AI expert tells New Mexico lawmakers to go further with regulation • Source New Mexico
An expert on artificial intelligence told a legislative panel that state lawmakers were right to protect themselves from AI’s potential harmful effects on their election campaigns, but now they need to offer the same protections to everyone else.
Melanie Moses is a University of New Mexico computer science professor and special adviser to the vice president for artificial intelligence research.
At the Science, Technology and Telecommunications Committee’s first meeting between legislative sessions on Wednesday, Moses said lawmakers took a good first step with the deepfake law passed in the 2024 legislative session.
Deepfakes can be used in any type of media that is edited or manipulated to make fake information appear real. Artificial intelligence can be used to easily create sophisticated media that can make a person appear in places they’ve never been, or do and say something that didn’t happen.
The new law in New Mexico will go into effect later this month. It requires political campaigns and candidates to tell the public whenever they use false information generated by artificial intelligence in a campaign ad.
However, Moses said it’s “a small step on what’s going to be a very long journey.”
“You’ve protected yourselves; you’ve protected people who are running for election,” Moses said. “The rest of the population also needs this sort of protection.”
She pointed to deepfakes that create pornography or slander as two examples of real harms. “We can’t let (those) stand, both for the particular individual that is going to suffer, but more broadly, because all of us will learn to distrust reality if these deepfakes proliferate,” Moses told state lawmakers.
“I think we’re already pretty close to never being able to believe what we see, which is a really hard way to run a democracy,” Moses said.
Moses said she thinks the Legislature would best serve New Mexico by considering what to regulate to protect people from algorithms’ potential harms, and where to invest money in artificial intelligence research for the economy and public education.
Transparency and equity
Moses said AI laws passed by other states contain many loopholes, especially around “trade secrets,” where governments and corporations can withhold information about the models they use to make decisions by claiming it is proprietary.
“I think that the state of New Mexico should take the position: if an algorithm is making a choice that affects someone’s freedom, their financial wellbeing, those sorts of highly consequential decisions, there can’t be a loophole that says, ‘This is my trade secret, so I can’t tell you why I’m denying you this loan or keeping you in jail,’” Moses said.
A tale of two Martins: Artificial intelligence takes center stage in Santa Fe
Machine learning models like the AI-driven chatbot ChatGPT, for example, are “not explainable,” Moses said.
“If the model is not transparent and can’t be explained, it shouldn’t be used in this kind of high-stakes decision making in our state,” Moses said. “I think that’s a line we can draw.”
Two committee members, Reps. Debra M. Sariñana (D-Albuquerque) and Christine Chandler (D-Los Alamos), were co-sponsors on a bill in the last session that would have required all state agencies to track their artificial intelligence systems, including whether the system was used to support a “consequential decision,” like determining someone’s public benefits or imposing punishment like jail or prison time.
Moses said she thinks that bill is “on the right track.”
Electricity and water usage
It is difficult to find a place to put the large data centers which host the supercomputers that help create these AI models, Moses said, because “there is no state that will say, ‘We can offer you enough power and enough water to cool this data center that you can put it here.’”
“We just don’t have the power to actually drive these,” Moses said. “It’s expected that 10% of the electricity in the United States will soon go to powering these AI models. In an era of climate change, that is an astonishing — and really not a future we want to go to.”
Chandler asked if she should assume greater demand on electricity and water because of a new supercomputer installed by Los Alamos National Laboratory.
Moses said yes and no. The lab’s machine is smaller than other supercomputers but New Mexico does not have much water, she said.
“I have not seen an assessment of how much water and electricity is in use for these particular machines,” Moses said. “It’s the kind of thing we should know the answer to before, as part of the decision making.”
Chandler asked if lawmakers should consider making some kind of public input requirement for supercomputing projects “to assure that the resources are there, and not they’re not going to negatively infringe on just living in a town.”
Study group
Moses said another important thing discussed in the previous session was some kind of “standing study group” to look at what is coming in the future, what lawmakers have already done, and the impacts of those new laws.
Moses said when there is radically new technology, “we really are terrible at predicting what will happen.” She said that’s why she thinks a study group is necessary to constantly monitor AI laws’ good and bad effects.
“With AI, it is really a complex system; you’re not going to be able to create a law that regulates AI and expect it to hold for years on end,” Moses said. “It really is going to have to be an iterative process.”
