"The Dark Side of Artificial Intelligence… Elitsa Britton in front of FACT

Artificial intelligence (AI) is increasingly entering our lives, but it turns out that it is very “hungry“ for water and energy. Elitsa Staneva-Britton spoke on the topic in front of FACT. She is an educational psychologist, graduated with a bachelor's degree from Sofia University “St. Kliment Ohridski” and a master's degree from the “Ludwig Maximilian University of Munich”. Her scientific interests are in the field of ecology and are related to the effect of climate change on human psychology and the concept of climate justice.

- Ms. Britton, when will the “dark side“ of Artificial Intelligence engulf us, because it turns out to be an insatiable consumer of energy resources?
- Before we talk about the topic, let's note that artificial intelligence (AI) has its bright side, namely its huge potential for creating innovations in the fields of ecology, education, medicine, engineering and finance.

The problem with the “dark side“ of AI comes not from the technology itself, but from how it is used. Often, AI companies focus on the benefits at a purely organizational level, such as saving costs by automatically introducing applications and selecting employees.

When using AI individually, its use is usually reduced to easy solutions – for creating content, often with unclear credibility, optimizing processes, even reading a book is reduced to a short summary. It is precisely because of the frequent use of AI for not very important things, that resource consumption increases, which in the long run would reveal The “dark side” of AI.

- Why does a ChatGPT search use about 10 times more energy than a regular search…
- While traditional internet search relies on retrieving already existing information published online, generative intelligence is based on a huge amount of data, through which it is trained and subsequently produces new content.
It is precisely for this generation of new content that the entire artificial neural network of AI is activated, in which energy consumption increases significantly.

- ChatGPT's energy consumption in 1 day is equal to about 564 MWh or the average amount of energy used by a household in 1 month. How will this consumption grow…
- This depends on how we use AI and whether companies will comply with the regulations entering the sector.
It is interesting that for the moment the data on The issue remains hidden from most tech giants. For now, according to the Hugging Face platform, the BLOOM model consumes 433 MWh to train a database, while other models such as GPT-3, Gopher, and Open Pre-trained Transformer (OPT) use 1287, 1066, and 324 MWh to train terabytes of data, respectively. Along with training models, energy is also spent on servers and chips, with the giant NVIDIA reporting a profit of $13.5 billion in the second quarter of 2023 alone. Experts suggest that AI can be more environmentally sustainable. This can be achieved through appropriate chips that prioritize energy efficiency over the computing power of AI itself. Such an action would optimize energy consumption by 100 to 1000 times. Another step is the development of technologies that reduce energy and water consumption (especially for cooling) in data centers. As early as 2023, the laboratory The Massachusetts Institute of Technology's Lincoln supercomputer center is developing techniques to reduce energy consumption.

- By 2026, the artificial intelligence (AI) industry will consume more than 10 times more energy than in 2023, because...
- According to some experts, companies are fighting a cold war over AI. Financial interests are stronger than social and environmental risks. Meta, Microsoft, Apple and Google have unabated plans to invest billions in AI infrastructure.
On an individual level, one seems to forget that behind digital terms like cloud space in AI infrastructure are physical devices that consume energy. And so we continue to use more and more AI.

- It takes about 8,300 liters of water to make a single chip, and about ½ liters to write an email from an AI bot. What does this tell us?
- Yes, the water consumption in building, training, and using AI models is also a serious waste of resources and risk. Additionally, data centers also consume water for cooling.
In a world where climate change is already limiting the availability of fresh water, this data suggests that we are not thinking about how more resource-hungry AI is putting additional pressure on already dwindling water supplies.

- Bulgaria is among the six European countries where an AI factory will be developed. The BRAIN ++ project is a huge economic opportunity, but it also poses challenges in terms of energy consumption and water consumption. And we only have the Kozloduy NPP…
- That's right. Bulgaria is also one of the six countries in Europe where new AI factories will soon be built, namely BRAIN ++. This is a huge economic opportunity. But such projects hide unclear challenges due to energy consumption and water consumption, especially given the water scarcity in Bulgaria, as well as the fact that we still rely largely on fossil fuels in our energy. According to EU data, our country is among the countries with the highest energy poverty at the moment.

- How much exactly does AI consume and how does it affect climate change?
- Apart from the aforementioned data on the consumption of natural resources, no one can say exactly how much AI consumes. Such data is not public and due to the multitude of companies, models, applications (in search engines, shopping, social media), the scope of such calculations remains an overwhelming task.
Interestingly, in the EU, AI companies are required to document the energy consumption of their models, while at the same time noting societal risks due to the negative impact of AI on health, security, equality and climate.

- What are the financial interests in the sector?
- Financial interests are definitely stronger than social and environmental risks. Meta, Microsoft, Apple and Google have unabated plans to invest billions in AI infrastructure. In organizational circles, there is talk of optimizing processes, saving time for creative activities, developing new abilities. And whether employees are so dedicated to self-development… remains questionable. Unfortunately, investments in AI in social activities, education and medicine, which can have extraordinary benefits, are lagging behind the plans of the big tech giants.

- Months ago, US President Trump - along with OpenAI - announced the $500 billion Stargate project to create data centers. How big is this business...
- Data centers are a huge business and data shows that by 2028, AI will consume over half of the energy in these data centers.
Unfortunately, the risks are also huge. In the aforementioned project, for example, forecasts indicate that each such center would swallow more energy than that used by the US state of New Hampshire.

- Water consumption in the development, training and use of AI models is also a serious waste of resources and risk. And the Earth is getting drier and warmer. What is this trend and are we paying enough attention to it, or are we just living our lives, and then after us…
- As we mentioned, this is a really serious problem, but we often distance ourselves from the problem or go to extremes. We need precise calculations of global energy and water consumption so that we can together and consciously help reduce electricity consumption and create a sustainable society.

- How can we reduce the energy footprint of AI?
- Against the backdrop of financial interests - while companies continue to attribute individual guilt - we can reduce the ecological footprint of AI companies: not to use AI for unnecessary activities in our daily lives, not to thank ChatGPT, to redirect to a more sustainable model. In fact, DeepSeek, with its launch on the market, proved that AI models do not necessarily have to be extremely expensive and energy-consuming. Individual steps are not enough. Companies also need to take action, and one of them is to move towards more sustainable models and technologies. AI has enormous potential in the fight against climate change, and in an ideal world, corporate interests would be focused on this.