By Shivm Mehta
It seems like every day someone prominent is talking about AI. Some praise its potential to solve many of humanity’s problems, while others warn against its unpredictable nature and threats to many aspects of our lives. With Nvidia becoming the most valued company in the world, it seems inevitable that the applications of AI will continue to grow and AI will increasingly impact our lives. However, not many are talking about one of the biggest impacts of AI: the environmental cost.
For starters, with these AI models only getting larger and more advanced, the computational resources needed to train and use these models are also getting larger. For example, these models will continue to use increasing amounts of electricity for development and operation, which will in turn increase the carbon emissions required for these models to be used.
Additionally, with more powerful AI and models being developed, the need for high-performance computer hardware has never been higher. Everyone seems to want to get their hands on the next big AI iPhone or computer. Thus, the carbon impact of these pieces of high-tech hardware will also increase, with more manufacturing, packaging, transportation, and electronic waste all coming from the hardware.
Now you may be asking yourself: how big is this energy cost? According to USA Today, a single prompt into any generative AI machine uses around 1-3 watt-hours of electricity. To put that into perspective, a fridge uses around 1-2 kilowatt-hours of electricity every day. That means for every 500 prompts put into something like ChatGPT, we are using the same amount of electricity as a fridge. Now that might not seem like a lot, but if you take the millions of daily users of something like ChatGPT and also the multiple prompts every day, the electricity cost adds up pretty quickly. For example, the CEO of OpenAI, the company that created ChatGPT, recently stated that the company loses tens of millions of dollars in electricity costs responding to the prompts “please” and “thank you.” Most of this energy is coming from the cost of burning fossil fuels, which further damages the environment.
Now what’s the solution? Unfortunately, many companies today are scaling up their AI development without considering the environmental cost. For example, Meta recently announced their plans for creating a data center similar to the size of Manhattan and spending hundreds of billions on AI development. Just think about the electricity, cooling, and maintenance cost of that center alone! Other companies such as Google are considering the environmental cost and deciding whether AI is necessary for basic tasks such as looking up a recipe or a simple math calculation.
What do you think? Is AI good, a tool to help improve many problems? Does the environmental cost even matter? But most importantly, will you continue to say your “please” and “thank yous?”