Pa. GOP lawmakers hear from ChatGTP as they consider laws addressing artificial intelligence

• By Nathan Wilson/ For LancasterOnline

Twelve Republican lawmakers sat in a committee hearing room at the state Capitol in Harrisburg facing an open laptop.

Next to the laptop sat a cardboard name card with a name printed on it – not the name of an expert using video teleconferencing technology to deliver testimony. The text on the card was simply: “ChatGPT.”

“ChatGPT has the floor,” state Rep. Joshua Kail, R-Bear and Washington, said to prompt the silicon witness at the Feb. 7 hearing.

With that, Griffin Caruso, a research assistant for the House Republican Policy Committee, used the laptop keyboard to inform the artificial intelligence program that it was giving testimony before Republican lawmakers and to introduce itself and give a brief explanation of current AI technologies.

Caruso read the generated responses aloud to the committee:

“Ladies and gentlemen of the PA House Republican Policy Committee, thank you for the opportunity to address you today. My name is ChatGPT, an AI language model developed by OpenAI.”

Thus began the first testimony by artificial intelligence to Pennsylvania legislators, delivered by perhaps the best known version of the technology, which was released to the public in November 2022. ChatGPT is a large language model, an AI algorithm that uses the vast amounts of text scanned into its library to answer questions from users.

As artificial intelligence evolves from a tech novelty to a powerful tool, lawmakers across the country are grappling with the revolutionary potential of AI and its implications for everything from hacking to disinformation.

Gov. Josh Shapiro signed an executive order in September creating an AI governing board to shape how commonwealth agencies would use artificial intelligence in their operations. The board, which includes senior officials from the Governor’s Office and other agencies, has no oversight authority over how the private sector uses the technology.

Meanwhile, 14 bills addressing aspects of AI regulation have been introduced in the Legislature since January 2023, but none has come close to being passed.

At the House hearing where ChatGPT appeared, Republican lawmakers asked questions about AI’s potential. ChatGPT responded with organized lists of positives and negatives, only refusing to answer when a lawmaker asked it to provide information about how to commit identity theft.

“It’s remarkable the depth of the conversations we are having with something that doesn’t even exist,” said Rep. Torren Ecker R-Adams and Cumberland, after ChatGPT listed five risks associated with using AI. “It’s incredible where this technology is going, but I have a healthy concern about where this technology is going as well.”

The Republican panel members also heard human testimony from tech industry executives and an academic from Harrisburg University.

In a hearing held a week later, Democratic lawmakers spoke with representatives from the tech industry and educators, as well as medical professionals who are already using AI to assist with patient care.

“I see this as the beginning conversation, the first step for many of us,” said Rep. Chris Pielli, D-Chester, who led the hearing. “This is a daunting task, but this is one that we are going to have to do. I can’t think, historically, of a greater episode facing humankind — perhaps maybe Gutenberg’s press might be up there.”

Scams and misinformation

Lawmakers are particularly concerned about the likelihood that AI tools will be used to commit crimes.

Legislation introduced in January by two Senate Democrats — John Kane of Chester and Delaware, and Jimmy Dillon of Philadelphia — was prompted by the story of a constituent who fell victim to an AI scam.

According to Kane, the constituent received a phone call from a family member asking for a money transfer. The voice sounded authentic, but it was actually an AI impersonation created using audio of the real individual’s voice.

Kane and Dillon have since co-sponsored legislation that would make it a misdemeanor offense to use AI to impersonate an individual without their consent. It would rise to a felony if used to commit fraud.

“Anyone hearing the voice of a family member who is in desperate need for money is going to want to help them out,” Kane said. “People can get manipulated and scammed. How far this is going to go, we really don’t know.”

AI’s talent for impersonation is already being used in the political realm.

Ahead of last month’s New Hampshire primary, Democratic voters in the state received a robocall with AI generated audio of President Joe Biden encouraging Democrats not to vote.

Generative AI’s ability to create realistic images and audio is only expected to improve, adding further pressure on lawmakers to try to respond.

Justin Brody, an assistant professor of computer science at Franklin & Marshall College, told LNP that efforts to regulate AI will likely fuel a high-tech game of cat and mouse.

“As people develop methodologies for detecting AI, the people who are generating these are going to use that data to improve their models,” said Brody. “I think it becomes a cat-and-mouse game, and I don’t know for how long we’ll be able to detect these.”

In response to the New Hampshire primary robocall, a bipartisan group of Pennsylvania state senators released a co-sponsorship memo in January seeking to prohibit the use of AI-generated impersonations in political campaigns.

AI Registration

A key challenge to AI regulation is the decentralized nature of the industry.

Companies like Meta, which owns Facebook and Instagram, are taking an open source approach, allowing other companies and hobbyists to use the companies’ core AI algorithms to create their own custom versions.

With so many creators working with AI, lawmakers and regulators can’t easily track them or create legislation that would apply in every case.

“I’m going to have to lean toward a no,” Brody said when asked if comprehensive regulation of AI was possible. “I can download something on my computer right now that can make it possible for me to generate a fake image of a person very quickly. I’m not going to say it’s pointless to have laws on the books, because if someone does something unethical and you can trace it, I think there should be some consequences for it.”

Several state lawmakers have proposed creating an AI registry to keep a record of businesses that are developing and operating AI software within the commonwealth.

In 2023, Rep. Robert Mercuri, R-Allegheny, who attended the Republican Policy Committee hearing, proposed a registry that would include the name and physical address of businesses using AI, including any IP addresses, the type of computer code they used and the intent of their software.

Mercuri views a comprehensive registry as the first step in regulating the industry as a whole. “If we don’t know what’s out there, it’s hard to understand, let alone guide this issue as it evolves,” he said.

There is bipartisan support for the creation of a state registry, with House Democratic Reps. Pielli and Robert Merski, of Erie, proposing similar legislation in June.

‘Our laws have to keep up’

The AI industry is expected to rapidly expand in the next few years. Bloomberg Intelligence estimates that the generative AI market could be worth $1.3 trillion by 2032. With the expected growth and the ease of access to AI tools, lawmakers have made regulation a priority for 2024.

“I want to be very clear, the ideal time to have had these standards was probably before now, with the rapid development and adoption of this in the public world,” said House GOP Leader Bryan Cutler, of Drumore Township. “It really is impressive how quickly it’s been adopted.”

Last month, Cutler announced the creation of an Artificial Intelligence Opportunity Task Force, composed of five House Republicans tasked with studying the issue. The House’s top Republican said he wants the task force to tackle a variety of issues involving AI and keep pace with the technology’s rapid growth.

“I think it’s going to be a pretty long effort, because I think AI is here to stay,” Cutler said. “But I also think it’s going to continue to evolve as we go forward. Our laws have to keep up.”