News for hackers. Stuff that matters.
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by cowmix on Friday June 09, 2023 @10:12AM (#36258090)
A close friend of mine recently found himself in the middle of an unexpected tussle with his former water utility provider. At the heart of the contention were fees that the provider insisted were my friend's responsibility, fees that, in truth, had been accrued by a subsequent tenant. A strict interpretation of city policy, combined with a few unusual loopholes, provided the city the justification to insist that he owed them thousands of dollars.

Navigating through this predicament, we engaged ChatGPT in the majority of our communications with the city. It not only streamlined our dialogue but also served as an indispensable resource for us. In the midst of this process, ChatGPT uncovered state utilities codes that appeared to bolster my friend's position. These were not codes we managed to discover through our initial Google searches. They were revealed thanks to the in-depth analysis capability of ChatGPT (of all the communications).

However, it's worth noting that even ChatGPT grappled with interpreting the state codes correctly. There were moments of back-and-forth, as the AI evaluated whether the codes could indeed help our case or if we were misinterpreting their legal implications. This dynamic examination, reflecting the inherent complexity of the issue, only served to enhance the value of the insights that ChatGPT provided.

Ultimately, powered by the findings and the intelligent assistance of ChatGPT, we managed to successfully challenge the charges. The city officials, who had been resolute in their claims, were forced to reconsider and ultimately dropped the fees. Without the involvement of ChatGPT, it's hard to say if we would have been successful in this dispute.

Re:Lawyers blame ChatGPT for tricking them into citing bogus... by COGlory on Friday June 09, 2023 @10:26AM (#36258294)
This comment is settinh of my ChatGPT alarm bells. Something about the sentence structure and adjective usage.
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by weego on Friday June 09, 2023 @10:30AM (#36258361)
Their recent comments have a few overly glowing and slightly unrealistic scenarios using ChatGPT.
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by jkea on Friday June 09, 2023 @10:39AM (#36258506)
Last two paragraphs starting with "however" and "ultimately" SCREAM chatGPT generated to me
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by cowmix on Friday June 09, 2023 @10:44AM (#36258607)
As least I didn't add "I hope this response finds you well."
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by cowmix on Friday June 09, 2023 @10:42AM (#36258569)
Heh, your spidey senses are correct. I actually wrote the response and then ran it though ChatGPT for clean up. I do agree, it actually reads weird over my original draft.
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by VoodooJuJu on Friday June 09, 2023 @10:34AM (#36258424)
This comment is fraudulent.

Not only is the style quite brazenly ChatGPT, this part is a huge red flag:

>ChatGPT uncovered state utilities codes...

ChatGPT cannot uncover anything, especially the municipality-specific corpus of codes that the comment ostensibly claims to have access to.

Re:Lawyers blame ChatGPT for tricking them into citing bogus... by cowmix on Friday June 09, 2023 @10:40AM (#36258520)
OP here. That's exactly what happened.
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by garblegarble on Friday June 09, 2023 @10:45AM (#36258626)
>ChatGPT cannot uncover anything, especially the municipality-specific corpus of codes that the comment ostensibly claims to have access to.

It's possible the training set includes those codes (or a citation of those codes). It's possible that they used GPT4 with Browsing, or that it gave them the right terminology to search for & they then pasted in sections of the code and asked it to work off that.

It's also entirely possible ChatGPT hallucinated these utility code sections to support its case and the people working for the utility didn't call their bluff... which is essentially exactly what this story is about, except they got caught out because ChatGPT's bluff was called...

Re:Lawyers blame ChatGPT for tricking them into citing bogus... by pyuser583 on Friday June 09, 2023 @10:28AM (#36258322)
Did you write this comment with ChatGPT's help? It's a helpful and well written post. I'm not trying to insult it. But I'm curious. The text seems to say "AI", and I'm wondering if I'm seeing things not there.
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by cowmix on Friday June 09, 2023 @10:45AM (#36258624)
Yes, in the spirt of talking about ChatGPT I did run my response through it. It is probably the worst thing it has ever produced for me. I ring this up to ChatGPT's "declining quality". :)
Re:Lawyers blame ChatGPT for tricking them into citing bogus... by butler14 on Friday June 09, 2023 @10:16AM (#36258143)
Great anecdote, thanks for sharing. Forcing ChatGPT to go back and sense check its overarching initial interpretation is such a good way of doing things in these types of uses cases.