The amended bill, which next goes to the Senate floor, would criminalize so-called deep-fake pornography and false political messaging. It also would create a new, free "Online Academy" for state residents to learn about the technology.
The new academy would first focus on teaching the basics of the technology, then possibly expand learning to include small business applications and marketing, said state Sen. James Maroney, D-Milford, co-chairman of the General Law Committee who led a statewide task force on AI last year. Certificate programs similar to courses now taught at Connecticut State Community College could also be possible, he said, adding that the General Assembly may have to revisit training on an annual basis as the technology changes so swiftly.
"By helping make our businesses more-efficient, that can help increase their profitability which could potentially lead to tax revenue," Maroney told the committee. The legislation, which was modeled after a similar law in Colorado, could allow AI to help locate health inequities in Connecticut towns and cities, and even train health professionals from nurses, to nurses' aides and EMTs, he said.
As such, there is a definite need for transparency and safety as AI's social and economic effects spread, Maroney told committee members, referring back to the work done by the task force.
"The first thing everyone wants to feel is safe, and I think we need to feel safe in using this," Maroney said. The bill would prohibit certain manipulated images, such as tactics that have been used to create synthetic nude or pornographic photos of people. It would also ban the distribution of deceptive political or election media.
A companion bill aimed at criminalizing deep fake porn was among proposed legislation that was the focus of a public hearing on Wednesday.
Safeguards are also necessary when it comes to discrimination, lawmakers warned.
"A lot of times when we think of AI, we're thinking of ChatGBT or generative AI that's really come on and become more popular, but AI has been around since 1956 when the term was coined," he said. "We use it in lots of ways to make decisions about people. We have seen that sometimes those decisions have disparate impacts. Since we have some biases in our world, it's kind of picking up on those biases without us necessarily knowing it, or realizing it."
There's still a learning curve to negotiate, however.
"We're going to learn as things go on, as time goes on," said Sen. Paul Cicarella of North Haven, a top Republican on the committee. "And we're at that point now where we're starting to really understand what this does, how it works. It changes every day. Just as we think we have our hands around it, it changes. That's why some of this is required and necessary, so we can protect individuals from using it and informing when it is being used."
Cicarella pointed out parts of the bill that would help state agencies take advantage of the technology to make services more efficient for residents.
While it may be possible to make state programs more accessible, Maroney said it shouldn't affect the state spending package lawmakers are now adjusting for the new fiscal year starting July 1. He also noted that Congress seems unlikely to approve AI regulations anytime soon.
"At the end of the day we're going to have to have a more cohesive — down the road — policy," said Rep. David Yaccarino, R-North Haven.
© 2024 the Connecticut Post (Bridgeport, Conn.). Distributed by Tribune Content Agency, LLC.
