Sending fewer texts on Chat GPT can help save planet; here's how

Most people think information is circulated through cables but few know it is sustained through a continuous supply of water. Every word on an AI bot can cost gallons of water and energy, experts have said. 

The data processors used by cloud services make use of large quantities of water to counteract the heat generated by the ultra fast computation. This cooling water is naturally drawn from surrounding water bodies and circulated through the hot processors to prevent them from getting overheated.

As per Bloomberg, tech giants are raising to expand data centers to support artificial intelligence. Some of these data centers are present in areas with an existing water stress.

Reports suggest that India is also pushing more on infrastructure of AI data centers, with initial development concentrated in few cities like Mumbai, Chennai, Visakhapatnam, and other growing hubs like Noida and Hyderabad.

Major cloud service providers like Google and Reliance are reportedly co-locating data centers in regions where renewable sources of energy like solar, wind and green hydrogen can be used to meet the monumental energy demands.

A thirsty computation trap

All web services make use of computation, an algorithmic process to process information, solve problems or calculate figures.

As opposed to CPU (Central processing unit) which has fewer processing cores that take one task at a time, a GPU (Graphics Processing Unit) uses thousands of tiny core processors at the same time, generating vast amounts of heat.

For AI to run, GPUs are essential to perform parallel processing, processing a large amount of information at the same time.

Each time one drops a query on an AI model, it has to perform a fresh computation to generate a response. This means with every word or number or sentence, a gigantic processor is consuming electric energy which is ten times the amount used for a simple Google search, such as asking the weather of a place.

As per reports, artificial intelligence systems are using about 500 milliliters of water (the size of a bottle) for every brief interaction with the user.

An expert speaking to The Conversation said that AI or language learning modules are not limited to simply responding to prompts but also understanding the context of the question asked by the user.

To cool down the enormous amount of heat generated by these servers, water has to be sprayed over these pipes or cooling towers have to be built for the excessive heat to evaporate. This includes both potable water and other cooling systems using other renewable sources of energy such as solar heat or wind energy.

Cloud services or third party platforms, providing computing services to businesses or individuals, build these massive data centers, preferably in locations where cooling systems don’t drain out the region’s water supply, for instance coastal areas. 

Amazon Web Services (AWS), Microsoft Azure and Google Cloud are to name a few.

Many cloud service providers are now exploring newer strategies to reduce energy consumption.

For instance, the Amazon Web Services has stated using recycled water in 20 data centers around the world. It is also finding a way to return it back to the community.

A report by Economic Times stated that upcoming data centers are moving inland rather than established coastal hubs like Mumbai and Chennai.

With more cities running short of fresh water, there is an increasing risk of consumers computing with these data centers to meet the demands of water.