Artificial intelligence: What is the carbon footprint of generative AI models?

Generative artificial intelligence (AI) has revolutionized technology, from chatbots to image generators. However, the increase in power of these models raises the question: what is its impact on carbon footprint and global warming?

The power consumption of a single AI model is difficult to estimate, as it involves the energy used in the manufacture of the computer equipment, the creation and use of the model in production.

According to euronewsin 2019 researchers found that creating a generative AI model called BERT, with 110 million parameters, consumed the same energy as a round-trip transcontinental flight for one person.

The number of parameters refers to the size of the model, and generally, larger models are more powerful. For example, it is estimated that the creation of GPT-3, with 175 billion parameters, consumed 1,287 megawatt hours of electricity and generated 552 tons of carbon dioxide. This equals the emissions from 123 gasoline passenger cars driving for one year.

It is important to note that these figures only correspond to the preparation of the model for launch, before consumers start using it.

Other determining factors

Model size is not the only determinant of carbon emissions. The BLOOM model, developed by the BigScience project in France, is similar in size to GPT-3 but has a much smaller carbon footprint. It consumes 433 megawatt hours of electricity to generate 30 tons of CO2eq.

A study conducted by Google revealed that, for models of the same size, Using a more efficient architecture and processor, along with greener data centers, can reduce your carbon footprint by 100-1,000 times.

As chatbots and image generators become more popular, and companies like Google and Microsoft incorporate AI language models into their search engines, the number of queries generated daily could grow exponentially. If chatbots become as popular as search engines, lEnergy costs to implement AI could rise significantly.

Another challenge is that AI models need to be continually updated. For example, ChatGPT has only been trained on data up to 2021, so it is unaware of the most recent events. The carbon footprint of creating and maintaining ChatGPT has not been made public, but it is likely to be much higher than that of GPT-3. If the model had to be recreated regularly to update its knowledge, the energy costs would increase even more..

The future is uncertain, but generative AI models are here to stay and are likely to be increasingly used for information and insight.