How to Use ChatGPT to Ruin Your Legal Career
ฝัง
- เผยแพร่เมื่อ 28 ก.ย. 2024
- Maybe A.I. isn't ready for primetime. 📌 Check out legaleagle.lin... for free career advice that will change the world from 80,000 Hours.
⚖️ Need a lawyer? Get a FREE consultation with my firm! legaleagle.lin...
⚖️⚖️⚖️ Do you need a great lawyer? I can help! legaleagle.lin...
Welcome back to LegalEagle. The most avian legal analysis on the internets.
🚀 Watch my next video early & ad-free on Nebula! legaleagle.lin...
👔 Suits by Indochino! legaleagle.lin...
GOT A VIDEO IDEA? TELL ME!
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
Send me an email: devin@legaleagle.show
MY COURSES
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
Interested in LAW SCHOOL? Get my guide to law school! legaleagle.lin...
Need help with COPYRIGHT? I built a course just for you! legaleagle.lin...
SOCIAL MEDIA & DISCUSSIONS
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
Twitter: legaleagle.lin...
Facebook: legaleagle.lin...
Tik Tok: legaleagle.lin...
Instagram: legaleagle.lin...
Reddit: legaleagle.lin...
Podcast: legaleagle.lin...
OnlyFans legaleagle.lin...
Patreon legaleagle.lin...
BUSINESS INQUIRIES
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
Please email my agent & manager at legaleagle@standard.tv
LEGAL-ISH DISCLAIMER
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
Sorry, occupational hazard: This is not legal advice, nor can I give you legal advice. I AM NOT YOUR LAWYER. Sorry! Everything here is for informational purposes only and not for the purpose of providing legal advice. You should contact your attorney to obtain advice with respect to any particular issue or problem. Nothing here should be construed to form an attorney-client relationship. Also, some of the links in this post may be affiliate links, meaning, at no cost to you, I will earn a small commission if you click through and make a purchase. But if you click, it really helps me make more of these videos! All non-licensed clips used for fair use commentary, criticism, and educational purposes. See Hosseinzadeh v. Klein, 276 F.Supp.3d 34 (S.D.N.Y. 2017); Equals Three, LLC v. Jukin Media, Inc., 139 F. Supp. 3d 1094 (C.D. Cal. 2015).
Special thanks:
Stock video and imagery provided by Getty Images and AP Archives
Music provided by Epidemic Sound
Short links by pixelme.me (pxle.me/eagle)
Maps provided by MapTiler/Geolayers
⚖ Was I too harsh on these guys?
📌 Check out legaleagle.link/80000 for a free career guide from 80,000 Hours!
You're always honest and telling it like it is and that's why we love You!😊😊❤❤❤❤
Im early somehow
Please Do A JFK 1991 FILM REVIEW on it's LAW ACCRUCY PLEASE PLEASE PLEASE!!!!!!!!!!!!!!!!!
You were perfect as usual. Adore your channel. Thank you for bringing laughter to us in these stressful times
Confess, you had a moment where you would have liked to just beat these two knuckleheads around the courtroom with the Federal Reporter.
This will be used as reference in law schools for decades to come. Ethics professors have just gained hours of material for presentations.
2023 edition textbooks are gonna go insane over this one xd
The lawyers will finally make their mark on history! 😅😂
I once read an ethics board case about a lawyer who got into a brawl with a judge and a court reporter. He got disbarred.
Id say they have about 29mins
Or alternatively, you can invent your own references!
Imagine paying a lawyer thousands of dollars and they use ChatGPT. I'd sue them in addition to the original lawsuit to get my money back.
I would bring these lawyers right through their Bar discipline to get them disbarred ASAP!
Word
Plaintiffs' lawyers are paid if they win, so there wouldn't have been money given to him.
I mean, would you trust yet another lawyer to handle yet another case after these guys did this? Although, if they defend themselves, it may be an easy case.
@@Tomas81623 I would but only because I'd know the idiots I hired the first time have just made sure no one else is stupid enough to try what they did especially not with the same client.
Honestly, even if ChatGPT didn't exist, it really seems like these lawyers would've still done something stupid and incompetent that would've gotten them sanctioned
😂 they didn't even check the source 😭 rookie mistake.
ChatGPT clearly states it can make stuff up.
Schwartz explained he used ChatGPT because he thought it was a search engine and made several references to Google. If only it was a real search engine like he apparently usually uses he could be certain it would only say the truth ;)
@@ericmollison2760 I see what you did there ;)
Tbh if the claim of the lawyers working together since 1996 is true they've been handling it for a good while, this may have been a slip-up by the elderly
@@sownheard He says _he_ did try to check, but couldn't find it and assumed it was just something Google couldn't find and assumed ChatGPT must have given him a summary.
As a paralegal, this whole case got under my skin in the worst way. From the unverified citations, to the fact that he didn't know what the Federal Register is, to lying to the judge. If I did even one of the things they did on this case, I would throw myself at the mercy of my boss, because there's no way in hell I would even let him sign something that wasn't perfect, I sure as shit wouldn't file it.
I just cannot imagine the embarrassment. I mean how do you even survive the level of embarrassment from using Chat GPT to write your documents and it getting everything wrong lol
Maybe this Schwartz guy is an imposter?
Best part was about F.3d
It's not a department it's a book
Asking Chat GPT to validate its own text is like asking a child if they're lying. What do you expect?
That's seriously the best bit, "are you sure this is all true?" "of course! check anywhere!"
And then they DIDN'T CHECK. Because how could anything on the internet be false?
The source is literally "I made it up"
@@genericname2747source: trust me, bro
Honestly this is particularly bizarre. If they had unquestioning faith in AI and didn't think they needed to validate, well that's bad but I can understand the train of thought. So imagine if one of them called an expert testimony, he sounded good and decided that didn't need to be validated. But maybe the so called expert seems a bit shady or his documents didn't seem to be in order. If you decided to validate that expert, would you ask _himself_ about his work?
This could be said for literally every human. It is extremely bad argument against AI. The person creating the fact can't be the one validating it. That's exactly why there is something called "peer reviewed" in academics.
I'm a medical student and one day the residents and I used ChatGPT for fun. I cannot even articulate how bad it is at medicine. So many random diagnoses and blatant wrong information. I'm not surprised the same is true for law
Not surprised. I don't know what data it was trained on, since I'm not in the field, but it does not appear to have been fed research.
@@rickallen9099Why are you copy pasting this everywhere?
@@chickensalad3535it's a bot
@@chickensalad3535dudes trying to look good for our inevitable AI overlords
@@rickallen9099yes but it ain’t here for like at least 5-10 years
The fact that ChatGPT has warnings about it not being a source of legal advice is the most damning evidence that these lawyers did not read through what they presented to the court. Perhaps if they had been more observant, they would have followed ChatGPT's advice to "consult a qualified attorney".
I use ChatGPT as a tool to narrow stuff down, basically to find out what I should google, but I know to ALWAYS CHECK EVERYTHING. And if my question ever gets too specific, it always states: 'I'm an AI model, I'm not qualified to advise on this, ask a professional. Seriously, I can't believe they'd thought they'd get away with this...
My immediate first thought is a pretty common set of phrases that internet comments use: "IANAL", "You'd have to check with a lawyer", "Get a lawyer to check this", "This is not legal advice.".
You know, the type of language ChatGPT probably was trained on, and probably had in its results somewhere.
@@ZT1ST Possible, but I think this response might have been implemented intentionally, for the same reason that all thise phrases are common in the first place. Kind of like how there are certain topics GPT will avoid (unless asked very nicely)
@@valdonchev7296 ChatGPT is specifically programmed to warn people that they shouldn't use it as replacement for proffessional advice.
their warning about not able to produce reliable code has never stopped my students from trying to use it... then fail the course. human ability to selective filtering the text is just...
I finally have confirmation if the background is a greenscreen. Seeing him pull a book from behind him made me happy
Everybody's talking about ChatGPT but this tiny little nugget was the most fascinating part of the whole thing. Also the car alarm sirens after he yeets the book into the background going on for several more seconds while he's talking made me laugh.
When he grabbed that book it broke my entire brain. Now I want to know what all of the books are.
"These books behind me don't just make the office look good, they're filled with useful legal tidbits just like that!" -- Lionel Hutz, attorney* at law
@@silveryin4341They look like reporters (the books of case law he describes around 10:56 ).
@@typacsksome of those books are from the 70s
As a Machine Learning Engineer, seeing Devin explain Chatbots better than 99% of the people in the world who think it's magic or something made me tear up
It's because he's smart and he and his team do their research. That's why he's in The Bigs. P.S. Congrats on being a Machine Learning Engineer, that's amazing! Please help keep us safe from them? Or at least keep it obvious when someone is being an idiot when they use it. Thanks, Your Friendly Content Writer and IT Specialist -
@@jooleebilly Thanks 😊
While Julie made that really nice comment, I just have to say that at first I read your name as Brazil.
He understands it better than these two lawyers did.
As a hobbyist programmer I knew where this was going from the very start, I use ChatGPT to help me learn and write code, I ask it how to perform a specific action in Python and it tells me the answer, but I am always double checking it just to make sure it's not bullshitting me, I simply do not trust it since I know it's just predicting text. I this is one where it is very good but I still am completely suspicious of it since I am very aware of the chatbots habit of making things up.
It’s because he is a very good lawyer that does his research and doesn’t make up citations.
Imagine calling up your lawyer to see how the case is going and finding out he's now in bigger legal trouble than you ever were.
That would be my 13th reason 😩 legal stuff is already so stressful, the costs are ridiculous, so finding out my attorney went and caught a case would be brutal 🤣
@@henotic.essence these would be no-win-no-fee lawyers for sure. Real money buys real lawyers
Tbf… a judge might go lenient on you if it turns out your lawyers doing this. Bigger fish ya know.
It happens
@@jackryan444 If you are a defendant (and lose), you may get a mistrial out of your lawyers being... Incompetent. If you are a plaintiff, you are probably SOL.
The realization that Devin is actually sitting in a library in all his recordings and isn't just using a green screen was by far the biggest plot twist in this video
Edit: why are people arguing about whether or not it was real or edited
why would he go through all that effort getting a book that looked identical to one in his green screen if that was what he was using
And he waited… Not the first, or the second time he mentions case books, but the *Third*. The storytelling in those videos…
It's a green screen.
@swilsonmc he picked up the book bruh, off the bookshelf behind him
I looked at it again and you're right.
Glad I am not alone in this. I almost jumped when he pulled out the book.
I’ve never worked directly with a judge, but I’m going to guess that making a judge research several cases that you refuse to research yourself (not to mention the AI crap) is going to make them very very angry.
Making a judge do work you should have done is like doing the same to anyone but judge has many ways to get back at the person and yeah, it does make them mad.
It's not just that CGPT *can* make stuff up, it's that that's *all* it's designed to do. It's a predictive text algorithm. It looks at its data set and feeds you the highest match for what you're asking, and literally nothing else. It looks at the sort of data that goes in a particular slot, fills that slot with data, and presents it to you. It can't lie to you because it also can't tell you the truth, it just puts words together in an algorithmic order.
Chat GPT is trained to generate text which humans see as looking real. That´s it. There´s no implementation of truthfulness in it´s training, at least not originally.
it's truly mind boggling how many people don't understand the basics of how these models work. "It'S LyInG!!" no mate, the predictive language model doesn't have an intention, it's just stringing words together based on an algorithm...
@@ApexJanitor It can't lie, because it can't think or have intent. Nobody fully understands how these models produce their results, but they do understand the kinds of things that are happening and what its limitations are.
@@ApexJanitor there's a difference between not fully understanding something and having no idea what's going on. I don't think this model is close enough to sentient to be able to "lie" in the moral sense or "want" anything (though it certainly does a good job passing the Turing test, so I can understand the confusion). It's utility function is essentially a fill in the blank algorithm, so of course if you ask it subjective questions, as the idiot lawyer did, it's going to seem to lie.
also what's with the tone of your message? Seems kinda hostile, and the "Hahaha"'s make me feel like The Joker has had a hand in writing this, why not LOL?
@@ApexJanitor I see what you're driving at, but the fact that a neural network of this scale is not comprehensible does not mean that we don't know what it is doing. It's predicting words, nothing more and nothing less. It's not some new and unfathomable way of thinking and responding to the world, it's just mimicking human language (and not very well, at that). You wrote "... it lies if it wants" but that assumes some sort of mind that "wants". ChatGPT and its ilk don't have minds.
Doing this in Federal court was bold (or just plain stupid.) The rules and standards are SO much stricter in Federal court!
I pick "stupid."
Boldly stupid
@@moehoward01 legit
I think lazy is also a valid option.
@@caseyhengstebeck1893 Well, how about all 3?
I'm not a lawyer, but I used to work with the local government with some quasi-judicial hearings where some appellants would retain lawyers to argue for them. One of the funniest cases I had dealing with lawyers, the lawyer quoted a particular case in a written brief which was old enough that it wasn't in the legal databases and he didn't have the full case to provide for review. I walked down to my local library, grabbed the book with the decision, and actually read the decision. The lawyer was then surprised when I forwarded the scanned copy of the case on to him, and I had to point out that it would appear the quote was out of context, and that the decision actually supported the Crown's position. The appeal was then abandoned shortly thereafter.
Begs the question though, how did he find said case? Also, clearly a number of lawyers are not reading the cases they cite, very concerning.
@@jeanmoke1 The original decision was probably cited in a later decision or a secondary source.
That is a legitimate way to do legal research, but, as noted, it is necessary to actually _read_ a decision before citing it.
I did legal research for government lawyers for more than a decade. I would summarize the salient case law and provide excerpts as applicable, but I always attached the full text of the decisions as well. I know that some (but not all) of the lawyers carefully reviewed my work.
@@jeanmoke1
Good lawyers can argue a ruling to make it appear that it supports their client. 🫡
On a list of things that never happened
I’m not a lawyer, but I think a judge’s order that repeats the word “bogus” three times in one sentence in response to your legal filing is probably not good.
The best way I have heard ChatGPT described is "ChatGPT knows what a correct answer looks like." At a surface level, it looks like a legitimate answer until you dive into the details in this case.
My understanding is Chatgpt will give you the answer YOU are looking for. That's what it did for these guys.
Props for the judge for keeping calm while asking these clearly mental lawyers confirmation and not just bonk them in the head with the case book he didn't know about
As a judge, you're supposed to bonk them with the gavel.
@@silentdrew7636I guess "throwing the book at them" was never literal, huh?
I was unsure why judges are treated with some kind of reverence in lawyer circles until I've seen/heard some of their interactions and opinions.
They sure are very composed, tactful and professional, yet absolutely brutal when it comes to scathing remarks.
@@LodinnIt feels like the judge was more dumbfounded than anything. I mean, the responses were so idiotic it makes you wonder how he even passed the bar.
@@warlockd Not sure I agree - by the time they've produced these made-up cases using ChatGPT, the damage was already done. Coming clean was probably the least dumb decision overall in that situation.
...granted, the F.3d moment sounds like a really, really bad knowledge gap, but IANAL. The rest didn't particularly stand out to me, they were pretty screwed by then already anyway.
One thing I love about legal drama like this is how passive-aggressive everything needs to be as it must be kept professional. A judge isn't gonna erupt on someone but if they make a motion to politely ask what you were thinking, you know you're in one heck of a mess.
@@cat-le1hf Ah yes the trial of Chicago seven.
You should see British parliamentary debates. There are strict rules of conduct which dictate how to address people and forbid, among other things, accusing another MP of lying. Even if they are blatantly speaking utter falsehoods, it's forbidden to accuse them of it - because MPs, being the highest and most honourable of society, are surely above such things and it would be an insult to the institution to so much as suggest the possibility of deception. This has lead to a lot of passive-aggressive implications. An MP can't accuse another of intentional lying, so they will instead suggest "The right honorable gentleman appears to be mistaken' giving the most respectful and formal of words while making it clear in their tone that the intended meaning is more 'liar liar pants on fire.'
as an engineer, “if your name is on it, you’re responsible for it” is a HUGE concept. there’s a lot of red tape in working for companies who deal with government contracts, and a lot of specific record-keeping programs you have to use. it’s important for process cycle tracking, but if you’re actually on the development/build side, it can seem pretty tedious. typically you need to be trained on these softwares, so it isn’t uncommon for only one or two people on your team to actually have the authorization to use them. instead of training everyone else, typically that person’s name is just put as the RE (responsible engineer) and then they’re the one who has to sign off on it. for my current program, that ends up being me a lot of the time. in most cases, it isn’t a problem to just go in and sign off on something, seeing as there’s an entire team of people who need to approve before it gets to you. but there’s always the chance that everyone in the upline may also have the same perspective, and my failure to thoroughly review a document before signing off could make or break a multimillion dollar defense contract. and even if it wasn’t even my design so any failures weren’t technically my fault, guess what? if my name on it, I’m the one who has to deal with the fallout. the abundance of approvals and review stages may seem overbearing and unnecessary at times, but that’s how we avoid catastrophic engineering disasters like we’ve seen so many times before. those checks and balances are there for a reason, and if your name is on it, you BETTER have taken the time to complete your check !!
Computer engineer here, it is very smart for you to assume that a screw-up could still slip through the cracks because it absolutely can. I know because I was once responsible for one. Back when I was just moved up to lead developer, a software my team developed and tested hard-crashed while demoing it to management. As it turns out, one of the new guys submitted his component of the software he worked on without verifying that it works. Since I was new to leading a dev team, I unfortunately just assumed that he verified it so we went ahead and put it together with the rest of the software and it passed our tests. That component dealt with installing the software, so when we tried to demo it to management on a computer that used a different OS, it wasn't properly installed. I got in A LOT of trouble for this (I got yelled at by everyone in management) because they planned official deadlines after I mentioned in an official document that the software was ready to demonstrate to management when it clearly wasn't, which meant they had to further delay a multimillion-dollar asset. This gave me the worst job-related scare of my life because they said that they had grounds to not just demote me, but to "let me go" (their words) because of the amount of money involved. I assume their superiors expressed to them how "unhappy" they were about the delay. Thankfully, I only got a warning because the problem was fixed quickly, but since then I've been too paranoid to not make sure that every word I write in official documents is 100% confirmed as true without a reasonable doubt. So it blows my mind how these lawyers did every single little thing you could do to do the complete opposite
I think legally it's (usually) the fault of the company rather than the individual. Or at least based on the cases I've heard. The reasoning being that the company processes should've caught it in the first place, and so they're equally liable.
@@supersonic7605 I am assuming, if only because the one lawyer asked if it was lying, that these lawyer didn't understand what a GPT model program is. I think they assumed it was an ACTUAL Artificial Intelligence. aka an Artificial Mind, one that could actually think on its own and not need input to generate any answers.
I think, given that none of these lawyers did any actual lawyering, thought that the GPT could do all of their research because it would collect data from various sources, read it understand it and synthesize a legal document for them.
The law firm itself, at the very least, should have terminated these guys, just for the sheer embarassment. This has certainly cost that law firm millions in revenue. They should also be debarred for failing to actually act as a lawyer. I wonder if the judge actually imposed a sanctin on the lawyers as well. hopefully they have to pay all the legal fees out of pocket for everyone involved and not take any pay, and perhaps get debarred or something.
I must admit that when the lawyer admitted, under oath, that he lied to the judge about going on vacation, I had to get up and walk around I was so stunned. Lying to a Federal judge? Sheesh! How did that lawyer ever pass the bar?
Because the US system allows pay to win for literally everything
Also, humans can know the information contained in an ethics class and answer questions based around it. Without actually understanding or agreeing with the information.
Passing the bar has nothing to do with practicing law
Passing the bar shows you know how to write a really hard test. That's kind of a separate skillset from learning how to navigate court without angering a judge.
There's always the people who pass at the bottom of their class.
After working for the Sacramento County Superior Court of California, it's crazy that attorneys would try to lie to a Judge. Judges are like gods of their court. NEVER mess with them. They're smart enough to figure it out. They started out as attorneys themselves. I got this from nine months of working as an IT specialist for the Court. Judges can be very nice people, but don't try to mess with them. They are not amused by legal shenanigans.
I even overheard one Judge in chambers who was speaking with a woman suing due to being injured in a car crash. He actually went out of his way to tell her that "he didn't want to speak ill of her attorneys, but it seems to me that your settlement should be far higher based on the photographs of your injuries. This is not legal advice, so if I were you, I'd consider making sure your attorneys have these pictures and are taking them into consideration." Okay, I'm paraphrasing, but he was oh-so-slyly suggesting that this woman get better lawyers. He was also one of the smartest, no-nonsense Judges I'd ever met. And he didn't suffer fools gladly. But the fact that he went out of his way to help this woman was incredibly good of him. Considering how short he could be, for example, when his computer wasn't working the way he expected, I was surprised to find out how generous and gentle he was with helping plaintiffs out.
It sounds unethical to me that the judge offered such "not legal advice".
@Jack You’re absolutely right. I wouldn’t say it’s “usual” at all for judges to be attorneys first. On the other hand, he was a federal appointment.
Upon wiki-ing him, he did practice privately in NYC for 26 years.
This is what I'm most concerned about with our judicial system given the political climate and the way judges were selected in the last administration. Judges are human and fallible, yes, but generally speaking the system has honed itself so that most judges are like vigilant guards watching over those symbolic scales. Sometimes it's out of personal interest that they are VERY not okay with someone/a group tipping those scales whether through bias, incompetence, ideology, etc and sometimes it's genuinely caring and taking their role in democracy seriously but whatever the motivation it plays a critical part in our lives.
Hoping that at least now many more people recognize how important this branch of government is
The first rule of practicing law is “don’t piss off the judge that is hearing your case.”
@@cparks1000000
If it's unethical to tell someone they deserve more money for their injuries than what their hack lawyers are trying to get them I don't want a ethical judge who will let me get screwed over.
Being raised by 2 attorneys I grew up with my parents talking about cases (obviously when or if they could) and the specific details of them. I find your videos incredibly fascinating and entertaining, thank you for being so thorough with every detail!
As an accountant, this video caused me physical pain. This sounds like a literal nightmare anyone in a legal or finance profession could have. I am genuinely surprised neither of these men broke down sobbing on the stand.
Who says they didn't?
I imagine it's not any better when the entire legal community is pointing and saying "Ha ha!"
*shudder* dealing with a client's lousy OCR system is bad enough. I cannot imagine the disaster that would ensue if someone let a generative AI near financial records or reports.
@@TheGreatSquark You will likely see it first in the investment side of things.
@@TheGreatSquark I imagine "the ai made a mistake" could be a nice excuse for fabricating numbers. At least would expect less trouble than "yeah we lied to mislead investors".
Me, seeing a Legal Eagle video: An analysis of the Trump indictment already?
Me, watching the Legal Eagle video: *Never mind this is so much better.*
He 100% should cover the Trump stuff, but it's nice he sprinkles in these sillier stories between them
I agree. He will get it done he's just taking time to get it right.
I wouldn't be surprised if it's already up on Devin's nebula. He does say a lot that his videos go up first there and there's a delay before they go down to TH-cam
Imagine you had to film this, and you are barely done reviewing the edits, that the Trump thing comes out…
Wouldn’t you just have a Spa day, before swimming in the… what’s the German word again?
@@bertilhatt Schadenfreude?
Seeing this a second time, it's even worse! I was just telling a coworker about this last night and he was blown away that a lawyer did this.
The judge was straight up savage.
I think the most eye opening thing in this whole video, is discovering that the book shelves are actually real, and not just a green screen lol
Ong
same
I didn't even notice. I just assumed he had it as a prop ready for this moment.
Came here hoping to see that I wasn't the only one who thought this!
Same 😂
I have some minor sympathy for the lawyer claiming he thought chat gpt was a search engine, given all the hubub and publicity about google and microsoft introducing so-called "ai search engines" a while ago. But the fact that he simply did not check *any* of the information provided is aboslutely mind boggling. He didn't even understand what the citations meant! It seems likely to me that he's been merrily citing cases without reading them for years, and this is just how he got caught. What a mess.
By the sounds of the description Devin gave, Mr Schwarz was not a federal lawyer, hence getting Mr LoDuca to file on his behalf. It is plausible (though given he's apparently practiced law for 30 years, something of a stretch to believe) that he simply wasn't aware of the federal nomenclature.
I have none for those lawyers. They should have checked to see if the cases were real if they couldn’t find what they were looking for in other places. I got a lot of sympathy for the guy who hired these morons though.
@@KayDizzelVidsyou have sympathies for a guy suing an airline three (!!) years after he got bonked with a serving cart? Really?
@@KindredBrujah Maybe his law practice never really extended to courts any much and he was perma-stuck in the ghostwriter position, signing papers for the firm and the like?..
@@Jehty_ Even the dumbest parties deserve proper legal counsel. A better lawyer would have told him not to bother.
I am a mechanical engineer, and run into this situation recently. I was trying to use ChatGPT to shorten my initial research into a topic, it gave me the equations, everything. But since they were sloppy and missing pieces, i asked it to give me the sources for these equations so i can go to the original articles and collect the missing parts. Oh boy i was in for a big surprise. It just kept apologizing and making up new article titles, authors, even DOI s. It was eye opening to say the least.
As a fellow ML engineer i am surprised your are relying on the chatbot for anything related to research it may help shorten and make pre existing concepts more concise but it is merely a tool for research not the spearhead of said research
@@shahmirzahid9551 well, "relying" is a bit misleading of a term. it was a low priority topic which i were to take based on if it's feasible to do in a short timeline, and i decided to try out chatgpt on a "if it works works" basis. it didnt work, and i haven't used it since for this purpose whatsoever
What is a DOI?
ChatGPT is NOT a search engine!! You cannot use it as such
@@Tyrim ah i see i did the same when i do some calculus theory study but i just made a engineered a prompt for it to give some detailed explanation of things and it works like a charm i too had my doubts but yeah i wouldnt still blindly believe everything it said as it could be outdated or completely wrong
I went to check myself what 925 F.3d 1339 actually was; it's a page within a decision by the US Court of Appeals D.C. Circuit (the full case actually starts on page 1291) called J.D. v. Azar, one that had to do with the constitutionality of a Trump-era restriction preventing immigrant minors in government detention from obtaining abortion services. It was actually kinda interesting to skim through, if completely irrelevant to airline law.
thank you for looking it up and sharing a quick summary with us! Was curious to see if someone looked it up or not.
It may be relevant when these minors are transported via chartered airlines. Human trafficking itself is a major issue that airlines look out for, so there seems to be relevance.
The fact it's not actually a real case, just a page in a case starting from an earlier page, helps explain why a cursory glance didn't raise the red flags you get when you actually read the page in front of you.
Tbf the biggest surprise to me is that is indeed a valid citation, and not some hilariously out of bounds non-existent thing.
Has anybody offered an explanation of WHY ChatGPT gave the false reference and was so adamant that it was a real source? Could ChatGPT be pulling from a fake law source itself? Did the programmers do this on purpose? I use ChatGPT regularly for work, and while not perfect, it's about 80% accurate in the IT space. So why would it be so far off in the legal space? It has been successfully used in the academic space also, to the point that some teachers and professors can't tell a real paper from a ChatGPT paper apart.
This is how one of those "He never went to law school but he's practicing law like a pro" TV shows would actually go
I remember the actual Zicherman v. Korean Airlines case, it was 1996 not 2008 like ChatGPT cited. A Korean Airlines flight entered Soviet airspace in 1983 and was shot down killing all 269 on board. It's a poor case to cite even if they'd gotten the citation correct and would have only hurt their case.
Jeez that's rough. Those poor people. Their last few moments must have been spent terrified and angry...
And they were citing that to make an argument about someone's knee injury... 🤦♀
Didn't the Soviet Union dissolve in 1991? Was it still considered Soviet airspace back then?
@@andrewli8900 the shoot down was in 1983, the court case happened in 1996 against the airline, which would be why ChatGPT chose to reference it, it was more than 2 years after the event, but the 2 year limit doesn't apply to willful misconduct. That's why it was a terrible case to cite, because it didn't apply in the current case and would have only served to further support the airlines position.
Superintendent Chalmers: “Six cases, none found on Google, at this time of year, in this part of the country, localized entirely within your court filings.”
Principal Skinner: “Yes.”
Superintendent Chalmers: “May I see them?”
Principal Skinner: “…no.”
Thanks for the laugh!
i could hear their voices
Seymour! Your career as a lawyer's on fire!
@@ThePkmnYPerson No, Mother! That's just the Northern Lights!
Well, Loduca, I'll be overseeing this case _despite_ the statute of limitations.
Ah! Judge Castel, welcome! I hope you're prepared for an unforgettable docket!
Egh.
(Opens up Fastcase to find legal citations only to find the subscription has expired)
Oh, egads! My case is ruined! ... But what if... I were to use ChatGPT and disguise it as my own filing? Hohohohoho, delightfully devilish, Loduca.
I don't think I've ever seen Devin this apoplectic. Not only is he ashamed for these clowns, he's visibly angry that they tried this crap.
Wait a couple of days so he recovers from the latest Trump thing…
Partly because this is the thin edge of the wedge. Chat GPT will be used more often to "improve" writing and these fake references will become common.
It's not just Fremdschämen, it's that this makes the legal profession look stupid 🤣
As he should be, for a whole ton of reasons.
@@ianb9028 Easy enough to automate basic verification of references though: program parses a set of references, queries a law db about them, and reports which ones it couldn't find. Human then goes looking for these references, to see if they exist.
I mean this wouldn't be good enough for verifying your own references, but it's absolutely good enough to catch most fakes from opposing counsel. And I can't see judges ever deciding that fake references are acceptable.
It's not just law. When "discussing" scientific issues, chat_gpt creates references to scientific papers and books which do do exist.
As a current engineering student, I feel like I'm going insane seeing so many other students rely so blindly on this stupid thing, it's gonna produce so many morons
Chat GPT doesn't actually "know" anything, it just produces things that sound realistic.
The language module has a concept of what realistic sounds like based on its input data, it has no concept on what is real or how reality works.
It is a very good parrot with no internal understanding of what it says.
@@EWSwot Yep... In a sense it's like those "How English sounds to non-English speakers" videos: It _sounds like_ it's answering the prompt - but that's all. Which may sometimes overlap with a sensible response, or other times make no sense at all.
As someone who was following ChatGPT's development, witnessing its sudden arrival into public consciousness has been... what's that word for secondhand embarrassment?
People keep trying to get a clever program to do things it was never designed to do, couldn't do if it was programmed to, and would be questionably legal if they could. Seriously, if AI is still struggling with how many fingers humans have, how do you expect it to understand legal issues?
Yeah it is excellent at metafiction
When he reached back and grabbed a book, I gasped. I always assumed the background was a green screen. I’m sorry for selling you short, Devin! Your content is great!
1000% same.
It looked too good to be a green screen
i cant believe those are real books XD
It actually is still a green screen, but he had the book available within arm's reach. You can actually tell by how he reaches the book, and how the book is angled when he pulls it out, as well as the off lighting from the background compared to his face.
@@temi19 i don't think so, but i haven't watched the part where he takes the book because i just skimmed through, do you have the timestamp
Chat GPT: great for generating plot ideas for my 9 yr. old's D&D games.
Chat GPT: not great for actual legal court cases.
This shit needs to be dealt with in the harshest way possible.
Imagine if the defendant didn't have or could not afford adequate legal representation.
This case might have gone straight to a default verdict, without anyone checking anything.
I think the more reasonable take on this is that it’s a good thing that someone mucking up with ChatGPT happened in a case clearly without merit. The judge can fairly eviscerate counsel without depriving the plaintive.
I'm not a lawyer, but this definitely qualifies as malpractice. LoDuca and Schwartz are FFFFF'd. They submitted everything under penalty of perjury, so they are definitely getting hit hard.
And in this case - LE seems pretty sure that the plaintiff's case is nonsense but imagine if they had had a legitimate case. They would've lost because their lawyers screwed up.
Well he could always copy past everything to ChatGPT and ask it to check if the content was AI generated.
@@randomwerewolf1099 I'm no legal expert, but aren't plaintiff's counsel here liable for something like fraud or malpractice to the detriment of the plaintiff?
Lol…as a medical student the amount of confidence chatGPT has with explaining disease pathologies that’s completely wrong is concerning. It does a good job of coming up with an answer that sounds right but isn’t.
That's because, technically, it is. There's a reason that most actual AI researchers call it a Language Model and not an AI, because that's all it is.
It know the language of law books, or the language of medical opinions. It does not have the facts, let alone up-to-date ones.
Yes actually, blaming that is completely fine, because the simple fact that it is capable of lying to you with confidence means that its work is literally useless. In your own last example, the scope of what you're suggesting it should be used for doesn't even make sense. You would get the AI to do 5 minutes worth of work, just so that you can spend 50 minutes fact checking it. Do it right the first time instead lmao@@A-wy5zm
Someone on the Gardening subreddit recently used ChatGPT to try and answer someone's question about pet-friendly plants and was SO CONFIDENTLY WRONG the mods actually had to step in, because the advice from ChatGPT could've literally killed this guy's pets. I had to go on a rant about Language Model hallucinations and the demonstrably failing accuracy of the output from these systems.
It's really validating when the mods leave your factually correct, even if angry and spitefully written, comments and delete the moron's 😅
In essence, chatGPT is your drunk uncle. It hears half of what you said, and spins off a long story based on something a friend told it 20 years ago with "facts" sprinkled in to support the argument it wants to make
It's become a bit of a meme in the crochet community to ask chat gpt to write a pattern (usually a plushie because they're small and quick) and laugh at the mess it produces. It only looks like a pattern if you've never seen a pattern before and think crochet is done with needles.
When I was reading the tweets on my own, the part that made me cringe the hardest as a non-lawyer was absolutely the “Were you on vacation?” “No, Judge” exchange. Wanted to be swallowed by the earth. Wanted to disappear from existence. He lied about LITERALLY EVERYTHING holy shit.
Basically career ending answer.
I took psychological damage listening to that exchange.
@@tbotalpha8133 I took emotional damage.😁 But yeah, when even complete strangers wince when he stated "no", it was not looking good.
Remember folks, what ChatGPT can and can't do is literally in it's name: it's a CHAT bot.
All it does is... keep up a conversation it thinks you want to have. That's where it starts and ends. It makes up the "facts" that you want to know because it doesn't really know anything, feeding it the "knowledge" just tells it how these facts are linguistically structured, so it can create a text that RESEMBLES what you are looking for to keep up the conversation.
In a basic sense, it is trained on how a correct response “should sound”. It doesn’t comprehend language and information like we do, it doesn’t have an abstract understanding as to why those documents it’s trained on are structured like that like we do, it just knows that they are, and frames a response accordingly. That’s why, as he said in a previous video on AI lawyering, GPT is known for “eloquent bs”. It sounds right, but it doesn’t have the ability to understand “this sounds right because it contains factual information”
It's basically a slightly more coherent version of if you kept hitting words from auto correct and used grammarly to check for a mistake.
Nothing of sustenance will be said, and it will fall apart the longer it goes on.
LITERALLY. the creators (wildly dishonest) marketing hype didn't help, but I'm still amazed that people apparently just need to see a 'style' or 'sound' of typing to immediately think "wow. this thing must be factually correct". Bro
It seems to me that the use of the term AI is too loose, when applied to these types of program, at least to me as a layman. AI implies that there is some sort of reasoning going on, whereas in fact it is just language modelling.
It does do programming though. And programming syntax is an exact business.
With the right question and understanding of its limitations it can do some excellent work for you.
You have to verify everything, but even then it can save a lot of time.
Basic boilerplate, examples in how to use a new library.
It's a great tool, but if you are a poor programmer, it won't make you a great programmer. If you don't understand what it gives you, you are likely to fail just like these lawyers.
Always read the caselaw cited at you in a brief my friends. On many cases when responding to motions, I discovered the authority being cited at me said the exact opposite of what opposing counsel was using it for. Nothing is more satisfying than going into a hearing and throwing opposing counsel's caselaw back at them.
Hear..hear. Not only opposing counsel, district attorneys often guilty of this too. (Even some judges, but you did not hear this from me).
"If you want to get technical, they make stuff up"
This one got me in stitches.
Truly a technical way of putting it.
As a New York defense attorney; I wish I could say I haven't run into plaintive attorneys this stupid and incompetent. But I have, many times
I guess ChatGPT at least allows us to find them quicker?
At least ChatGPT is better at spelling
Does AI know how to spell "plaintiff"?
@@annthompson6600 when you realize “plaintive” is an actual word, and that autocorrect is a thing. This isn’t a brief or motion, this is a comment section
@@Dr.MikeGranato That’s very true, but the opportunity to review what one writes is a reason replies aren’t live-posted. Even after posting, editing remains possible, if one recognises one’s error.
Anyone who's dumb enough to use ChatGPT to completely do their work, especially something as critical as law, doesn't deserve to be in that position. As an accountant, I've been encouraging my coworkers to use the AI for things like drafting emails, writing excel formulas and VBA scripts, etc... rote things. However, I VERY specifically emphasize that it is a tool to add to your arsenal, NOT a replacement. You always have to test or verify the info it gives you.
I tell non-tech people that using ChatGPT is worse than using an enthusiastic unpaid Teenage Intern.
Would you let such a Teenager handle all your writing without review? Would you trust their "research", without a review?
Why are you doing those things with ChatGPT? And not reviewing its output.
At least with a Teenager they can actually learn, and explain their work.
ChatGPT doesn't understand anything, and is an Authoritative Idiot (AI).
As a software engineer, the most I've used ChatGPT for is coming up with prompts for backgrounds for NPCs in my D&D campaign. I let it plant the seed, then I warp it to fit my story and my creativity. Even I do more due diligence than these lawyers for my fantasy campaign...
@@preo720 Oh yeah, it's great at rubber ducking with creative ideas.
Oh yeah. Whenever I use it to write something, I only give it very specific facts and check it over afterwards! NEVER trust it to be completely correct if you give it free reign to do whatever it wants.
Yeah. I always check if the things chat GPT puts out is bs or not. Very useful tool but you have to actually verify if what it’s talking about is real and factual. Often times, it spouts fictional nonsense.
I spent over two hours trying to stop Chat GPT to hallucinate over the contents of Anna Karenina. No matter what I asked or how much I tried to correct, it was never able to produce a single paragraph from the actual book without adding on made up characters, relationships or events. It always sort of sounded real if you've never heard the story but It was always wrong. Even something as simple as feeding it a line from the book and asking what chapter that line was from it never got right.
That's because it's not a database. It hasn't even ever seen the material it was trained on.
AI training extremely oversimplified: A program grabs some text, takes the beginning and shows it to the AI, asking it to continue. Then it takes the AI's response and compares it to the real text. Depending on how much the response looks like the original, it either slaps or cuddles the AI (i.e. changes the AI's internals). Repeat for as long as possible.
ChatGPT has closed training data. It will not respond to you trying to correct it. No amount of hours you spend correcting it will actually change the model of GPT4. You can't teach chatGPT new stuff, only openAI can do that. And that takes millions of dollars and weeks, because they used the wrong technology for rapid learning.
@@NineSun001 I thought ChatGPT is GPT 3.5. Isn't GPT 4 the one that includes images and videos?
Chatgpt is like a human being. Could _you_ cite a page from Anna Karenina perfectly? Well neither can chat gpt.
@@zhelmd it's nothing like a human lmao. it's all just predictive texts and cannot think or reason as people can
This was my favorite LegalEagle upload to date. The build up to the judge grilling the lawyers is great storytelling. The situation depicted -- there's no facepalm big enough to cover the gargantuan irresponsibility on display.
It's like smashing a house of cards with a hammer
Im not in law but a peer of mine was bragging to me that they used ChatGPT to cite and write their politics paper and last I saw them they were having a breakdown because ALL those citations were false and the teacher caught them. I did feel really bad but also you did this to yourself.
"You told me that ChatGPT supplemented your research, but what was it supplementing?"
STOP, THAT LAWYER NEEDS 5TH DEGREE BURN CREAM
And definitly some oxygent supplement...he will have heart attack or passed-out foaming at the mouth.
This is a great example! It went wrong in so many ways. We also did a video where we experimented with how it would handle doing legal research and tested it on some scenarios related to autonomous weapons and international humanitarian law. The problems with its legal reasoning turned up really quickly as well!
I am a physicist and I kind of feel 20% bad for these lawyers because I can see exactly how this sort of thing could happen. In my own experiments with CGPT, I asked it for references about a topic and it gave me 4. These looked legit with names of people I recognized working on the sort of topics they usually work on. I went to those references and couldn't find them, but they were convincing and if I hadn't been paying attention when writing a paper (and if the way we put citations in papers wasn't automated in such a way to make this impossible) I could have put the citation in as a placeholder with the intention of checking it later. With 40-60 citations per paper, it would have been easy to miss that one in the checks.
The reason I only feel 20% bad for the lawyers is because citations in science are a very different thing than in law. In law, you don't cite a case without describing in detail why it is pertinent. In science, you might cite 60 papers with most of them just being "here are examples of other work on this topic". You generally only write in detail about the findings in your citations a few times in the paper when you have a very specific point you need to make.
I am a grad student in biochemistry. One of my first experiences was asking the bot about an old virology paper I knew kinda well. It told me that it was written by people that were not the authors. I distrusted it ever since…
This is a well known facet of the way Chat GPT works. It uses a prediction system to form its output and doesn't have any idea about what the contents as a whole are. It doesn't think about anything except what the next word should be based on the prompt it was given and wouldn't have any idea what any of it means because it's just a language model. It's a very, very specialized AI but it's results are convincing enough in casual conversation that a lot of people are fooled into thinking it knows what it's talking about.
You should check your citations in science too. I know it's not always done well, but at the *minimum* read the abstract and skim it to check it is relevant to the place you cite it. I used to always check a few of the references randomly (plus the key ones) when refereeing papers. Occasionally they cited completely irrelevant (or even contradictory!) material: that was an immediate recommendation to reject from me. You lose all credibility and trust. Checking citation chains is an absolute pain, if the authors of a paper don't even check their direct citations, then the whole edifice collapses for everyone.
So, yeah, writing papers is a pain, and referencing it is nobody's favourite part of doing science. But you've still got to do it.
@@QuantumHistorian Absolutely. I'm not saying I would have made this mistake, but I can see how someone who wasn't skeptical of GPT and didn't know how it worked could. This is new technology and people aren't used to the idea that something could be totally convincing AND completely fabricated.
The first time I used it, I was actually just trying to use it to find information on a topic that I was struggling with in a paper. I was excited when I saw the titles of the list of citations it gave me, because I was like "YES! That's exactly what I was looking for." Then when I couldn't find the actual paper, I asked it if it was real, and it said "sorry, that was an example of what a paper on that topic would look like. Here's a list of REAL citations." Those weren't real either. Every time I told it so, it was like "oh, okay. Well, THESE are real" and it kept giving me more fake citations. It was pretty hilarious.
@@Phoboskomboa it’s so funny watching chatGPT double down when it’s wrong, just like a real person would. Because it’s emulating a real person first and a good scientist second. I worked in a university math library for help and I had a guy come in asking about linear algebra. Turned out he tried to use chatGPT to explain the stuff to him, but chatGPT sources all it’s knowledge from Wikipedia. The field is so large+complex but specialized so the course was slightly more basic at our uni just to explain the math. GPT had no idea how the slightly easier but more complex problems should be done or how to explain them, and could only cite harder theorems and examples. Such a nightmare lol
I was so pleasantly suprised to find 80,000 hours sponsoring this channel! It's a great resource and all free, and I have genuinely been telling my fellow young and lost graduates to get on it
There is one thing I've learned about the legal system.
Do not, under any circumstances, waste a Judges time. Especially a Federal Judge.
Volcanic is probably an understatement for how irate he was.
I've tried using Chat GPT to figure out Tip of My Tongue answers on Reddit, like "what is the movie where a kid makes a bunch of weird contraptions and uses them to fight 2 adults" and ChatGPT will be "oh, it sounds like you're referring to and episode of Friends, episode The One With the Couch where Joey and Chandler engage in nonsense and hijinks", where it could be legitimate characters and show, but an absolutely wrong synopsis.
Someone once said "ChatGPT doesn't know facts. It only knows how facts look like"
The second most amazing thing about this video is finding out the bookshelf behind him is not a set he just uses to film, but actually has federal reporter books on it that he can reach back and show to the camera 😄
ChatGPT writing code is really hit and miss too. I tried to use it to solve a server side problem easily and was happy how fast the code came out, I still spent all day inputting error messages before giving up and solving the solution myself. Much quicker.
The hardest thing is it can't say it doesn't know without further information.
GPT is trained by scraping data from the internet. And on the internet, no-one /ever/ admits ignorance. So neither does ChatGPT.
As a paralegal myself; this entire thing was hilarious. Did they not have staff to vet any of this? We are required to run our attorney's stuff through the woodchipper to find shit like this. Our group had a laugh over this entire situation and felt real bad for any paralegal that was assigned to these attorneys.
Schwartz: "I thought there were cases that could not be found on Google."
Judge: "Did you look for them in a law library? Or check with another attorney to see if they could find them?"
"No, your honor."
"So when the court told you these cases were suspicious, did you then check with someone else to determine whether they were real?"
"No, your honor."
"When you pulled down the full text of the cases, did you read them? Did they make sense to you?"
"No, your honor."
"Did you do any research, given the full text of the cases to determine whether the people in the case were real, or to find out why you couldnt find the case anywhere else?"
"No, your honor."
"Disbarred."
Yeah, I do wish it had gone down that way. Hindsight is 20/20 and all that. The lawyers will probably be put on probation, given community service legal work & fined -no lasting consequences for their long term financial success in law.
The phrasing of this really brings to mind the interaction between Tom Cruise and the doctor on the stand in A Few Good Men. This was great. I just can't help but LOL at this whole ChatGPT debacle. I don't even feel bad for those idiot lawyers. It's the most basic common sense that if you sign something, you've read over and confirmed that everything is as you would do it yourself. I also thought about how empowered their paralegals must be who have their work out there and signed without the lawyers having a clue.
I honestly wouldn't have been surprised if the judge literally threw the book at them.
And THIS is why OpenAI has these terms of use indemnifying them against the information that ChatGPT gives you. Imagine if those lawyers could sue them!
i hate to break this to you but not even the terms of use will stop people from doing dumb shit if it means cutting corners to go to the final product
@@leaveeeevee5762but they do make it near impossible to sue them for that stupidity, so ultimately that's it's actual job, legal defense.
In our law firm, the litigation team that I am on has briefs, answering and replies, ran through more than one attorney. Sometimes one attorney will write it and go second chair while the other attorney argues it, but all attorneys pass it for review. Then it hits us, the paralegals, and it undergoes a third/fourth/fifth review when it's getting prepped for uploading and linking on the docket. WTF is wrong with those people?????
That sort of quintuple-checking should mean that by the time something is filed, almost all errors have been caught and eliminated.
Wow. I can almost picture the judge turning to the court stenographer and saying, "Watch, I'm about to end this guy's career."
Super piece -- just one small correction, since we know you're a stickler for details: in German "Fremdschämen" is pronounced "Fremt-Shaymen", because the ä in "schämen" is similar to our long A, as in "hay".
17:48 Asking ChatGPT whether the cases it cited are real is like asking someone who just lied to you whether they lied.
25:38 Today I learned that the legal eagle is not sitting in frobt of a green screen but that he actually has a big book shelf with thick law books in it
Looking at it carefully, it actually looks like CGI when he picks the book off the shelf.
He is a lawyer. They usually have a bunch of law books
All of this sounds like high school shenanigans. The petty lawsuit, using AI to do your homework for you, and getting caught by the principal.
Just hire a frickin paralegal next time!
13:36 I love how the case randomly turns from a wrongful death suit to a breach of contract suit mid-page. 😂
On the bright side, their client probably has grounds to sue them. Those lawyers were actually so good, they ensured their client still got a payout! Incredible.
Yes i automatically think that.
Ah, but that's assuming this whole mess hasn't already completely ruined them financially. Can't get blood from a turnip
If the original claim was out of time I doubt he can collect damages for the loss of the claim
I do feel like it’s fair for the client to demand a refund from these lawyers
@@davidarnold2456 If the claim was out of time, the lawyers should have known that, and should have clearly stated so to the plaintiff. If the plaintiff chose to go ahead anyway, then that's on them. It sure sounds like they just took the case without considering the merits, given how they handled writing their submission.
I was not expecting such a precise and comprehensible explaination of what a GPT model is and does from a lawyer with, presumably, no AI background. Really nice work!
I've used ChatGPT to get information for a university paper. The thing is that when you ask and get the references for the stuff you ask, you actually READ through the references.
Yeah, its incredibly good at making up references to papers that don’t exist.
usually the references do not exist because chatGPT is not intelligent
@@Corsaka Yeah, its mentioned briefly here but chatGPT is a predictive text program. It doesn't know anything except what words typically follow the previous. Unless a lot of people have typed up whatever source you are looking for, and those sources have been used for training, it won't show up.
@@barryfraser831 Exactly. You have to look through the sources it gives you. You know, read them.
So I'm in law school now and just wanted to let you know that this was in our CivPro casebook. So thank you to the attorneys involved in this case for giving an immortalized example of what not to do!
Chat GPT gets the answers to online quizzes wrong about half the time. And I am talking about a basic stats class. I can’t imagine something like Law or Medicine relying on such a wonky ChatGPT
TBH I was really impressed that the books in Devon's set were real. With that said, It really reinforces the point "Don't Lie to the Judge!"
Yet none of the ones in frame seem to be recent Federal Reporters
I'm actually not surprised at all, not really anyway. Given he didn't seem to use chromakey in the one video I saw over the winter where he went over lawyer memes (I believe he said he was on a ski retreat?), and he probably spends a lot of time in his office anyway (lawyers tend to do a lot of paperwork), any suspicions I had were reduced. Furthermore, it seemed reasonable to assume he would have shelves like that somewhere on his firm's premise.
That said, seeing It confirmed like this when my normal assumption is that all backdrops are chromakeyed was remarkable, if not surprising.
The biggest problem with AI right now is that it's not at all AI as any lay person understands that fairly well-established concept, they just keep calling it AI because the sales team wants big, big hype out of this and something like "statistically predicted data engine" doesn't sound sexy enough for their tastes. So here we have a fairly ignorant lawyer expecting it to act like human-level intelligence and it just is not that. LOTS of people are making the same mistake, just not with such high stakes.
Tbf, having no intelligence is still higher intelligence than some people we have here. Exhibit A and B being these fools.
Exactly. "AI" is a misnomer.
I discovered your channel through your video on the submarine accident. I love the way your format your information and how you explain these things. Also, the pitch midway through for your own lawform is *chef's kiss* proper modern day marketing. Short, sweet, and simple.
I look forward to more videos :D
As a lawyer myself, I am feeling sick to the stomach hearing all of this - starting with filing out of time. It just got worse and worse.
When the judge said they he was glad they were finally on the same page, I had to pause the video and take a second. Thats got to be the politest way I’ve ever heard someone say “I am now about to tear you SEVERAL new assholes.” Whole new level of intimidating coming from someone who has to be as composed as a JUDGE.
I’m a Canadian lawyer. We have CanLii, which makes it comically easy to search for cases, but people still do it. There were a few articling students who got “disbarred” using AI to cheat on their provincial bar exam equivalents.
When I use chat gpt, I use it relatively if that makes sense. Most of the time I ask like "how do I go about answering this question" (like what are the steps) in comparison to "what's the answer to..." and I think it's a good tool if used that way, because you actually learn. Since I've been doing that, test scores improved greatly, and I even got the most growth last school year in my tests out of the others. (Hence why my whole class was rewarded -- simply because I and another person achieved excellent growth) but I also use it in terms of reading and I feel like it betters my comprehension.
Like if I'm reading (away from school) and I find just a random article that I find interesting. I would read the article, and reread, and then develop my own conclusion. Then I ask chat gpt how they would conclude it, and their answer really makes sense and relates to mine but the best thing is, they use different words that I'm not entirely used too and so now I can become familiar with those words which I end up doing, which ties into me saying it bettered my comprehension. Not only because it made me view my own point of view differently, but because it showed me another one that just made sense.
In conclusion (I know this was a lot) If you want to use chat gpt, please use it right. Chances are, you're going to get caught if you're flat out asking for 100% answers, and if you do it that way then you truly aren't even going to learn anything, and I know teachers say that all the time but it's true. Thanks for reading my opinion on this if you did lol.
Great vid. You were NOT too harsh on these lawyers! BTW, I teach college and yes, ChatGPT does create fake sources and attributes fake dialogue to real people.... which makes it useless for college students' research essays.
Also makes it pretty easy to find students that don't actually do the work, it seems, if a quick search for a source in a place you should be able to find the source... returns nothing.
Helping with essays, papers, lab reports, etc, like how to better phrase something, or how to set up an outline, etc, is great. It works well for **helping** based on the massive amount of data out there it has to draw from to know how a paper should be formatted, etc. A good student can definitely use it to better write their own actual paper and reports, but that's all they should do with it.
For actual facts and the like... or pulling from actual direct information... you gotta do your own work and figure out what it actually got right, because most of what it does will be wrong or nonexistent.
Okay, as a German I have do admit that your pronounciation was quite good. Just one tip: Fremdschämen has an 'ä'. The closest thing to this is like the 'a' in apple. The rest was spot on.
I am lawyer in mexico, a few weeks before this, I used CGPT to look for precedents for a case. The IA outputed several precedents with the legal solutions that I needed to support the position I was defending. I was amazed at the IA finding exactly what I needed.
Of course I was taught in Legal School to always check on the legal source, that's how I found those precedents where invented by the IA.
I shared my experience in social media with other lawyers, and I wasn't the only who had already found out that CGPT it is not reliable for researching legal sources.
With further test I found Chat GPT wil even made up laws!
Do your job, do not delegate it without checking.
Good for you, I imagine Mexican judges dislike idiots who tie up court with nonsense just as much as US judges. And I thought making up laws was the department of the police, they try to do it here.
I'm a student and I use Chat GPT everyday as a sort of personal TA for computer science and discrete mathematics, and its only right first try like 60% of the time. I constantly have to call it on its mistakes, a lot of which are very obvious and can be spotted immediately. I still vouch for its usefulness though, I mainly use it to check my work, and ask clarifying questions on the subject matter and for that its really good, just don't trust it to do it all for you.
It has the same value as autocomplete when texting, saves time, sometimes.
I found its useful for when your trying to look up something with very little to go on. A bit like to going to a librarian to start research when you have no idea where to start, they will know enough to point in the right direction but are not an expert. An example of this was, I saw a road service truck with a crash attenuator on the back, and I wanted to know what the hell it was. But without even knowing what words to use, google was useless, but chatting back and forth with gpt with vague descriptions I was able to find out what it was.
Same. I’m a senior developer.
I treat it as a junior coder to whom I can delegate repetitive or mundane tasks, but who needs its work validating.
I know nothing about law but man does this guy keep me engaged haha 😊
As a law student, I got curious about ChatGPT and used it to look up a case... and then found out it made a completely fake case even after I put in everything INCLUDING the full citation. Ever since I have not touched that thing
My fed courts professor was showing us where she asked it questions about Burford and it just completely whiffed on softball questions about the case facts that even a student who didn’t do the reading would have had an idea of
Devin makes it sound a lot like the tool is not precisely to blame but that in the case for this video definitely and perhaps in many other situations the main threat is what if a lawyer is stupid enough not to do a basic double-check.
I could imagine that spending even 5 minutes could yield results of "well okay I can't use it for That but it might be able to do This" and the ChatBot comes out of it looking a whole lot better when you toss out the examples of absurd negligence and cartoon buffoonery.
There might still be tons of limits on what the tech is useful for but it seems to come with warning labels for what it is and isn't designed for. And maybe following those would help.
I'm sure some tasks it's good for they just used it for totally the wrong ones. And it's pretty clear they were enough at fault personally as people that blaming the robot would make as much sense as blaming the printer.
It’s because it isn’t a search engine. Text generators like ChatGPT are just designed to create plausibly written sentences, not true ones. I’ve worked with similar programs for creative writing, where I would write a prompt, then feed the prompt back into the machine and give it its own response as a new prompt. The machine would “hallucinate” often, generating all sorts of weird nonsense with gusto, because it’s only function is generating sentences that are grammatically correct and somewhat related to the prompt you give it. It doesn’t “know” the difference between accurate and false information. Nor does it know anything about mortality, which is why it’s awful at writing dramatically satisfying stories. This seems to be hard be for people to understand. And I don’t blame them. I don’t think companies have been forthright about the fact that their program is interesting but far more “artificial” than “intelligent” in the human sense.
I am a A&P student. Went through and made Chatgpt answer questions.
It was wrong 80% of the time.
Majority of the questions were basic, from Hazmat to Electricity to Powerplant. Stuff that anyone could know off the top of their head if they studied.
I had past work sheets that were graded. Yeah it did not pass 8 out of the 10 times. Word for word, with answers given.
I sometimes use it to get formulas or to write notes on basis of what I've typed up so far.
If you ever feel like dropping out of school because things are getting too difficult, remember that these idiots not only got through law school, but passed the Bar exam.
Yeah: clearly they cheated! Haha!!
There's a good chance they did work hard and know their stuff... once upon a time. After practicing for 30 years, they don't care anymore and are just collecting a very large paycheque without doing the work.
ChatGPT can pass the bar exam.
15:03
OK but at least the judge and the guys he spoke to at the 11th circuit probably had a good laugh at the absurdity of this all.
As a layperson, I've learned that anytime you hear the word "novel" used in referring to a legal theory or strategy, it's usually code word for "bat shit crazy", and a good way to piss off a judge.
"It's a bold strategy, Cotton, let's see if i pays off for them."
Glad this man was a lawyer. Imagine if he'd been a doctor.
Proctologist, proving it was an inside job.
They’d be great at euthanizing abandoned and rabid critters at the county animal shelter… maybe if they didn’t drop the responsibility to the secretary.
"You see, ChatGPT told me to cut your heart in half to make both halves work 50% better"
just wait for it
Just wait, it's coming.
21:28 never seen such a profecional way to describe "Cringe" in a different language.
I am not involved in the medical field or the legal field. But I do know that I was making a game (and thus didn't feel like I really needee to do research for it) on _history_ on ChatGPT back when it first came out.
And it told me that Juan Peron was the President of Argentina...
IN 1939!!!!!
I don't know much about law... but this feels like someone cheated their way through law school somehow and is only now being caught. Especially at the "Do you know what F.3d stands for?"
Yep
Devin reaching behind him to grab one of the books is something I wasn't expecting
We should stop misslabeling ChatGPT as an AI and start calling it was it really is:
A Language Model.
It's marketing
a machine learning algorithm. idk a language model wouldn't make so many referencial mistakes XD
I've been calling it a probability calculator for words
I'd say it's definitely intelligent, there's still a way to go but it can pass a lot of college-level tests and whatnot, which is something an average human can't even do. Following that to its logical conclusion would mean that the (below) average person is simply collecting words and memories in a chunk of neurons, nothing to call intelligence which I feel like is super dehumanizing
You mean text generator.
Once again, I'm completely reassured as an editor about how much my job still matters in the era of ChatGPT. What a train wreck!
just because it's awful doesn't mean someone won't try to replace you with it.
@@tsm688 Yeah, but that will bite them in the ass.
sure you wont be saying that when GPT 5 rolls out
@@crepooscul if fools use a tool like that it will but prompt engineers make sure the bite them in the ass part doesnt happen
The funny thing is, this isn't even just "because he used ChatGPT". It's because he didn’t double check ANYTHING. This is like going to your lawyer friend and saying "Hey do you have any good cases I could cite for this?" and then when they give you some names you're like "Alright cool I'll just write this down without even looking up the names myself". Literally all he had to do was a few Google searches.
Worse! Honestly at least in that case the person you were talking to was a lawyer. Forget a lawyer, at least they'd be HUMAN.
@@derekprospero literally
in this episode, i realised this wasn't a green screen but an actual book case in the background. The green screen was on the lawyer tho... well done CGI... weeeeeeell done...
Wow!!!! I have never seen a lawyer (either real, or played on TV) pull a book from the shelf behind them and open it!!!! Bravo!!
I always assumed that was a greenscreen backdrop :)
@@jatkinson85 i still think it is, throwing a really expensive book don't seem likely to me for a TH-cam video.
"How about that? I looked up a case! These books behind me don't just make the office look good, they're filled with useful legal tidbits just like that!" -Lionel Hutz
15:40 Considering what happened to the lawyers in this video, I suggest changing your slogan to "You don't just need a legal team, you need a *legal* legal team."
Not going to lie, I've been assuming your background was a green screen because it would be easier to film than sticking your desk in a corner. Props for being able to actually grab one of the books cited (and explaining what so many of those books you see in attorney's offices are!)
Legal Eagle isn't Phoenix Wright: he actually *reads* the books in his law office.
Though I think Phoenix Wright would also be heavily ashamed to see these lawyers encountering sanctions this way, or the possibility that he may be compared to them; and that's saying something.