Whether you’re using a search engine or AI to answer a query, there is a cost to the planet. Find out which options are the most sustainable.
The impact of a Google search
Every Google search has a carbon footprint:
- 0.5g CO2e = one simple search
- 5.6g CO2e = five minutes web browsing from a smartphone
- 8.2g CO2e = five minutes web browsing from a laptop
According to writer Gerry McGovern, “Google receives 80,000 requests per second or 6.9 billion requests per day. Every day, the Google ‘car’ travels 6.7 million km. That’s 1,325 tons of EqCO2 every day.
“Every year, 483,552 tons of CO2 are caused by us searching Google. You’d need to plant 48 million trees to deal with that sort of pollution.”
How you can help
- Go dark – use Blackle rather than Google
- Use an eco-responsible search tool: eg Lilo which funds social and environmental projects, or Ecosia, which uses searches and ads to fund tree-planting
- Aim to halve the number of Google searches you do each day. Could you try to solve that query instead? Could you focus on creating the best search query to find the answer you need first time round?
- If you’re a content creator, ensure your content is easy to find and navigate, so that your audience can reach it as quickly as possible.
Consider the impact of AI
“We’re seeing this shift of people using generative AI models just because they feel like they should, without sustainability being taken into account,” Sasha Luccioni, the climate lead for the AI company Hugging Face, told The Guardian.
There is limited data on the carbon footprint of AI. Companies that develop them, such as Google, Microsoft and OpenAI, will not disclose how much energy and electricity it takes to train and run their AI models and data centres. [Source: The Guardian]
However, a single generative AI query is estimated to have a carbon footprint that’s four to five times higher than a search engine query. [Source: Euronews]
AI is thirsty
AI is also very thirsty. The water footprint of AI models is enormous.
- Training GPT-3 at Microsoft’s state-of-the-art US data centres can consume 700,000 litres of freshwater.
- ChatGPT needs to “drink” a 500ml bottle of water for a simple conversation of roughly 20-50 questions and answers, depending on when and where it’s deployed. This is a huge volume of water when you consider that ChatGPT has billions of users. [Source: arxiv.org]
How you can help
- Use tools such as ChatGPT and Bard but only when necessary
- Use search engines as a more sustainable alternative to AI when possible
- Research the alternative low-impact, efficient AI approaches and methods if you’re considering building generative AI into an app