Lawyers fined for filing bogus case law created by ChatGPT (power, hardware)

[moguldreamer]

Lawyers fined for filing bogus case law created by ChatGPT

https://www.cbsnews.com/news/chatgpt...s-who-used-ai/

A federal judge on Thursday imposed $5,000 fines on two lawyers and a law firm in an unprecedented instance in which ChatGPT was blamed for their submission of fictitious legal research in an aviation injury claim.

In another example of "AI Hallucinations," ChatGPT cited non-existent judicial opinions that it fabricated out of whole cloth to bolster the AI-generated positions.

This comes on the heal of ChatGPT also fabricating convictions for sexual assault when tasked to uncover dirt on an political opponent.

ChatGPT and its kin are not tasked with accuracy. They are tasked with creating grammatically correct prose.

I can't wait for mistakes that will be made when ChatGPT is employed in customer support.

[SnazzyB]

IMO, $5000 isn't a big enough fine. I think the fine should be so big that it scares law firms from even thinking about using fake legal research.

[Mircea]

I can speak to that.

ChatGPT synthesizes. That's what it does. It has no application in the legal field and the attorneys should have known that so the fines are justified.

Hopefully the ABA bars its use.

I was on a zoom meeting for AI intended for the legal world and it impressed no one. You have to tell it what to do which is the hugest misconception because AI will not replace jobs. You have to babysit AI to make sure it does what you want it to do.

It did generate a Motion for Summary Judgment but it was worded goofy. It pulled too much case law, some of it not relevant and some of it not on point. It had a difficult time crafting a legal argument and it couldn't integrate the facts of the case with the case law in support of the MSJ.

AI might be able to generate a generic Complaint, but if it required specificity, AI would fail and if you have to meet the standard in Iqbal for a fraud/misrepresentation claim, AI could never craft plausible fraud scheme.

And the writing was goofy. An attorney's individual style is reflected through their arguments and AI robs attorney's of their identity the same way MS Word robs you of your identity.

Microsux tries to tell you what words to you and even if you turn off the auto-feature, you still end up with dozens and dozens of blue underlines because you didn't write it the way Microsux wants you to write it.

Anyway, my advice: Whether you're in a civil or criminal case doesn't matter. Get a statement from your attorney in writing signed by your attorney that they are not using AI for anything.

[moguldreamer]

Originally Posted by Mircea

I can speak to that...

...It did generate a Motion for Summary Judgment but it was worded goofy. It pulled too much case law, some of it not relevant and some of it not on point. It had a difficult time crafting a legal argument and it couldn't integrate the facts of the case with the case law in support of the MSJ.

Did it engage in AI Hallucinations - outright fabrication? Have you run across this?

[Mircea]

Originally Posted by moguldreamer

Did it engage in AI Hallucinations - outright fabrication? Have you run across this?

No, but then my experience is limited and this particular product is not designed to synthesize. It pulled the right cases but didn't synthesize them into a single citation. Unless the court decision is formatted in such as way where there is a section marked "Held" or "Holding" it was unable to identify the holding.

It couldn't merge the facts of the case with the citations it pulled.

Law firms use different software and some don't use any software at all. You have to manually input the caption info (court, plaintiff(s), defendant(s), case number, judge, title).

In other words, you can't tell it to go to PCLaw and pull caption data from matter number 23-0607 or go to Clio and pull this matter info.

So, AI will eliminate exactly ZERO jobs in the legal industry.

ML (Machine Learning) is different than AI. But, ML is of no value unless it can access and digest 1,000s of family law cases in a particular jurisdiction and it would have to be able to recognize and understand that other jurisdictions are different.

For example, there are 88 Ohio counties but not all of them have a Domestic Relations Court. Whether they do or do not depends largely on the population and how much tax revenue the county generates. Even those counties that do have a Domestic Relations Court (the actual name in Hamilton and Clermont Counties) other counties call them Family Court (like Stark County).

Then there's the issue of juvenile courts. In some counties they're under the jurisdiction of the Domestic Relations Court or Family Court and in other counties they're under the jurisdiction of the Probate Court.

AI will never know that unless you tell it each and every time you use it. ML might actually learn that over time, but it would blow a gasket trying to find the Domestic Relations Court in a county that doesn't have one.

I'm not saying there's no value to AI or ML, only that it is currently over-hyped and over-sensationalized.

AI and ML might have some utility, but that will be 10-30 years from now.

You could probably use AI to generate sales letters/emails and for anything that is of a mass generic nature, but customizing it for individuals is going to be a chore.

[burdell]

Originally Posted by SnazzyB

IMO, $5000 isn't a big enough fine. I think the fine should be so big that it scares law firms from even thinking about using fake legal research.

IF they knew it was bogus it would seem disbarment would be appropriate.

[AntonioR]

I used to think the whole "AI is going to take over the world, the sky is falling" mantra was an exaggeration. Yet, every month there seems AI reaches new heights, such as this which has never been seen before. Hmm... I still think it's an exaggeration, but now there is a but.

[GearHeadDave]

Originally Posted by AntonioR

I used to think the whole "AI is going to take over the world, the sky is falling" mantra was an exaggeration. Yet, every month there seems AI reaches new heights, such as this which has never been seen before. Hmm... I still think it's an exaggeration, but now there is a but.

"AI" was already taking over the world, long before it got this latest moniker. That process began with the first microchip and the first lines of code.

The speculation about what AI will do is not an exaggeration - it's not even close to the reality that will be coming for all mankind. Look ahead 20 years - the increase in AI capability will roughly follow Moore's law which states that computing capability doubles ever two years; so that in 20 years you are looking at hardware/software intelligence that is hundreds of thousands of times more powerful than it is today.