capable of altering images. However, we need to think about its impact on the environment now, rather than waiting decades like we did with the industrial age.
Many people still believe that digital technology can operate without consuming resources, but that’s not true. As Kate Crawford discusses in her book, *Atlas of AI*, AI is made from various resources, including natural materials and human effort. Calculating the contribution of workers, especially in lower-income countries, is a significant challenge, and governments often struggle to represent these figures accurately.
Let’s break down how AI works. Its life cycle includes two main parts: hardware and software. The hardware begins with extracting raw materials, creating components like processors and memory, assembling devices, and setting them up in data centers. The software side involves data collection, model development, initial training, and testing before deployment.
You may wonder how this process harms the environment. Current information is limited, but AI’s energy consumption is noteworthy. For instance, during the training phase of AI models, substantial energy is consumed—sometimes in the hundreds of thousands of kilowatt-hours. Manufacturing hardware also creates environmental impacts, as it often involves mining rare earth metals.
When we run AI models, there are ongoing energy costs too. Every action taken—whether it be querying an AI or maintaining the required infrastructure—results in a carbon footprint. It’s estimated that one query to a service like ChatGPT uses enough energy to power a light bulb for about 20 minutes. When millions of users engage with these tools daily, the total energy consumption becomes substantial.
Additionally, the tech industry is projected to contribute to a significant portion of global greenhouse gas emissions in the coming years. For example, a study indicated that just training the GPT-3 model produced as much carbon emissions as thousands of round-trip flights. Moreover, data centers demand enormous amounts of water to cool server systems, similar to the daily water consumption of a small town.
Experts have highlighted the pressing need to assess how different AI models impact the climate. Various models require varying amounts of energy to operate. For instance, larger models may produce significantly higher emissions than simpler ones, making it a complex task to fully understand their environmental effects.
In recent years, large tech companies have seen their emissions increase as they expand data centers to support growing AI workloads. Microsoft and Google, for instance, have reported climbing emissions tied to their AI developments. Data centers are already consuming as much electricity as an entire country and that demand is expected to skyrocket in the coming decade.
Looking to the future, experts stress the need for innovative approaches to make AI greener. Running data centers on renewable energy and designing more efficient hardware are just a few ways we can help mitigate the environmental impact. However, the rapid advancement of AI technologies poses a challenge, as it often outpaces these green initiatives.
As we embrace the benefits of AI, it’s crucial to confront its environmental challenges with urgency and transparency. If we continue down this path without awareness, we may find ourselves facing consequences we could have mitigated with foresight.
Source link
AI,age of AI,digital technology,artificial intelligence,environmental costs,data centres
