Sasha Luccioni: The hidden environmental costs of AI models like ChatGPT
As we rely more and more on technology in our daily lives, it's important to consider the environmental impact of the devices and services we use. One area of concern is the energy consumption of AI models like ChatGPT. These models require significant amounts of energy not only during training but also during their everyday use.
The Energy Consumption of AI Models
Training large language models like ChatGPT involves processing terabytes of data over months, emitting large amounts of carbon dioxide in the process. In fact, one estimate suggested that training ChatGPT could emit as much carbon as five cars over their lifetimes.
Recent research has shown that the energy consumption of AI models is substantial. For example, training ChatGPT 3 was found to consume as much energy as 130 American homes in a year. This has raised concerns among climate scientists about the environmental impact of technology.
Measuring and Mitigating Environmental Impact
To address these concerns, AI researcher Sasha Luccioni and her colleagues developed tools like BLOOM and Code Carbon to estimate the energy consumption and carbon emissions of AI models. These tools can help companies make more sustainable choices when developing or deploying AI technologies.
Code Carbon, for instance, allows developers to compare the energy consumption of different models and make informed decisions about their environmental impact. By using tools like Code Carbon, companies and individuals can mitigate the environmental costs of AI technologies.
Choosing Sustainable AI Solutions
Luccioni also advocates for the development of an energy star rating system for AI models, similar to the ratings used for appliances and vehicles. This rating system would allow consumers to choose AI models that are more environmentally friendly for their specific tasks.
While AI technology holds great potential for solving various problems, including environmental issues, it's essential to consider its energy consumption and carbon emissions. By making informed choices and prioritizing sustainability, we can minimize the environmental impact of AI models like ChatGPT.
