Generative AI and its potential environmental impact

What are potential environmental impacts of generative AI?

One of the most significant potential environmental impacts is the energy consumption of generative AI. It requires enormous computational power to process the large amount of data, which is needed to generate new information. In January 2023, ChatGPT had roughly 590 million visits. With approximately 5 questions per user, this amounted to the same energy consumption as 175,000 people consume in the same period of time. While this enormous figure only takes into account the electricity of the actual usage, another couple of thousands of megawatt have been used to train ChatGPT. The development of ever more powerful components needed to generate content increases electricity consumption significantly To date, this electricity is still mainly generated from non-renewable energies. Hence, generative AI is right now contributing to climate change and enhancing environmental damage.

Another potential environmental impact is the increase in e-waste. Many devices are being exchanged due to outdated hardware. As a result, the demand for new devices, which in turn require new resources for production, is increasing tremendously. If hardware is not properly disposed of, large areas of the environment are or will be polluted. Bosch wants to take its share in climate protection and is already recycling many electrical parts, as you can read here.

Generative AI also has high (negative) potential to contribute to the decimation of natural resources. The infrastructure and production of generative AI applications requires the use of rare minerals and metals, which are either highly intricate in production or often mined in critical social and working environments.