The more powerful the AI, the more energy it consumes. So what does the rise of increasingly powerful generative AI models mean for the future of our carbon footprint? “Generative” refers to the ability of AI algorithms to generate complex data. Another option is “discriminative” AI, which chooses between many options and produces only one number.
Generative AI can create more complex outputs such as sentences, paragraphs, images, and even short videos. It has long been used in applications such as smart speakers to generate audio responses or autocomplete to suggest search queries. However, it has only recently gained the ability to generate human-like language and realistic photos.
The exact energy cost of a single model is difficult to estimate, including the energy used to manufacture computing equipment, create models, and deploy models in production. A study conducted by Google found that for the same size, using more efficient models and processor architectures and greener data centers can reduce the carbon footprint by 100 to 1000 times. However, larger models do consume more power during installation.
Read also: ChatGPT’s Potential to Aid in the Search for Extraterrestrial Life on Mars, Experts Discuss
Going back a few years, many people outside of research labs didn’t use models like BERT or GPT. That all changed in November 2022 when OpenAI released ChatGPT. According to the latest available data, ChatGPT had over 1.5 billion visits as of March 2023. Microsoft incorporated ChatGPT into its Bing search engine and made it available to everyone in May 2023.
If chatbots become as popular as search engines, the energy cost of running AI may increase significantly. However, AI assistants have many uses beyond simple searches, such as writing documents, solving math problems, and creating marketing campaigns. The carbon footprint of ChatGPT’s creation is not publicly known, but it is likely much larger than GPT-3. If it has to be updated regularly to keep the information up to date, the energy cost is even higher.
The future is hard to predict, but large-scale generative AI models are here to stay, and people can increasingly use these models to gain information and knowledge. Even if a single large AI model does not harm the environment, energy consumption can become an issue if a thousand companies develop slightly different AI bots for different purposes and each is used by millions of customers.
More research is needed to make it more effective. The good news is that AI can be powered by renewable energy. Likewise, public pressure can push companies and research labs to publish the carbon footprint of their AI models, as some companies have already done. In the future, consumers may even use this information to choose a “greener” ChatGpt.
Source: euronews.com
