Please create an account to participate in the Slashdot moderation system

 



Forgot your password?
typodupeerror
×
Government AI

Brazilian City Enacts an Ordinance That Was Secretly Written By ChatGPT 41

An anonymous reader quotes a report from the Associated Press: City lawmakers in Brazil have enacted what appears to be the nation's first legislation written entirely by artificial intelligence -- even if they didn't know it at the time. The experimental ordinance was passed in October in the southern city of Porto Alegre and city councilman Ramiro Rosario revealed this week that it was written by a chatbot, sparking objections and raising questions about the role of artificial intelligence in public policy. Rosario told The Associated Press that he asked OpenAI's chatbot ChatGPT to craft a proposal to prevent the city from charging taxpayers to replace water consumption meters if they are stolen. He then presented it to his 35 peers on the council without making a single change or even letting them know about its unprecedented origin.

"If I had revealed it before, the proposal certainly wouldn't even have been taken to a vote," Rosario told the AP by phone on Thursday. The 36-member council approved it unanimously and the ordinance went into effect on Nov. 23. "It would be unfair to the population to run the risk of the project not being approved simply because it was written by artificial intelligence," he added. [...] Keeping the proposal's origin secret was intentional. Rosario told the AP his objective was not just to resolve a local issue, but also to spark a debate. He said he entered a 49-word prompt into ChatGPT and it returned the full draft proposal within seconds, including justifications.

"I am convinced that ... humanity will experience a new technological revolution," he said. "All the tools we have developed as a civilization can be used for evil and good. That's why we have to show how it can be used for good." And the council president [Hamilton Sossmeier], who initially decried the method, already appears to have been swayed. "I changed my mind," Sossmeier said. "I started to read more in depth and saw that, unfortunately or fortunately, this is going to be a trend."
This discussion has been archived. No new comments can be posted.

Brazilian City Enacts an Ordinance That Was Secretly Written By ChatGPT

Comments Filter:
  • by OrangeTide ( 124937 ) on Friday December 01, 2023 @03:41PM (#64047247) Homepage Journal

    Then why would anyone bother reading it? I swear these glorified Markov-chain text generators are not a good basis for your art, culture, and government. Future generations are going to pick through our ashes and look on this like we look at the Romans drinking leaded wine.

    • How do we look at the Romans drinking lead wine? It wasn't known to be harmful and, given the short lifespans of the time, was a relatively minor danger. We still use it in pipes in parts of the US!
      • Lead poisoning is a longstanding academic debate as to the cause of the decline of the Roman empire. I suspect it's more of an urban legend than strictly a cause of the decline, but the concept should be pretty familiar to laypeople born in the 20th and 21st century.

        • OTOH we KNOW that IQs went up and crime rates went down after we stopped putting lead in gasoline.

      • Old pipes are OK. They form an oxide layer that saves you.

        Cooking in lead pots, drinking from lead cups, eating off lead plates? Not so much.

    • by Ksevio ( 865461 )

      The LLMs like chatGPT are quite different from previous generation Markov models. They're quite capable of creating stuff like laws where it's basically just boilerplate text or expanding on ideas.

    • Don't worry, we can just ask ChatGPT to summarize it and tell us what it says.

  • by Aero77 ( 1242364 ) on Friday December 01, 2023 @03:42PM (#64047249)
    Generating an Ordinance modeled after other typical Ordinances of this type. This has to be one of the most obvious uses of AI, copying other human generated work for the common good. AI has many pitfalls and downsides, but I don't think this is something that we need to complain about.
    • by HiThere ( 15173 )

      Not necessarily object to. This may be a perfectly good measure. But it sounds as if he skimped on quality control, and that's a very necessary step. Sometimes you can skip it, but you never know when.

  • by XaXXon ( 202882 ) <<xaxxon> <at> <gmail.com>> on Friday December 01, 2023 @03:55PM (#64047289) Homepage

    and presumably the people voting read it, right?

    The source is irrelevant, the contents are what matter.

    • by Chris Mattern ( 191822 ) on Friday December 01, 2023 @03:56PM (#64047291)

      "and presumably the people voting read it, right?"

      I like you. You're an optimist.

    • by Junta ( 36770 ) on Friday December 01, 2023 @04:06PM (#64047317)

      While none of the people voting read it, they did paste it into ChatGPT and asked "Can you sum this up in 50 words or less?"

    • This simply isn't true in legislation at least not here in the US. Maybe it is in Brazil, I don't actually know. I'm not a lawyer either but my understanding is that, if a statue is clear courts interpret it using the plain text. But, if there is some ambiguity, the courts resolve the ambiguity by interpreting in a way consistent with the legislative intent. This is done first by looking at the legislation as a whole and, if still not resolvable, as a last resort, legislative history will be examined.
      • by srg33 ( 1095679 )

        Sorry, but this argument fails. Your description of the process is mostly correct. But, the whole legislation is available. Legislative history can be very thin. Some laws are passed with practically no debate (whether or not AI is involved). All legislation SHOULD be reviewed and debated prior to vote, but . . . Congress people and their staffs SHOULD read and verify all bills, but . . .
        XaXXon is basically correct.

      • The LLM didn't invent it, it was directed to write up an idea. It's likely that the 49 word prompt adequately described the idea, but it just wouldn't be "proper" to have such a short, straightforward ordinance. So the LLM can "verbosify" the idea to be consistent with what legal documents are supposed to be.

        The crazy part is not that an LLM could extend the concept to this extent, the real crazy thing is we can't bring ourselves to just deal with the 50 word version through and through.

    • It's probably better than a lot of ordinances written by people.

  • Why can't the law/ordinance just say "The city can't charging taxpayers to replace water consumption meters if they are stolen." In Portuguese of course. Why does it have to be any more complicated than that, sans an identifier number or something?

    This is the problem with our legislative systems.

    • by Draeven ( 166561 )

      The problem with our legislative system is that everyone and their mother will do everything they can to work the system to favor only themselves.

      So now you have to rigorously define what "stolen" means.

      • The problem with our legislative system is that everyone and their mother will do everything they can to work the system to favor only themselves.

        So now you have to rigorously define what "stolen" means.

        I would counter that if the "short version" of the language isn't clear enough, it isn't a good law in the first place.

      • And by whom. Without that, residents could "steal" their own water meters and sell them back to a supplier who would then bill the city to re-install the meters. Also probably need to define what kind of evidence is necessary for something to be considered theft.
    • by kqs ( 1038910 )

      Well, most of the code I write would be much much shorter if I just never added error checking, bounds checking, fallback code, graceful degradation, unit tests, corner-case handling, and a host of other things which separate a good reliable program from an error-ridden piece of garbage.

      Laws are the same way. We'd all love short trivial laws, but experience has shown that short trivial laws have loopholes and special cases that unscrupulous people use for ill effect. And it turns out that there are a lot

      • I wish I had seen your post before I commented. Yours are much better than mine and I would have modded you up instead
    • My guess is that the city councillor picked this topic specifically because this is such a simple case that it could be drafted by ChatGPT and submitted without a single modification. The only way to understand his comment "without modification" is that he's trying to make a point, make himself famous in the news for being the first in having it fully done by AI, so he picked a simple case.

      In my news there are guesses at which of our national politicians use generative AI to partly or fully write proposals

    • What if the taxpayer participated in the theft? Writing legislation has parallels to software. The "happy path" is generally easy and the number of corner/exception cases is daunting.
      • Perhaps there's a point there, but I'm skeptical the LLM extended form of the brief idea would add meaningful guards rather than empty verbosity.

        • I don't think anybody knows but my hypothesis is that the LLM would expand meaningfully if the training data were meaningful. i.e. with current LLMs that have the whole of the internet as their training data, who knows what you might get. But if the training data were actually just legal documents (legislation, court opinions, et cetera), maybe it would do really well.

          Bringing it back to a more slashdot-relevant conversation, although nobody knows, I continue to think that LLMs will be able to produce g

          • by Junta ( 36770 )

            What I've found is that when the model can produce workable code, I would be better off pulling in a maintained library to do the same. If GPT can write it, then it's almost certainly already a library.

            If there isn't a lot of existing code to do *precisely* what you ask, then it will spit out.. something that may hit a couple pieces of the problem that would be part of a computer programming class, but then useless mistakes otherwise.

            If there is existing code to do precisely what you ask... well why isn't

            • I think you are correct in many cases. When writing proprietary software, you often have similar but not identical features and nobody has written a library specific to your app. The next evolution would be for the LLM to be able to recognize that it generated the code using some very similar code and that maybe a library would be a good choice. What we are talking about, at some point, is an AI-driven refactoring. Given the number of non-AI refactoring tools that are useful, there will be some value the
    • Why can't the law/ordinance just say "The city can't charging taxpayers to replace water consumption meters if they are stolen." In Portuguese of course. Why does it have to be any more complicated than that, sans an identifier number or something?

      This is the problem with our legislative systems.

      I just draped a woman’s scarf over mine and haven’t paid a thing. Water utilities hate this one weird trick!

  • All of my /. comments are now written with 100% pure, organic, gluten-free, vegan, non-GMO human repetitive strain injury to ensure our valuable but replaceable human labor is represented with solidarity like not using the self-checkout line. Boycott, divest, and sanction tactics should be applied to AI scabs replacing human labor. Proletariats of the world such and such! Maybe the Unabomber had a valid point but the entirely wrong prescription like Hamas al-Qassam.
  • I mean when lawmakers are flat out admitting they aren't doing the job they are paid to do and instead just offloaded it to a computer, why should they get paid?

  • city councilman Ramiro Rosario revealed this week that it was written by a chatbot, sparking objections and raising questions about the role of artificial intelligence in public policy.

    Make the most of your opportunity, in the future, objections too will be raised by AI.

  • A human can always be trusted to write something that will be misunderstood or vague so as to ensure future abuse.
    AI might actually write something that's clear and legally air-tight.

  • The next generation is going to look back and know this was a good idea, and a perfectly good use for AI
  • Actually being able to have ALL law codified in a way that computers can understand and generate it would be pretty useful. Or, have a computer smart enough to figure it out, and give access to ALL the laws and regulations to it. However this is a law created by a bullshit generator. He wasn't using a logical reasoner. Just yesterday I was beating my head against a configuration thing and asked Claude.ai which gave me an incredibly useful response which unfortunately was based on the existence of a NON-EXIS

  • seems to me you'd get a better dramatic payoff if you told no one but saved evidence to be revealed on your death. Then it might be notable and interesting. A gamble with more stakes, IMO. But does anyone think like that anymore? Humanity bores me I think.

  • The objections would come from those who want to retain power and control, as opposed to those politicians that are in service to their constituents, and want to do the best for them, as opposed to the objections who want to do the best for themselves.

Beware of all enterprises that require new clothes, and not rather a new wearer of clothes. -- Henry David Thoreau

Working...