AI: The New Nuclear Supervillain?

Did AI just become a nuclear supervillain? Possibly—but not in the traditional Marvel Comics sense. If we’re not careful, AI might just consume all of our extra energy and turn us into paper clips, as illustrated in Nick Bostrom’s Superintelligence. Unless we put on our energy-saving superhero suits and smart prompt-writing gloves, we just might be able to save the world.

There’s a little secret about artificial intelligence that isn’t so secret anymore. It’s very, very hungry. It’s easy to think that using AI has very little cost, since most users don’t experience the expenses associated with training and operating the various AI systems we’ve all come to enjoy in our daily lives.

The Insatiable Energy Appetite of AI

So, just how hungry is AI, you ask? It’s as voracious as an entire city. Let me explain.

Last month, Microsoft signed a deal with Constellation Energy to restart some of the nuclear reactors at Three Mile Island (yes, that Three Mile Island) to secure additional energy for its AI development and operations. Under the terms of the agreement, Constellation will sell 100% of the energy produced to Microsoft for twenty years. Do I have your attention?

Keep in mind that this is just one of many companies currently operating that will need to find ways to develop, maintain, and operate their AI models in the coming years. However, not all of them will be able to open nuclear power plants to support their efforts. So what will they do? Most likely, other companies will need to rely on non-renewable energy sources—which could have enormous impacts on the environment if consumers (you and me) and companies alike are not mindful of how these energy-hungry systems are used.

Now, how much more energy do these models use? While these are rough numbers that can vary depending on the model, a solid base estimate is that generative AI models use 10 times more energy per query compared to a traditional Google search. Some may be lower; others may be significantly higher—but these are the numbers and ratios that are generally accepted.

Energy Challenges in Mexico

As we’re still in the process of bringing more people into the AI world, this means that, as a global community, we need to start seriously considering how we will provide all this energy to power everyone’s daily lives—plus the AI products they will be using.

If we take the context of Mexico in particular, stable electricity and water supply are two major challenges that have slowed the development of more advanced technologies and manufacturing, such as semiconductor plants. However, it’s worth noting that there are some exciting new projects in the works that we’ll discuss in future articles.

It’s only natural that people in Mexico want to participate and thrive in the technology era driven by artificial intelligence. But for that to happen, we need to take a closer look at the limitations that could hinder the development of more advanced technologies and manufacturing in the country—namely electricity and water. So, what should we do when we’re dealing with such a power-hungry AI?

Becoming AI Superheroes

There are three things we can all do to become real-life AI superheroes:

  • Conscious use of AI: Make sure that you and your coworkers are being thoughtful in how AI systems are used. Are you using the system the way it’s meant to be used? Are people in your office trained to use generative AI efficiently—maximizing results with the fewest prompts possible?

  • Political participation (Political warning): Vote for measures and candidates who focus on the perhaps boring—but crucial—topics of infrastructure (both traditional and digital), such as sustainable water and energy production, or those who support the development of tech infrastructure at the national and local level.

  • Education and outreach: Share what you’ve learned about the environmental impact of these AI models with people you know. Now that you’re aware, help others understand that there are things we can do to make the world a little better—just by adjusting some of our AI habits.

Greatness is the accumulation of many small acts done well, consistently, over a long period of time. So let’s be great together and see what a sustainable AI world could look like—starting with how we use these amazing tools and technologies.




Originally published in Spanish for Fast Company Mexico:

https://fastcompany.mx/2024/10/17/ia-supervillana-nuclear-christopher-sanchez/

Christopher Sanchez

Professor Christopher Sanchez is internationally recognized technologist, entrepreneur, investor, and advisor. He serves as a Senior Advisor to G20 Governments, top academic institutions, institutional investors, startups, and Fortune 500 companies. He is a columnist for Fast Company Mexico writing on AI, emerging tech, trade, and geopolitics.

He has been featured in WIRED, Forbes, the Wall Street Journal, Business Insider, MIT Sloan, and numerous other publications. In 2024, he was recognized by Forbes as one of the 35 most important people in AI in their annual AI 35 list.

https://www.christophersanchez.ai
Previous
Previous

From Chatbots to Agents: The Evolution of Artificial Intelligence

Next
Next

The Mexican Moment and Artificial Intelligence