Generative AI Tools Effect on the Environment

The impact of artificial intelligence is taking center stage, questioning its impact on humanity and social order. The discussions need to examine whether ChatGPT and other generative artificial intelligence (A) tools will massively affect the environment. Despite the likelihood, there is still a reason to remain optimistic.

Extending Artificial Intelligence Impact on the Environment

There have been conversations concerning the dangers posed by generative artificial intelligence tools, especially in how they will replace human jobs. Additionally, Kate Saenko, a professor at Boston University, has further raised concerns regarding the tools’ effect on the environment.

According to the scholar, whose sentiments were published by The Conversation, Saenko is more concerned about the energy costs associated with developing AI tools. In this case, a robust AI is highly likely to consume more energy.

So far, several studies and debates have focused on the energy use of blockchains such as Ethereum and Bitcoin. However, the impact of quick AI development on the globe has yet to attain much attention.

Insufficient Data on Carbon Footprint Attributed to Generative AI

According to Professor Saenko, making crucial changes is necessary, but he still admits that there needs to be more data regarding the carbon footprint of any generative artificial intelligence query. Despite this, she claimed that according to research, the data is four to five times greater than a simple search engine query.

A 2019 report by Saenko revealed that a generative artificial intelligence model referred to as the Bidirectional Encoder Representations from Transformers (BERT) used a significant amount of energy. Specifically, its energy use was equated to a round-trip continent-wide flight for a single individual.

For artificial intelligence models, parameters refer to variables acquired from data guiding the model’s predictions. The higher the number of parameters, the higher the intricacy of the model, which ultimately translates to the need for additional computing power and data. Further, the adjustment of parameters occurs during training to reduce mistakes.

GPT-3 Model Yielded Tons of Carbon Dioxide

Professor Saenko claimed that the energy consumption of OpenAI’s GPT-3 model, comprising 175 billion parameters, was equal to 123 gasoline-enhanced passenger cars driven for a year. Alternatively, this figure is approximately 1287 megawatt-hours (MWh) of power. Besides, she claimed that 552 tons of carbon dioxide were generated. These figures are based on the model before its introduction to customers.

Senko stated that the energy costs of AI deployment might rise significantly in case chatbots gain similar popularity as search engines. Her statement was based on Microsoft’s inclusion of ChatGPT in the Bing web browser.

Another major challenge is the introduction of mobile applications by several artificial intelligence chatbots, for instance, ChatGPT and Perplex AI. These situations ease their utilization and enhance exposure to a vast audience. Professor Saenko stated a Google study that revealed that a reduction in carbon footprint could be attained via the utilization of a more effective model architecture and processor as well as a greener data center.

Miguel Faus, a Spanish Film Director, reveals how he used NFTs to fund his Calladita film. According to him, blockchain and Web 3 enable creators to handle some major complexities associated with the film industry. Particularly, getting funding is a major challenge in the industry.

Unrestricted Development of AI Poses Major Threat to Environment

Saenko claimed that one big AI is incapable of destroying the environment. However, if several organizations create slightly different bots for various purposes and use by millions of clients, energy utilization becomes a major issue.

Finally, she concluded that additional research would be vital in enhancing the efficacy of generative AI. She wrote that AI is capable of running on renewable energy. The introduction of computation to places with ample green energy or planning for computation during the availability of renewable energy can significantly reduce emissions by 30 to 40.