AI As a Force for Good: IEEE Members ShareInsights at CES

January 15, 2025

Artificial intelligence is both powerful and mesmerizing. It can automate mundane tasks that most of us would rather avoid and create powerful images that captivate the imagination.

But those capabilities come with a cost. Training and deploying AI models require vast computational power, leading to significant energy consumption and a growing carbon footprint. Companies looking to deploy AI often cannot find candidates with relevant expertise. And AI brings with it a host of other ethical concerns, including the rise of deepfakes, the value of intellectual property and the trustworthiness of AI systems.

The conflict between AI’s costs and its promise was front and center at a panel entitled “Can AI Deliver Technology for Good?” hosted by IEEE Spectrum Editorial Director Glenn Zorpette at the 2025 Consumer Electronics Show (CES) in Las Vegas. The panelists included IEEE Senior Members Hasala Dharmawardena and Nita Patel, alongside IEEE Members Mark Riedl and Gloria Washington.

You can watch the full discussion here. Check out some of the highlights:

How Large Language Models Work

You may have heard people say that large language models are simply systems that are very good at predicting the next word in a sentence. When you hear the word “blue” you might think of the word “berry.” But if you hear the phrase “deep blue” you might think the next word is “sea.” Large language models use context to predict the next word in a phrase or a sentence based on massive datasets, and lots of math. The explanation paved the way for a discussion about responsible use of AI, and why generative AI systems can provide incorrect answers.

“Every word that goes into a language model is a clue as to what’s going to come next,” Riedl said. “And with just three words, you can come up with a pretty good idea of what I’m going to say. Now if you have a hundred words or a thousand words, the word that’s going to come next becomes kind of self-evident. A language model tries to find the patterns in all of these words, all the clues, and how they might relate to what’s going to come next.”

How Much Energy Do AI Systems Use?

The more complex the large language model, the more energy it takes to generate text or images. U.S. energy consumption has remained relatively stable over the past two decades, despite population and economic activity increases. The panelists explained why the rise of artificial intelligence is expected to increase power consumption.

Emerging Use Cases

AI models are great at answering questions. They can even plan vacations for you, but there are limitations to their capabilities. While they can write out a plan, they can’t book a hotel room or choose a flight for that vacation. The next step for AI will likely be AI agents capable of writing out a plan and putting it into motion. Agentic AI will also have a big impact on fields like healthcare, manufacturing, marketing and cybersecurity.

“For me, AI is about enabling humans to do more,” Patel said. “The idea of a calculator was very simple, but that enabled people to do a lot more. AI is like that for programming. You can say you want something to happen and then it’s created.”

Can AI Deliver Technology For Good?

The answer, of course, is yes. However, panelists pointed out that AI is a dual-use technology, one that the user shapes. Ultimately, it’s about striking a balance between innovative development and the need for human oversight and thoughtful policies to guide its development and use.

“Everything is wonderful with these technologies, but oftentimes, underrepresented and marginalized groups are affected in ways we don’t anticipate,” said Washington. “We need to truly think through fairness, accountability and transparency when we develop and deploy these tools.”