AI can be used to provide better services for New Yorkers

The Queens Assembly member made history when he became the first lawmaker to use a generative AI program to research and write a bill that he introduced in July. Critics were aghast but Clyde Vanel, who chairs the Assembly Subcommittee on Internet and New Technology believes the exercise demonstrates the limitless potential of the software to identify shortfalls in state rulemaking. He introduced another bill – this one he wrote himself – calling for the governor to create an AI task force to determine how the state should regulate the technology. This interview has been edited for length and clarity.

Can you tell us more about why you decided to use Auto-GPT to write state legislation?

One of the things we have to do when it comes to new technology is to wrestle with it, try it. Especially if I’m in the position to influence policy, I should truly understand the capabilities, the benefits and the downfalls of these technologies. In light of that I said, “Hey why don’t I see if it can do my work.” Not just a part of making policy but trying to figure out how it will affect workers, the creative economy, white-collar jobs and blue-collar jobs. What better way to understand it than to use it. So that’s what I did.

What’s interesting is these types of artificial intelligence that are available now. These large language models have only been available to the mass public for less than a year now, and we are still trying to understand what it will mean for many aspects of society. So how can I not use it? My question is, how can we not try to figure out the possibilities of these policies? And I got some backlash for that.

What was the reaction to doing this?

Some people, even some of my colleagues, said, “We’re elected, not computer programs. We should be writing bills, not computer programs.” Some complained about the substance of the bill. 

The computer is a technology. I don’t write bills without a word processor. And what’s important when it comes to generative AI is that these technologies are human-centric, which is what I showed going through the process of using generative AI to write my bill. It shows the level of misunderstanding that’s there on how and if we use technology.

How did it know to write a bill requiring landlords to provide a lease to renters? Did you train or prompt it in any way?

I wanted to see how large the language models are and how broad of a prompt I could put in to see what comes out. So the prompt that I gave initially found a gap in New York state law and found something that should be a law. And what’s interesting with the exercise is that it found something that is not explicit in any data or laws or articles. It was able to have that level of thinking to say this is happening now, and it should be this. I was amazed by that. It’s not the Magna Carta. It’s not the most seminal bill but I was amazed that the technology was able to identify a gap in the law not on the internet or anywhere.

Have you tried it again?

I didn’t do it again. I’m going to use the tools more, but I haven’t done it again yet.

Did the bill pass?

Some of the other reactions were substantive reactions, some of the people in housing were talking about the substance of the bill. We don’t have a session until January. The bill was just introduced.

Did anyone make an argument that because a computer wrote it, that it has no standing?

What I made sure to do was I made sure to credit that this bill was co-written by artificial intelligence. We have to figure out how to do it with credit and citation so people know when artificial intelligence is used in conjunction with humans to create content. I’m making very certain that I cited that this bill was assisted by a human to write and research the bill.

We have to have a conversation about bill drafting and authorship in the world. We had initial conversations about the bill drafting process. I have different educational institutions that take different standards on these programs. Some embrace it and say, “Yes you can cite it.” And others are on the different end of the spectrum about the use of these technologies.

What’s the difference between ChatGPT versus advocates that spoon-feed bills to multiple legislatures?

I personally don’t take bills that say here’s this bill take this carrier. For people who feel OK with bills drafted by outside groups, the logic moves forward that they shouldn’t have an argument against ChatGPT. At the end of the day, I used Auto-GPT to help write a bill, that doesn’t mean the bill doesn’t go through the whole process. Even if someone gives you a bill from Mars, it still has to pass committee in both houses, get on the floor, get voted on in the Assembly and Senate, and the governor has to sign it. Using it to write a bill, it still has to become a law.

What’s the latest on your legislation creating an AI task force in New York? Has it met yet? What do you expect to come out of it?

The bill is still on the desk of the governor and we’re hoping the governor signs it. She has until Dec. 31 to sign it. There are many people in the industry and a lot of discussion in the media about the bill. The importance of providing a task force is to have stakeholders that includes industry, academia, policymakers and futurist ethicists figure out the proper guardrails when it comes to this technology. There are many implications. There are societal implications, ethical implications, workforce implications, productivity implications to make sure we use the technologies for all the benefits that it has, have an environment that builds and supports growth and innovation but also minimizes potential harms. 

What are some other ways that the state government could utilize AI in coming years?

The state government has and is using a lot of artificial intelligence. There are many different levels of intelligence. Everyone is very familiar and everyone is pissed at the government right now for an AI technology we have been using for the past five years to address speeding and red lights. So when you speed and get a red light ticket, that’s AI that’s doing that. We don’t have toll booth operators anymore. And that has implications also. We have how New York City is about to implement a program to reduce noise. The point is there are going to be noise detection cameras that can identify loud vehicles or construction sites that emit a certain level of noise that can get ticketed. We’re using more and more of these technologies to provide better service for New Yorkers.