Is Generative AI Bad For The Environment? Explore The Impact
Is Generative AI Bad For The Environment
When talking about artificial intelligence (AI), we often focus on issues like fake news and robots taking over jobs. However, there's another concern that we need to pay attention to—Is generative AI bad for the environment?
Kate Saenko, a computer science professor at Boston University, shared her concerns: "As an AI researcher, I often worry about the energy costs of building artificial intelligence models. The more powerful the AI, the more energy it takes."
Generative AI is a type of AI that can create complex things like sentences, pictures, and even videos. It's different from another type called discriminative AI, which makes choices from a limited set of options.
Lately, Generative AI has gotten even smarter. They provide responses that sound just like a human and make images that look real. But here's something crucial we need to think about: how does AI development affect the environment? In this talk, we'll discuss whether Generative AI harms the planet and what we can do to help.
Is Generative AI Bad For The Environment?
Estimating the precise energy cost of an individual AI model is challenging as it contains multiple factors. They are manufacturing computing tools, model creation, and production usage.
Back in 2018, Google created a generative AI model called BERT. It had around 110 million parameters. They claimed that larger models like BERT are more powerful and effective. However, researchers found that developing this model required as much energy as a continuous flight around the world for one person.
Generative AI Bert
How about bigger AI models like GPT-3? Building it required about 1,287 megawatt hours of electricity, which is a huge amount. This caused the release of 552 tons of CO2, which contributes to climate change.
The emissions from making GPT-3 are similar to what 123 gasoline-powered cars would produce in a whole year! These numbers are just for creating the AI model. We even have yet to mention how much energy it uses when people use it.
But the model's size is not the only thing that affects carbon emissions. There's another AI model called BLOOM, which is similar to GPT-3 in size. Yet, it has a much smaller carbon footprint.
This AI model, created by the BigScience project, only uses 433 megawatt hours of electricity and produces 30 tons of CO2. So, even though it's big, like GPT-3, it doesn't release as much carbon dioxide into the atmosphere.
Moreover, we don't have much information about how much carbon dioxide is created by a single question asked to an AI. Yet, some experts think it could be 4 - 5 times more than asking a question to a search engine like Google.
Meanwhile, more and more people use AI chatbots. AI models are becoming a part of search engines like Google and Microsoft. As a result, the number of questions asked to AI every day could grow really fast.
Generative AI Models
Here are a few examples that show how generative AI harms the environment:
- Data centers: When training generative AI models, we need to use datasets in data centers. These data centers use up a lot of energy and produce harmful emissions, which can contribute to climate change.
- Computational resources: Generative AI models require a lot of powerful computers to train and work properly. This means they use up a lot of energy and produce more emissions.
- Waste: Creating generative AI models generates a lot of waste, including plastic and electronic waste. During the process of making them, we end up producing a large amount of waste that harms the environment.
Use Generative AI Sustainably
Is generative AI bad for the environment? As we stated above, the answer is yes.
Nevertheless, besides thinking about how generative AI affects the environment, we shouldn't forget about the good things it can do too. Generative AI not only brings up environmental issues but also has the potential to solve them. Use it in the right ways to protect the environment and make things more sustainable.
For example, generative AI may assist with making eco-friendly goods and upgrading energy systems. It also is a part of finding fresh approaches to halting climate change.
Let's use generative AI to make our energy systems better and more efficient. This technology has smart models that can study how we use energy and find ways to improve it. They also have new ideas for breeding and distributing energy in a sustainable way.
Use Generative AI Responsibility
Generative AI is also helpful in designing eco-friendly products and materials. By using simulations and creating designs, AI models aid engineers and designers in making things that use fewer resources and are better for the environment. That's why we have eco-friendly consumer goods, transportation systems, and buildings.
How about fighting climate change? Generative AI allows scientists and lawmakers to analyze data and predict what might happen if we try different solutions. This pro helps them make informed decisions and take effective actions to reduce the impacts of climate change.
What if you want to use generative AI in a better way directly? There are several ways you should try:
- Renewable Energy: One way is using clean and renewable energy to power the training and use of generative AI models. That means using energy sources like sunlight or wind that don't harm the planet.
- Energy-efficient Design: Try to make AI models that use less energy. A study from Google suggests that we should use more energy-efficient designs and processors for these models. By taking these actions, we can reduce the carbon footprint of models that are the same size by 100 to 1,000 times.
- Limited Usage: Remember to use generative AI less often. Using it only when necessary reduces the overall energy it consumes. Be careful about using AI and finding other ways to do things when we don't need them.
Is generative AI bad for the environment? Yes. But while generative AI has some effects on the environment, it's critical to look at the bigger picture.
AI models are helping us in various tasks, and this trend will likely continue. And we can still find ways to make it better for the environment.
Let's use clean energy sources and create AI models that use less energy. Remember to use AI responsibly and think about how it affects the world. By doing these things, generative AI can be really helpful while also taking care of the environment.