Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
All right, good morning.
Speaker 2 (00:01):
AI is everywhere doing everything, and it's an amazing tool
and you can use it to make yourself more productive
and to start a company with three people instead of
thirteen people, and you can do all kinds of things.
But AI is not perfect, and when you are involved
in something even vaguely close to a mission critical operation,
(00:21):
you better be very careful when you're using it. And
some lawyers seem to be learning that even though it
seems like a thing they should have learned already. Joining
us to talk about AI hallucinations and other aspects of
AI use in trial and other court proceedings as Crawford Applebee.
Speaker 1 (00:40):
Crawford is a partner at Wisner Baum, a big law firm,
and he joins us on Kowa. Thanks so much for
making time for us.
Speaker 3 (00:48):
Appreciate it, all right, Ross, thanks for having me so, I.
Speaker 1 (00:54):
Was obviously not surprised.
Speaker 2 (00:56):
We've heard a lot of small stories, well maybe they're
not small to someone who's a lawyer, about lawyers using
AI for briefs and things like that where there are
hallucinations involved. And I want you to tell us more
about what's happening. But I didn't know until I was
reading up to talk to you that the actual federal
(01:16):
government got in trouble with federal attorneys doing the same stuff,
So can you please describe what the problem is and
then also.
Speaker 1 (01:25):
What you think the scale of the problem is.
Speaker 3 (01:30):
Yeah, so I think obviously we're only finding out about
the situations where attorneys are being caught by courts or
by opposing counsel. But I think what's going on here
is that AI creates a strong temptation for people to
cut corners. It quickly provides confidence sounding polished answers even
the most complex legal questions, and the words it writes
are very impressive at first glance. And so when you
(01:51):
combine that with how much pressure attorneys are under to
produce quality work as fast as they possibly can, lawyers
will fail to verify what ay I give them, and
then you end up in these situations with fake cases
in court documents.
Speaker 2 (02:05):
Just one quick follow up, how big a problem is it?
Speaker 1 (02:08):
Right?
Speaker 2 (02:08):
You know? Do you think this is a thing that's
happening every day but only being caught once in a while.
Speaker 1 (02:15):
Is this Are we making a lot out.
Speaker 2 (02:18):
Of something that's actually a small problem, or do you
think it's kind of a big problem?
Speaker 3 (02:23):
I think it's a big problem, but it's not going
to be big compared to what's coming. The reason that
it's a big problem now is because of ethical rules
that attorneys have to follow in order to be licensed
and represent people. And as the court pointed out in
that recent DJ case, there's something called a duty of candor,
which means that you can never represent false facts or
(02:44):
law to the court. So when you fail to verify
the law and your briefs and you file it and
your brief contains fake cases, that's a problem with the
duty of candor.
Speaker 4 (02:53):
So when I looked at this AI hallucination database, it
showed over eighteen hundred cases identified so far, a majority.
Speaker 1 (03:00):
Of those cases in the US.
Speaker 4 (03:02):
When you look at these cases, obviously any type of
hallucination is not good. But I almost want to ask,
is part of the case accurate? Is some of the
case accurate?
Speaker 1 (03:16):
Is it all made up?
Speaker 4 (03:17):
What is the situation of when AI is actually interpreting
a case in this situation.
Speaker 3 (03:24):
That's a good point because when people think of hallucinations,
they tend to think of things that are just completely imaginary.
You know, a fake case name with a fake site
that doesn't pull up anything. There's other types of hallucinations too.
I've seen hallucinations where AI will tell you that a
case is quoted for something when it doesn't contain those words,
but it's a real case. AI will tell you that
a case represents a certain rule of law or is
(03:44):
about a certain subject, and then when you read it,
it's not actually about that. So hallucinations can be They
can cover more than just a completely pretend case. They
can cover just misrepresentations of the case that exists. Right.
Speaker 2 (03:57):
And the first thing you said, where it just it's
a case that doesn't exist, that that also happens, right,
So it's that it's that whole range. Now, I don't
blame an attorney necessarily for wanting to use AI as
a as a tool, right, And I probably shouldn't ask
you in public whether you use AI. You can you
(04:19):
can talk about that if you if you want to.
But going back to your duty too, okay, So, but
going back to your duty of candor, obviously you have
to fact check everything that it does. But I wonder,
going back to the way you open this, how much
of this do you think happens? Because there are a
lot of attorneys out there who just aren't very good writers,
(04:41):
and they see that AI writes better than they do.
Speaker 3 (04:45):
I think what you're going to see is that this
happens at all levels. And I think we've seen that,
and like as you pointed out the eighteen hundred odd situations,
you're going to find some of the most prestigious firms
in there, and that's going to continue to happen. But
I think the important thing to understand is that the
cure for AI hallucinations right now is human verification. And
(05:06):
so until the technology gets better and we can find
new ways to check and catch these kinds of things,
it's going to just keep happening. But no matter how
great the output is from the AI, you have to
check it. And people may say, well, that sounds counterproductive
because AI is supposed to be making me more efficient.
Well let's think about it this way. Let's say that
it would have taken you about half an hour to
write a paragraph of something, and AI can write that
(05:28):
instead in about a minute or two. Now all you
have to do is spend about five to ten minutes
fact checking that and making sure the law is real,
and you've just saved yourself. Twenty twenty five minutes, So
you've got to play the long game with AI. You
got to use it appropriately. It's all about shaving off
minutes and at the end of the day, you save
yourself an hour.
Speaker 4 (05:46):
Are there any other safeguards you feel like law firms
should have in place when using AI besides just having
somebody that fact checks. And it's funny that you said
counterproductive because a lot of people think of AI as
eliminating a person and now was like, well, no, you
to make sure you still have a person to check
AI to make sure that it's accurate.
Speaker 3 (06:05):
Yeah. Absolutely, I think human verification of the output, especially
when you're going to do something like file something with
the court or represent something to say a jury. You know,
we're going to see some situations in the future where
there's falsified evidence where a client might submit a AI
created video or image to their attorney, the attorney want no,
They'll file that with the court and they might get caught.
(06:25):
Things like that that are going to start happening more
and more.
Speaker 2 (06:28):
We're talking with Crawford Apple Be, a partner at Wisner Baum.
I'm not an attorney, but I do love reading constitutional law,
Supreme Court cases, federal appellate circuit cases and so on.
And sometimes I'll talk about something on the show and
I'll go to AI and just ask it, has there
been a case about this topic?
Speaker 3 (06:47):
Right?
Speaker 2 (06:48):
And obviously that's a kind of thing that lawyers would
do all the time. And it's not that hard then
to go look up to see whether that's a real case. Now,
AI has never told never hallucinated to me, but I
haven't asked it very many of these questions. But I
can imagine it's saving an immense amount of time.
Speaker 1 (07:05):
We got just over a minute left.
Speaker 2 (07:08):
I imagine the range has been all over the place
on this next question. But when courts find that lawyers
have given them false data, hallucinated data, what have the
penalties been.
Speaker 3 (07:24):
Yeah, they have ranged from just simply admonishing the attorneys
as was in that recent DOJ case you were referring to,
all the way up to heavy fines and sanctions like
over ten thousand dollars or more and being referred to
their state bar for possible discipline. I mean, it's very serious.
Even if you just have a brief with one pit
case in it, you could end up with the most
(07:45):
of your sanctions and you're seeing that in the examples
in the database that we're talking about.
Speaker 2 (07:50):
Last, very quick question, have you ever had a client
to ask you if you use AI in doing your work?
Speaker 3 (07:58):
Not me personally? No?
Speaker 1 (08:00):
Interesting?
Speaker 2 (08:00):
Yeah, I think I think I would like to know that.
But I also think that some people would immediately infer
a negative if their attorney used AI, and I wouldn't.
So I think I think it's a complex question, just
like the use of AI anywhere else. I'll give you
(08:21):
the last nineteen seconds to say anything you want.
Speaker 3 (08:25):
I guess I would just say that, you know, no
matter how you feel about it, AI is here to stay.
We have to learn how to use it properly. It's
a major invention that's changing everything, like the Internet or
you know, another major invention before that, and it has
the power to enhance people in their jobs and make
them better at it. And so I think that we
just need to get through this rough part and figure
(08:45):
things out and understand it and use it properly, and
then everyone will benefit.
Speaker 2 (08:49):
Crawford app will be a partner at Wisner Bomb. Thanks
so much for your time, great conversation, great topic, appreciate it.
Speaker 1 (08:55):
Thank you, Ross.
Speaker 4 (08:57):
I think it's hilarious. That I just realized there was
a chat on that zoom. Did you read that chat? No,
so chat popped up on that zoom meeting and said
Crawford added read dot Ai meeting notes to the meeting.
Read provides AI generated meeting summaries to make meetings more
effective and efficient.
Speaker 2 (09:15):
From time to time, we get these guests and we
log them into the Zoom. We do a lot of
interviews by zoom because first, I like seeing the guests.
We like seeing the guests, but also the audio quality
is so much better than phone. And more and more
and more guests these days are showing up with their
own AI also logging into the Zoom, and it listens
(09:38):
to the conversation, transcribes the conversation, and takes notes for them,
and summarizes the conversation for them.
Speaker 1 (09:45):
That is quite quite a bit of irony.
All right, good morning.
Speaker 2 (00:01):
AI is everywhere doing everything, and it's an amazing tool
and you can use it to make yourself more productive
and to start a company with three people instead of
thirteen people, and you can do all kinds of things.
But AI is not perfect, and when you are involved
in something even vaguely close to a mission critical operation,
(00:21):
you better be very careful when you're using it. And
some lawyers seem to be learning that even though it
seems like a thing they should have learned already. Joining
us to talk about AI hallucinations and other aspects of
AI use in trial and other court proceedings as Crawford Applebee.
Speaker 1 (00:40):
Crawford is a partner at Wisner Baum, a big law firm,
and he joins us on Kowa. Thanks so much for
making time for us.
Speaker 3 (00:48):
Appreciate it, all right, Ross, thanks for having me so, I.
Speaker 1 (00:54):
Was obviously not surprised.
Speaker 2 (00:56):
We've heard a lot of small stories, well maybe they're
not small to someone who's a lawyer, about lawyers using
AI for briefs and things like that where there are
hallucinations involved. And I want you to tell us more
about what's happening. But I didn't know until I was
reading up to talk to you that the actual federal
(01:16):
government got in trouble with federal attorneys doing the same stuff,
So can you please describe what the problem is and
then also.
Speaker 1 (01:25):
What you think the scale of the problem is.
Speaker 3 (01:30):
Yeah, so I think obviously we're only finding out about
the situations where attorneys are being caught by courts or
by opposing counsel. But I think what's going on here
is that AI creates a strong temptation for people to
cut corners. It quickly provides confidence sounding polished answers even
the most complex legal questions, and the words it writes
are very impressive at first glance. And so when you
(01:51):
combine that with how much pressure attorneys are under to
produce quality work as fast as they possibly can, lawyers
will fail to verify what ay I give them, and
then you end up in these situations with fake cases
in court documents.
Speaker 2 (02:05):
Just one quick follow up, how big a problem is it?
Speaker 1 (02:08):
Right?
Speaker 2 (02:08):
You know? Do you think this is a thing that's
happening every day but only being caught once in a while.
Speaker 1 (02:15):
Is this Are we making a lot out.
Speaker 2 (02:18):
Of something that's actually a small problem, or do you
think it's kind of a big problem?
Speaker 3 (02:23):
I think it's a big problem, but it's not going
to be big compared to what's coming. The reason that
it's a big problem now is because of ethical rules
that attorneys have to follow in order to be licensed
and represent people. And as the court pointed out in
that recent DJ case, there's something called a duty of candor,
which means that you can never represent false facts or
(02:44):
law to the court. So when you fail to verify
the law and your briefs and you file it and
your brief contains fake cases, that's a problem with the
duty of candor.
Speaker 4 (02:53):
So when I looked at this AI hallucination database, it
showed over eighteen hundred cases identified so far, a majority.
Speaker 1 (03:00):
Of those cases in the US.
Speaker 4 (03:02):
When you look at these cases, obviously any type of
hallucination is not good. But I almost want to ask,
is part of the case accurate? Is some of the
case accurate?
Speaker 1 (03:16):
Is it all made up?
Speaker 4 (03:17):
What is the situation of when AI is actually interpreting
a case in this situation.
Speaker 3 (03:24):
That's a good point because when people think of hallucinations,
they tend to think of things that are just completely imaginary.
You know, a fake case name with a fake site
that doesn't pull up anything. There's other types of hallucinations too.
I've seen hallucinations where AI will tell you that a
case is quoted for something when it doesn't contain those words,
but it's a real case. AI will tell you that
a case represents a certain rule of law or is
(03:44):
about a certain subject, and then when you read it,
it's not actually about that. So hallucinations can be They
can cover more than just a completely pretend case. They
can cover just misrepresentations of the case that exists. Right.
Speaker 2 (03:57):
And the first thing you said, where it just it's
a case that doesn't exist, that that also happens, right,
So it's that it's that whole range. Now, I don't
blame an attorney necessarily for wanting to use AI as
a as a tool, right, And I probably shouldn't ask
you in public whether you use AI. You can you
(04:19):
can talk about that if you if you want to.
But going back to your duty too, okay, So, but
going back to your duty of candor, obviously you have
to fact check everything that it does. But I wonder,
going back to the way you open this, how much
of this do you think happens? Because there are a
lot of attorneys out there who just aren't very good writers,
(04:41):
and they see that AI writes better than they do.
Speaker 3 (04:45):
I think what you're going to see is that this
happens at all levels. And I think we've seen that,
and like as you pointed out the eighteen hundred odd situations,
you're going to find some of the most prestigious firms
in there, and that's going to continue to happen. But
I think the important thing to understand is that the
cure for AI hallucinations right now is human verification. And
(05:06):
so until the technology gets better and we can find
new ways to check and catch these kinds of things,
it's going to just keep happening. But no matter how
great the output is from the AI, you have to
check it. And people may say, well, that sounds counterproductive
because AI is supposed to be making me more efficient.
Well let's think about it this way. Let's say that
it would have taken you about half an hour to
write a paragraph of something, and AI can write that
(05:28):
instead in about a minute or two. Now all you
have to do is spend about five to ten minutes
fact checking that and making sure the law is real,
and you've just saved yourself. Twenty twenty five minutes, So
you've got to play the long game with AI. You
got to use it appropriately. It's all about shaving off
minutes and at the end of the day, you save
yourself an hour.
Speaker 4 (05:46):
Are there any other safeguards you feel like law firms
should have in place when using AI besides just having
somebody that fact checks. And it's funny that you said
counterproductive because a lot of people think of AI as
eliminating a person and now was like, well, no, you
to make sure you still have a person to check
AI to make sure that it's accurate.
Speaker 3 (06:05):
Yeah. Absolutely, I think human verification of the output, especially
when you're going to do something like file something with
the court or represent something to say a jury. You know,
we're going to see some situations in the future where
there's falsified evidence where a client might submit a AI
created video or image to their attorney, the attorney want no,
They'll file that with the court and they might get caught.
(06:25):
Things like that that are going to start happening more
and more.
Speaker 2 (06:28):
We're talking with Crawford Apple Be, a partner at Wisner Baum.
I'm not an attorney, but I do love reading constitutional law,
Supreme Court cases, federal appellate circuit cases and so on.
And sometimes I'll talk about something on the show and
I'll go to AI and just ask it, has there
been a case about this topic?
Speaker 3 (06:47):
Right?
Speaker 2 (06:48):
And obviously that's a kind of thing that lawyers would
do all the time. And it's not that hard then
to go look up to see whether that's a real case. Now,
AI has never told never hallucinated to me, but I
haven't asked it very many of these questions. But I
can imagine it's saving an immense amount of time.
Speaker 1 (07:05):
We got just over a minute left.
Speaker 2 (07:08):
I imagine the range has been all over the place
on this next question. But when courts find that lawyers
have given them false data, hallucinated data, what have the
penalties been.
Speaker 3 (07:24):
Yeah, they have ranged from just simply admonishing the attorneys
as was in that recent DOJ case you were referring to,
all the way up to heavy fines and sanctions like
over ten thousand dollars or more and being referred to
their state bar for possible discipline. I mean, it's very serious.
Even if you just have a brief with one pit
case in it, you could end up with the most
(07:45):
of your sanctions and you're seeing that in the examples
in the database that we're talking about.
Speaker 2 (07:50):
Last, very quick question, have you ever had a client
to ask you if you use AI in doing your work?
Speaker 3 (07:58):
Not me personally? No?
Speaker 1 (08:00):
Interesting?
Speaker 2 (08:00):
Yeah, I think I think I would like to know that.
But I also think that some people would immediately infer
a negative if their attorney used AI, and I wouldn't.
So I think I think it's a complex question, just
like the use of AI anywhere else. I'll give you
(08:21):
the last nineteen seconds to say anything you want.
Speaker 3 (08:25):
I guess I would just say that, you know, no
matter how you feel about it, AI is here to stay.
We have to learn how to use it properly. It's
a major invention that's changing everything, like the Internet or
you know, another major invention before that, and it has
the power to enhance people in their jobs and make
them better at it. And so I think that we
just need to get through this rough part and figure
(08:45):
things out and understand it and use it properly, and
then everyone will benefit.
Speaker 2 (08:49):
Crawford app will be a partner at Wisner Bomb. Thanks
so much for your time, great conversation, great topic, appreciate it.
Speaker 1 (08:55):
Thank you, Ross.
Speaker 4 (08:57):
I think it's hilarious. That I just realized there was
a chat on that zoom. Did you read that chat? No,
so chat popped up on that zoom meeting and said
Crawford added read dot Ai meeting notes to the meeting.
Read provides AI generated meeting summaries to make meetings more
effective and efficient.
Speaker 2 (09:15):
From time to time, we get these guests and we
log them into the Zoom. We do a lot of
interviews by zoom because first, I like seeing the guests.
We like seeing the guests, but also the audio quality
is so much better than phone. And more and more
and more guests these days are showing up with their
own AI also logging into the Zoom, and it listens
(09:38):
to the conversation, transcribes the conversation, and takes notes for them,
and summarizes the conversation for them.
Speaker 1 (09:45):
That is quite quite a bit of irony.
The Ross Kaminsky Show News
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!