The AI Explosion, Explained

The field of AI has been around for decades. So why is it suddenly everywhere, and what lies ahead?

Duke's Larry Carin, Lee Tiedrich and Matt Hirschey discussed what the future holds for AI Jan. 30 at the 2024 Duke Centennial Research & Innovation Week

Panel moderator Daniel Dardani

The demo was part of an AI-themed panel at Duke Research & Innovation Week 2024, held Jan. 29 to Feb. 2 as part of the Duke Centennial celebration.

What we now call AI has been around as a discipline for decades, said moderator and tech transfer expert Daniel Dardani of Duke's Office for Translation and Commercialization and the Office of External Partnerships.

But AI tools like ChatGPT have pushed the technology into public view.

Which raises an obvious question, Dardani said. “What’s different about AI today, and what could come next?”

This isn’t the first time AI has captured the public imagination, said AI and machine learning expert Larry Carin, former chair of electrical and computer engineering and former Vice Provost for Research at Duke.

Over its history, the field has gone through repeated cycles of booms and busts -- periods of remarkable progress and optimism followed by skepticism.

These cycles are known as the “seasons of AI,” Carin said, and “we’re definitely experiencing an AI summer.”

But this moment seems different, Carin added. It’s not just another bubble of inflated expectations. “We’re in a different phase,” Carin said.

Larry Carin, Ph.D., former chair of electrical and computer engineering and former Vice Provost for Research at Duke. York Wilson Photography.

“Most of the focus right now in AI is really on productivity,” said Hirschey, a professor of medicine who also directs Duke’s Center for Computational Thinking.

Imagine a harried physician. In a regular week, many doctors spend nearly half their time entering data and doing deskwork. AI-powered voice assistants that integrate with patients’ electronic health records could make charting easier so that doctors can focus more on patient care.

Or take any office job. Increasingly, companies are integrating “AI assistants” into office apps like Word, Gmail or Excel. These AIs could save users time by automating or offloading tedious tasks such as scheduling meetings or managing email.

“There's something deep about human nature that is looking not for shortcuts, but we’ll say for ‘productivity enhancement opportunities,’” Hirschey said.

And indeed the potential gains are impressive. According to the accounting and auditing firm PricewaterhouseCoopers, the boost in productivity enabled by AI could add $6.6 trillion to the global economy by 2030. That’s more than the GDPs of Canada and the U.K. combined.

“So it's not surprising that it starts here,” Hirschey said. “But I think the most exciting prospect is beyond productivity.”

In education, for example, AI chatbots could serve as students’ personal tutors, clarifying and reviewing concepts based on content from a syllabus or textbook.

Hirshey asked one such bot to “teach me the fundamentals of large language models from an undergrad level,” and the AI came back with an answer almost instantly.

“This is a new way to interact with course material,” Hirschey said.

Such tools aren’t perfect. When he asked the AI tutor for clarification, typing in “I don’t quite understand,” the chatbot choked. It came back with “error analyzing.”

“That just shows you it's a real demo,” he chuckled.

In another demo, Hirschey showed off a custom chatbot he created called “Ghost Scientist.”

It differs from regular ChatGPT in that it was programmed for a specific task, in this case to suggest experiments that could be carried out. After asking the audience to call out fields they were working in, he chose two answers at random and typed: “how would I use light microscopy to study ion channels?”

In an instant the AI came back with a list of ideas, ranked based on feasibility.

Some experts might skim this list and say, “that's totally bogus, or that's not going to work,” Hirschey said. But the point is “we can also build tools that help us come up with ideas,” Hirschey said. “We can use AI to help brainstorm.”

“Right now everyone is thinking about AI as a task completer,” Hirschey said. “But really we should be thinking about AI as an educator, an idea generator, or even as a thought partner.”

AI legal policy expert and Duke executive in residence Lee Tiedrich, J.D. York Wilson Photgraphy.

If you had told an AI-savvy audience even five years ago that such things would soon be possible, Carin said, “I think everybody would have just laughed. No chance.”

To tech insiders, Carin said, “This is unbelievable technology. It is a tremendous accomplishment of sustained research over many decades.”

“But now we have new questions,” Carin added. “Such as: who owns this?”

Tech giants like Google and Microsoft are facing a barrage of lawsuits alleging that scraping data from the web to train their AI systems violates people’s privacy or property rights.

“I've never seen the technology evolve so much quicker than the law,” said Duke alum Lee Tiedrich, who practiced law for 30 years at the forefront of AI and other emerging technologies before returning to Duke as an executive in residence at Pratt.

“Even the patent office and copyright officers are struggling with the implications of AI on traditional intellectual property bedrock concepts,” Dardani said.

“There seems to be some consensus across the aisles that big tech needs to be reined in,” Tiedrich added. Just a few months ago the White House unveiled a sweeping executive order on artificial intelligence, joining a number of countries around the world in calling for stronger guardrails around AI.

“It’s still the Wild West, but policymakers are trying very hard to tame it,” Tiedrich said. The goal is to “capitalize on all the great benefits that artificial intelligence brings, but also make sure that we mitigate the risks.”

One of the clearest lessons from the panel was that we've only just begun to see what's possible with AI.

“It's early,” Hirschey said. “We still don't know exactly how these tools work, what their limitations are, or the full implications of their use.”

Looking ahead, however, AI agents could be embedded inside every app we use. One study suggests that a third of college students are already using ChatGPT for schoolwork.

“There's no question that it’s going to be used,” Carin said.

Dardani brought the discussion to a close with one final quip: “If there’s any portion of today that you forget and want to double back on, don’t worry -- just ask ChatGPT.”

Q&A following the panel "AI: Glimpse of the Future". York Wilson Photography.