“ChatGPT, tell me…”: The Impact of Artificial Intelligence on the Environment

the artificial intelligence He entered the life of a modern person forever, however, it seems that he himself did not calculate the energy costs of his “friction” with artificial intelligence technologies in terms of the environment.

While scientists are sounding the alarm about the need to save energy, AI technologies are still moving along a particularly energy-intensive path.

Liberal.gr talked about this issue with Spyro LyuliAnd Member of the Forbes Technology Council and Senior Advisor to EYP MCFHeadquartered in New York, it is engaged in the design, engineering and implementation of data center facilities.

Interview with Christos W. Panagopoulos

Mr. Liolis, there is now an abundance of references on the Internet on AI technologies, but no one yet seems to have thought about the environmental consequences of AI systems. What is happening exactly? What is the environmental impact of artificial intelligence?

A recent discussion with industry leaders led me to consider the environmental costs of a productive AI model and its impact on data center facilities. My research began by entering the claims into the resource itself:

How much data center power in kilowatts (kW) does ChatGPT consume per day to answer typical user queries? On average, how many queries does ChatGPT answer?

ChatGPT: I do not directly consume power, as I am only on the computer servers and I use the computing power provided by said servers to generate responses to user queries. […] However, some estimates suggest that large data centers can consume anywhere from a few megawatts to a few hundred megawatts of power, depending on their size and usage. […] However, I can give you a rough estimate based on certain assumptions. […] So a rough approximation of the power consumption for a single front lane answering a typical query is about 0.00085 kilowatt-hours (kWh).

Ongoing research on this topic has revealed some surprising facts about the model’s energy consumption.

According to some models, it is estimated that Consume GPT-3 training 1.287 megawatt-hours (MWh)that they broadcast 552 tons of carbon dioxide equivalent (CO2e).

GPT-3 form, which contains 175 billion parameterscan require thousands of GPU chips (estimated at around 30,000) and millions of GPU hours to generate a single response.

The data scientist only estimated that in January 2023 ChatGPT’s electricity consumption can reach 23,364,000 kWh. That’s an impressive use of power from a single model of productive AI.

You talked about productive AI. What models are widely used nowadays?

There are many AI technologies and paradigms in use today, such as DALL-E and GPT-3, among others, and while the technology is still in its infancy, it is beginning to make its way into many industries, such as marketing and media. , financial services, games, 3D video, fashion, textiles and other materials science, to create new designs and experiences.

Technology allows engineers, scientists, designers, and artists to quickly experiment with different styles, compounds, and materials, resulting in more innovative, functional, environmentally friendly, and personalized products. Besides creating and testing digital twins, generative AI is expected to be one of the fastest growing markets.

What is the impact of productive AI on data centers?

The impact of produced AI on data centers will be significant, as these systems require large amounts of computing power and storage to train and create new content, and the foundation of productive AI is just that — to analyze massive amounts of historical data to create new content. content, which in turn can be used in the future.

As data centers continue to evolve and expand, traditional data center lifecycle approaches may not be enough. The level of complexity of the environment requires a new way of thinking. Creative AI is transforming the way data centers are designed, operated, optimized and maintained. The same benefits of productive AI can be used to revolutionize the way data centers are built and managed.

Therefore, data centers will be the beneficiaries and beneficiaries of productive AI.

How do data centers enable the productive AI model?

Data centers support information technology production through high-performance computing (HPC), hosting specialized hardware, data storage, and networking. One of the biggest challenges associated with generative AI is the need for computationally intensive resources, such as graphics processing units, to effectively process the huge amounts of data required to train these systems. This can lead to a significant increase in power consumption and cooling requirements for data centers, as well as the need for more advanced infrastructure and networks to handle the increased demand.

Building a modern, resilient HPC data center infrastructure that is also optimized for genetic AI workloads provides an exciting opportunity for data center operators to cater to this niche market.

Early adopters of the productive AI model are players in the general application fields. Typical use cases include basic or advanced searching and creating media and images based on existing public data from the Internet, such as creating documents, campaign emails, and even pieces of competitive information.

Early hysteria for AI is now giving way to more customized, industry-specific use cases across all industries, including the architecture, construction, and engineering (ACE) sector. These use cases are based on AI produced not only on the web, but perhaps more importantly, on large internal data sets, knowledge libraries and permanent staff, along with forward-thinking professionals in the organization.

While data center facilities were often seen as unattractive warehouse buildings, today architects and engineers are using AI solutions such as DALL-E and Midjourney to design conceptual building images for their projects. Designs and graphics that used to take weeks to create and hours of computing to render now take just a few keystrokes and clicks and can be instantly modified.

In addition, AI manufacturing can provide advantages during the detailed design and construction phases by analyzing design options, consuming standards as input (such as Uptime Institute, IA-942, or other mechanical electromechanical installations (MEP)) and optimizing space power cooling (SPC) outlets. It can also assist with scheduling options, material handling, construction machinery such as cranes, run “what-if” scenarios to check scheduling impacts and report progress against schedule, identify supply chain delays and even recommend recovery or fast-track steps.

What is the future of artificial intelligence on a financial level?

According to market research firm MarketsandMarkets, the global artificial intelligence market is expected to grow from $11.3 billion in 2023 to $51.8 billion by 2028, at a compound annual growth rate of 35.6% over the forecast period.

Much of this growth can be attributed to generative AI, or systems that can create new content, such as text, images, or sounds, in response to prompts made by users—versus predictive AI, which provides answers or predictions based on existing data.

In fact, there aren’t a lot of conversations I have today, whether professional or recreational, that don’t mention Open AI’s ChatGPT, one of the latest developments in the field of production AI. Topics vary from how to use them for writing emails and messages, planning events and vacations, or just for fun.

Does Productive AI Have Limitations? How plausible are the scenarios about the marginalization of modern man?

Although industry AI is a powerful technology for improving many aspects of data center facilities, it has its limitations and drawbacks.

FirstlyAs with other AI or data-driven technologies, generative AI models require a large amount of high-quality data to be effective. Processing massive amounts of data requires rapid access to computing, storage and networking toolkits. Especially for building data centers, this data can be difficult to obtain due to the complexity and variety of the construction process, as well as the variety of data sources over the years, including legacy infrastructure systems as well as CAD files, programming and spreadsheets, among others.

With such diversity and complexity, AI production models may not be able to take into account all the variables and limitations involved in building data centers, which may limit their effectiveness.

secondlyExisting AI models may not design solutions that are practical or feasible in the real world, in a given location. For example, a model could produce a design that is optimized for energy efficiency, but may not be structurally sound or meet regulatory requirements.

end, Productive AI does not replace human expertise or expertise. While it can help improve some aspects of building data centers, it should always be used in conjunction with human knowledge, practical experience, and practical judgment to ensure that the final product is safe, efficient, and efficient.