|
A recent study published in Nature sheds light on how the environmental impact of large language AI tools compare to human labor for the same tasks.
The study found that although large AI models like ChatGPT use a lot of energy, in some cases they may have a much smaller environmental footprint than humans do. For example, a person in the US who writes a 500-word document might have 40 to 150 times the environmental impact of a typical AI model. For simpler AI models, that difference can be even greater: up to 4,400 times.
However, these numbers depend heavily on where you are in the world. The gap is smaller in countries where the average person's carbon footprint is lower than in the US. For instance, a person in India might only have 3 to 16 times the impact of a standard AI model. This shows the importance of considering local energy sources, labor conditions and other regional factors when calculating AI’s true environmental costs.
Another important point raised by the study is the need for more transparency. Just like how we can now see estimated CO₂ emissions when booking a flight, we may soon expect similar disclosures when choosing between AI tools. Knowing the carbon footprint of different models could help users and companies make more sustainable decisions.
So, could AI help reduce emissions? Maybe, but only in certain situations.
As AI models become more complex, their energy use increases. There are also social impacts to consider, like the risk of job loss as AI replaces some types of work. As AI becomes more common in the workplace, both the environmental benefits and the human consequences should be evaluated.
| 
