THE HAGUE, 20th February. The intersection of sustainability and artificial intelligence (AI) has become a hot topic in global discourse. Governments across the United States and Europe are investing heavily in AI projects, recognising its potential to drive economic growth and address complex challenges. However, this “AI-mania” prompts critical discussions about the environmental implications of AI development.
The Environmental Cost of AI
The main concern of AI criticisers is the energy consumption it requires to train large language models (LLM). For example, training the ChatGPT 3.5 model costs around 500 tonnes of CO2 emission. 500 tonnes of CO2 is roughly equivalent to 1000 cars each driving 1000 km. This energy usage contributes to increased carbon emissions, exacerbating climate change and raising questions about environmental sustainability. To further put it into perspective, let’s compare a simple Google Search with a ChatGPT query. A ChatGPT query uses 10x the amount of electricity a Google Search uses. According to Greenpeace, a typical Google search emits 0.2 grams of CO2, while a ChatGPT query can emit 1 gram.
Making AI Greener
AI’s high energy consumption has sparked urgent discussions about sustainability, with researchers exploring ways to reduce energy use. One approach is model optimisation, where AI systems are trained with fewer parameters to lower energy demands without compromising performance. Another approach is efficient hardware, such as GPUs for minimal power usage.
AI as a Tool for Sustainability
Despite these concerns, AI offers promising solutions for environmental conservation. “Green AI” is artificial intelligence focusing on the development of energy-efficient models and solutions designed to minimise ecological footprints. One such company is Carbon Engineering. Carbon Engineering performs Direct Air Capture (DAC) where they take carbon out of the air and sequester it beneath earth’s surface. In their Canadian DAC facility, they use AI algorithms to optimise the process and increase efficiency.
Another company using AI to help our environment is Xylem, which uses AI to survey and manage water infrastructure like dams and irrigation to minimise water loss and enhance water distribution.
The potential of AI to optimise resource usage for more sustainable and equitable distribution promises a hopeful future of environmental and AI collaboration.
RESONIKS: Our Sustainability Promise
At RESONIKS, we are committed to contributing to a sustainable AI future. Our approach focuses on optimising production processes to eliminate waste, and selecting energy-efficient GPUs for our models. Remarkably, the energy required to train our models is equivalent to running a heater for just one our, demonstrating our energy efficient trainings.
By designing AI algorithms with sustainability in mind, we can harness the technology’s potential to address environmental challenges.
Check out this article from our friends at TU Delft for more insights: