Research & Discovery

Q&A: Is Artificial Intelligence Friend or Foe?

Artificial intelligence has been around for longer than you may think.

Email spam filters and autocorrect systems are considered among the earliest forms of artificial intelligence, or AI. And way back in 1997, IBM’s Deep Blue supercomputer made history, defeated then-reigning world chess champion Garry Kasparov.

The idea of AI dates to a conference at Dartmouth College in 1956, “where a group of famous scholars and mathematicians got together to ‘solve’ the problem of artificial intelligence over the weekend,” said Mona Slone, an assistant professor of data science and media studies at the University of Virginia. “That is often treated as the birth of artificial intelligence as a field with its own name.”

Many are familiar with ChatGPT, a generative artificial intelligence tool released in 2022. It can create talking points for presentations, write songs, brainstorm ideas and even help you plan dinner menus. But how can people use it and other AI tools responsibly? UVA Today turned to Sloane for answers.

Mona Slone is a frequent public speaker and commentator, and has written for The Guardian, MIT Technology Review, The Hill, Nature and other outlets. (Photo by Catarina Heeckt)

Q. ChatGPT is generative AI. What does that mean?

A. Generative AI is a deep learning technique that takes a ton of raw data – think of the internet for example – to learn essentially the logic and structure of human language. Based on that, it can infer what we mean when we ask it a questions.

Q. What does that mean?

A. That means generative AI produces statistically probable outputs when prompted, which is different from understanding meaning. Generative AI has a blueprint of the human language and then, in the specific context of a prompt, can produce an output that is contextually relevant.

Q. Is AI a super powered Google?

A. I would not say that and here is why. Google is about search, and search is information retrieval and, historically, information retrieval systems look for matches between the keywords you put into the search bar and keywords in the corpus that it has access to. So, all the internet. What AI search (does) is it interprets your input as keywords.

Q. Can you give an example?

A. Theoretically, you could say, “Give me the best restaurant in Charlottesville for a Thursday night dinner with my best friends.” AI can go and interpret your search and say, “Well, you said you want Charlottesville, but maybe you want Richmond. And you said Thursday, but maybe you want Wednesday.” So, there is an interpretation that is happening there.

Q. What are some other ways people can use generative AI?

A. Meal planning is one of them. We have folks who use generative AI to summarize text for them. Generative AI is used sometimes to produce images and promotional materials.

‘Inside UVA’ A Podcast Hosted by Jim Ryan

You can use it to plan gardens, write speeches. Students can use it in meaningful ways to help them understand concepts that are discussed in class and ask questions, really using it in a way that helps their learning.

Q. There are skeptics who ask, “How can you trust what AI is telling you? How do you know that it is verified information?” What do you say to that?

A. They’re right.

We absolutely have to understand that we are not dealing with absolute truth here. We have to go back and verify, and that is our responsibility. I worry that we are slipping into a sloppy use of these systems.

Q. Why is AI such an energy hog?

A. It is computationally energy-intensive. Whenever a person interacts with an AI model, this model has to be switched on to run and that takes enormous amounts of energy. Even more so, the creation of these models takes an enormous amount of energy. That is problematic because it has an environmental impact. It is problematic because it can drive up local energy prices for communities.

We don’t really talk about that because we still imagine AI to be this quasi-magical thing that appears wherever you look.

Q. How should people be using AI responsibly?

A. I can give you my personal take on that. I think AI use only happens when it absolutely has to happen. When you can do something quickly with a Google search or any other information retrieval system, you could do that rather than use ChatGPT.

If we are imagining ourselves as sustainability-aware consumers, we could choose a vendor that has these kinds of things in mind and … discloses energy usage of their model. We are really just beginning that conversation.