🌱 Environmental Considerations

AI uses a lot of compute power, which burns a lot of electricity. You must think about the thical implications of using a technology that requires this much power.

Environmental Impact

While AI has the potential to solve many challenges we have to acknowledge the energy consumption that the best AI models require. As AI proves itself over time, the size of models and their energy requirements increase exponenetially.

The GPT models are estimated to have produced over 500 metric tonnes of Carbon Dioxide to train. It is unknown how much energy they consume from their usage, particularly as it has grown exponentially.

Meta has several models it trained with over 1,000 tonnes of CO2 equivalents produced. The company does work to use renewable energy for a vast majority of their data warehouses. While these numbers aren't as jaw dropping as the numbers of other industries, they are large.

There are some designe deisions that can be made in the construction of models that can reduce the energy requirements of large models. This is an area, which still requires research and will have tradeoffs. Because Neural Networks aren't entirely legible, performance tradeoffs are not well understood.

Cloud Computing provides between 2 and 4% of global greenhouse gas emissions. This is larger than the aviation and shipping industries. The future impact of AI on the environment is not to be understated. All of this being said, cloud computeing does create economies of scale, and is actually 72-98% more effecient than hosting IT infrastructure on premesis.

Microsoft

Microsoft is invested in many of the most prolific AI products, like Github CoPilot and OpenAI. This means that their data centers use a lot of electricity, and water.

Microsoft has a policy to be carbon negative, replenish more water than they use and have 0 waste by 2030, and to have removed all historical emissions by 2050. They've aggressively acquired 2.5 million tonnes of carbon removal and have replenished 1.3 million cubic meters of water.

What's most impressive about Microsoft's initiatives is that they take Scope 3 emissions into account. They are taking responsibility for their entire supply chain. While some of this may be marketing, and it's unclear what offsets they're buying. Microsoft is working on its Emissions responsibilities.

Google

Amazon

AI Impact

Overall, AI must contend with the environmental impact of the technology. Many tonnes of Carbon Dioxide are released into the atmosphere every time a model is trained or called.

More coming soon...