Creating authentic works that originate solely from one’s thought process is a rare sight to see these days. But how can anyone take the blame? I mean, when you have a system so optimized that it can synthesize and create any prompt you give, it’s hard to say no to, especially when you can carefully engineer the work to give it your own touch. But what is said goes a long way: nothing is too good to be true, and, well, there’s always a catch.
It would’ve been so much easier if AI had been this magic machine that created amazing works and insights, so much so that they resembled authentic thoughts. This way, we could have dodged the catch that AI models entail. You might be thinking, “Wasn’t this the other way around? Doesn’t AI provide a solution for the environmental crisis by using historic knowledge and detecting anomalies to predict future outcomes, thus assisting in monitoring the environment and helping make planet-friendly choices?” While this may be the case, the impact of providing such solutions cannot be ignored.
AI’s Contribution to Raw Materials and Waste:
AI deployments can be found within large data centers operated by cloud service providers. Firstly, the manufacturing process takes a toll on the planet. According to the Digital Economy Report 2024 by the UN, such electronics rely on a high amount of raw materials, requiring a staggering 800 kg for a 2 kg computer. Not to mention the high energy consumption required for manufacturing chips, the water required for cooling such systems, and the chemical and electronic waste—such as carbon monoxide, mercury, lead, etc.—which makes one question if such technology is really necessary for us.
AI’s Water Footprint:
Moreover, with the world already facing a water crisis, AI’s water footprint is just making it harder. As reported in the research article “Making AI Less Thirsty,” it was revealed that training the GPT-3 language model can directly evaporate 700,000 liters of water, and that too in one of the best data centers that Microsoft has to offer in the US. And this is only the manufacturing part. The water required for the functioning and cooling of such systems is approximately 16.9 milliliters per inference, estimating millions of gallons annually.
AI and Energy Consumption:
You might think that this would be all, but it’s just the start, as the amount of electricity consumed by AI data centers will shock you. The IEA reported that data centers account for 1% to 1.5% of global electricity. Furthermore, EPRI reports that, with new data centers emerging that require the capacity of up to 1,000 megawatts, the same energy can be used to power 800,000 homes. It may come to mind that Google also requires data centers, so why isn’t the blame being held at the same level as AI data centers? It has been reported that Google searches require 0.3 watt-hours, but at the same time, a request made to ChatGPT consumes 2.9 watt-hours each. The EPRI report also outlines how the AI model development, training, and utilization account for 10%, 30%, and 60% of the energy footprint, respectively.
Balancing Tech Development with Sustainability Goals:
With governments such as the ones in the EU making their goal to increase their share of global chips to 20% by 2030, as reported by The Guardian, their goal to get halfway to net-zero carbon emissions by 2030 seems a bit too idealistic and, frankly, a bit ironic. As the EU invests a staggering €43 billion for microchips and semiconductors, according to a report by the European Commission, it really makes one question whether the sustainability goals are just a façade.
Need for Ethical and Sustainable AI Policies:
We should take into consideration the ethical use of AI models on a governmental and global level and make AI models sustainable and environmentally friendly. Moreover, global policies should be initiated with strict implementation to put this AI ambush on the environment on hold. We should remember that we won’t have any use for AI if there isn’t even a planet to live on.