In the fall of 2016, the Connecticut Department of Children and Families began using predictive analytics tools that promised to help identify children at immediate risk.
Using more than 20 data points, the tool compared unresolved cases in the Connecticut system to previous welfare cases with poor results. Each child then received a predictive score that flagged some cases that required more immediate intervention.
But even as more states began to implement the tool, some government agencies said the tool missed urgent cases or incorrectly flagged less serious cases. I noticed that it seems to beResearch published in journals child abuse and neglect It was later found that it did not improve the child’s performance. Connecticut and several other states have abandoned tools developed by private companies in Florida. In 2021, five years after the Connecticut Department of Children and Families first used the tool and two years after the state scrapped it, researchers from Yale University will release information about how it works. When asked, it concluded that the agency did not understand the tool at all.
“This is a very big public liability issue,” said Kelsey Eberle, a clinical lecturer at Yale Law School. “Agencies get these tools, they use them, they trust them, but they don’t necessarily understand them. And the public certainly doesn’t understand these tools because because they don’t know it.”
Connecticut is the latest state to pass clear regulations on artificial intelligence and other automated systems, thanks to its tradition of testing tools for children at risk. The bipartisan bill, which passed May 30 and is expected to be signed into law by Democratic Gov. Ned Lamont, calls on state agencies to investigate and evaluate all government systems that use artificial intelligence. , which mandates the establishment of a standing working group to recommend further regulations.
Many states already regulate aspects of these technologies through antidiscrimination, consumer protection, and data privacy laws. However, since 2018, at least 13 states have established commissions dedicated to researching AI, and since 2019, at least seven states have committed to reducing bias, improving transparency, and promoting both government and private sectors. Passed laws aimed at restricting the use of automated systems. .
In 2023 alone, lawmakers in 27 states, plus Washington, D.C. and Puerto Rico, debated more than 80 AI-related bills, according to the National Congress of Legislatures.
Artificial intelligence tools, broadly defined as technologies that can perform complex analytical and problem-solving tasks once performed by humans, are now used to understand what Americans are seeing on social media and how. It frequently determines whether students are admitted to college and whether job candidates are scored in interviews.
According to the IBM Global AI Adoption Index, by 2022 more than a quarter of all U.S. businesses will use some form of AI. In one notable example of the pervasiveness of AI, California’s recent bill regulating technology received comments from a variety of organizations, including grocery industry trade associations and the state nurses union.
But federal law has stalled and regulation has been left to local governments, creating a patchwork of state and local laws.
“For many years, the United States has been very lenient when it comes to technology regulation,” says Darrell M. West, a senior fellow at the Brookings Institution’s Center for Innovation and Innovation and author of a book on artificial intelligence. “But the social climate and policy-making environment have changed as we have seen the pitfalls of lack of regulation, such as spam, phishing and mass surveillance. People want this to be regulated. “
Lawmakers’ interest in regulating technology has surged during this Congress and is likely to increase next year thanks to the proliferation of ChatGPT and other consumer AI tools, said Jake, director of the Telecommunications and Technology Task Force. Morabito said. Conservative US Legislative Exchange Council (ALEC) in favor of deregulation.
“Extreme” Possibilities and Risks
Once the realm of science fiction, artificial intelligence has now permeated nearly every corner of American life. Experts and policy makers often define the term broadly to include systems that mimic human decision-making, problem-solving, and creativity by analyzing large amounts of data.
AI is already powering a range of speech and image recognition tools, search engines, spam filters, digital maps and navigation programs, online advertising and content recommendation systems. Municipalities are using artificial intelligence to identify lead water lines for replacement, speeding emergency response. Machine-learning algorithms introduced in 2018 reduced sepsis deaths at five hospitals in Washington, D.C. and Maryland.
But even as some AI applications yield new and unexpected societal benefits, experts document countless automated systems that yield biased, discriminatory, or inaccurate results. For example, facial recognition services used by law enforcement have repeatedly been found to misidentify people of color more frequently than white people. Amazon has retired its AI recruiting tool after it found it consistently penalized female job seekers.
Mark Hughes, executive director of Justice for All, a Vermont-based racial justice group, said critics described AI bias and error as a “garbage in, garbage out” problem. said there is something to do. Hughes sat on several state Senate committees last year and said lawmakers need to intervene to prevent the biases and systemic racism that often inherently manifest in training data from being perpetuated by automated systems. testified.
“We know that technology, especially something like AI, will always duplicate what already exists,” Hughes said. stateline. “And I intend to reproduce it for mass distribution.”
More recently, the emergence of ChatGPT and other generative AI tools that can create human-like text, realistic images, and other content in response to user prompts has raised new concerns among industry and government officials. increase. Such tools can fire workers, violate consumer privacy, infringe copyright, spread disinformation, and facilitate the creation of content that fuels hate speech and harassment, the policy said. Planners are concerned. A recent Reuters/Ipsos poll found that more than two-thirds of Americans said they were concerned about the negative impact of AI, and three-in-five said AI could threaten civilization. I answered that I am concerned that there is
“I think AI has great potential to revolutionize the way we work and make us more efficient, but it also has potential dangers,” said James, a Democrat and proponent of the state’s AI law. Senator Maloney said. “You just have to be careful when moving forward.”
Maloney said Connecticut’s new AI regulations provide one of the first comprehensive models for tackling automated systems, and the upcoming legislative session will see regulation expand from the state government to the private sector. said he expects
The act creates a new Artificial Intelligence Office in the state executive branch, tasked with developing new standards and policies for the government’s AI systems. By the end of the year, the agency will draw up a list of automated systems used by state agencies to make “critical decisions” like housing and health care, and that they meet certain requirements for transparency and non-discrimination. should be documented.
Maloney said the law is based on recommendations from academics at Yale and other universities, as well as a similar 2021 law in Vermont. The model is likely to surface in other states, Maloney said, with lawmakers in Colorado, Minnesota and Montana now working with Connecticut to develop parallel AI policies. Several states, including Maryland, Massachusetts, Rhode Island and Washington, have introduced similar measures, Maloney said.
In Vermont, the law has already created a new advisory board and state artificial intelligence department. The department’s head, Josiah Reich, found “about a dozen” automated systems in use in state governments in his first annual inventory. These include the Department of Transportation’s computer vision project, which uses AI to assess potholes, and popular antivirus software that detects malware in state computer systems. Neither tool poses a risk of discrimination, Reich said.
But even as emerging technologies improve government services, they may require more vigilance, he added. Raiche recently began experimenting with how state agencies can use generative AI tools such as his ChatGPT to help voters fill out complex forms in different languages. However, in a preliminary internal trial, Raiche found that for sample questions he found that ChatGPT produced better quality answers in German than in Somali.
“There is a lot of work to be done to ensure fairness is maintained,” he said. But automated systems, if run correctly, “can be very helpful in helping people interact with governments.”
A patchwork of regulations
Like Connecticut, Vermont plans to extend AI surveillance to the private sector in the future. Reich said the state will likely achieve that through the Consumer Data Privacy Act, which controls the data sets underlying AI systems and can act as a kind of backdoor to broader regulation. said to be sexual. California, Connecticut, Colorado, Utah and Virginia have also passed comprehensive data privacy laws, while a few jurisdictions have narrower ones aimed at sensitive and risky uses of artificial intelligence. regulations are adopted.
For example, New York City employers using AI systems as part of their hiring process will have to audit those tools for bias and publish the results by early July. Meanwhile, Colorado requires insurers to document the use of automated systems and prove that it does not lead to unfair discrimination.
A new patchwork of state and local laws is plaguing tech companies, and they are beginning to call for federal regulation of AI and automated systems. According to Brookings Institution’s West, most technology companies cannot customize their systems for different cities and states, so in the absence of federal law, many companies would be forced to operate most across geographic locations. That means stricter local regulations will need to be adopted, he said.
This is a situation that many companies want to avoid. In April, representatives from a wide range of business and technology groups lined up to oppose California’s AI bill. The bill would require private companies to monitor AI tools for bias and report the results, or face hefty fines and consumer lawsuits. The bill survived two committee votes in April before being defeated by the Congressional Appropriations Committee.
ALEC’s Morabito said, “Governments should work with industry and not take such a hostile approach.” “Let the market lead here…Many private companies want to do the right thing and build a trustworthy AI ecosystem.”
ALEC proposed an alternative state-based approach to AI regulation. Dubbed a “regulatory sandbox,” the program allows companies to work with state attorneys general’ offices to test emerging technologies that might conflict with state law. Moravito said such a sandbox could encourage innovation while protecting consumers and educating policymakers on the needs of the industry before drafting legislation. Arizona and Utah, as well as the city of Detroit, recently created regulatory sandboxes where companies can conduct AI experiments.
However, these programs do not prevent legislators in those states from pursuing AI regulation. In 2022, a proposed Republican bill would ban AI from violating the “constitutional rights” of Arizona citizens, and the Utah legislature recently convened a task force to consider a potential AI bill.
Yale University’s Eberle said policymakers no longer see AI as a vague or future concern, nor are they waiting for federal action.
“AI is here whether we want it or not,” she added. “It’s become part of our lives now…and lawmakers are just trying to get ahead of it.”
Stateline is part of the States Newsroom, a network of news outlets supported by a coalition of grants and donors as a 501c(3) public charity. Stateline maintains editorial independence. If you have any questions, please contact editor Scott Greenberger. [email protected]. Follow Stateline on Facebook and twitter.
