Latest / Coopers' Code / Law in the Age of AI: Is the Legal Profession Prepared?

Transcript

Law in the Age of AI: Is the Legal Profession Prepared? · Coopers' Code

0:00
Welcome to Cooper's Code. I'm very excited to be joined
0:04
for this first of what is planned to be a monthly session
0:08
with Sherrod Mellenfar and Marshall Cole.
0:11
We're going to be doing deep dives into AI and, and part of
0:14
the reason we're doing this and doing it in this fashion is
0:16
because things are changing so fast that we felt it was
0:21
important to have conversations about how firms are implementing
0:24
AI, both from a, a proactive stance and and also from a, a
0:28
malpractice perspective, with each of us having our own look
0:33
and, and background in terms of, of examining things.
0:37
Marshall Cole as as listeners know, is an esteemed lawyer
0:40
who's got a practice down in Southern California focusing on
0:43
legal Mal defense shred. Milan Farr is a hitter from
0:47
Walnut Creek who found the light came over from the defense to
0:50
the point of side and and is an early adopter of technology and,
0:54
and everyone knows me. So without further ado, let's go
0:58
ahead and, and turn it over to talking about AI and, and new
1:03
and current events in terms of what's going on there.
1:06
I, I suggested folks either pay attention to the news or come
1:11
forward with any sort of notes in terms of things that are
1:13
going on in their life with that.
1:15
Marshall Sherrod, is is there anything burning for you at, at
1:18
at the top for that? I, I think for me, I'm happy to
1:23
jump in here. What's, what's kind of burning
1:25
is I, I'm seeing in the legal community a divide that's
1:31
probably similar to e-mail came into existence and Westlaw came
1:35
into existence going from books to, to web-based legal research,
1:39
which is, you know, a group of lawyers who are too strongly
1:45
resist using AI in replacing things that lawyers do versus
1:52
probably a younger group of lawyers perhaps that has fully
1:55
embraced it in almost every part of their practice, if not
1:59
practice, if not every part of their practice.
2:02
And I think it's really, really interesting to see people sort
2:07
of clinging on to the past and not recognizing that, you know,
2:11
technology is always going to be there, it's always going to
2:14
change, It's always going to evolve.
2:15
And we as lawyers have to evolve with it or we're just going to
2:19
be left behind. And so when I was reading these
2:22
comments, I just thought it was very interesting that, you know,
2:25
some people would say nothing can replace, you know, the, the,
2:28
the advocacy of a lawyer and what a lawyer does and how a
2:32
lawyer writes and thinks. And while much of that is true,
2:36
I like to characterize AI as a tool, as I've kind of used that
2:40
word before as sort of another weapon, if you will, that you
2:44
can have in your arsenal to help advocate on behalf of your
2:47
client. So I just thought that that was
2:50
just a really dated way of looking at things when I was
2:54
reading these emails going back and forth.
2:57
And, and also, not surprisingly, some of them would say, oh, a
3:02
lawyer was using Westlaw and then do the AI based Westlaw and
3:08
then cited cases in a brief and the court found that the cases
3:12
didn't exist because it was quote, hallucinating.
3:15
Well, I think there was a number of errors within whatever was
3:19
conveyed to the person writing that e-mail, because if you do
3:22
your due diligence of using AI and use a closed source platform
3:27
versus an open source platform, you don't have hallucination
3:31
problems. And so when I saw someone saying
3:34
Westlaw created hallucinations, I thought that was really
3:38
troubling because really Westlaw shouldn't then can't create
3:42
hallucinations because it's a closed source platform.
3:45
They've only fed their model with, you know, accurate data,
3:49
the cases, the treatises that they already have.
3:52
So it's actually extremely reliable to use.
3:56
But of course, I always recommend double checking
3:57
anyway. So I just think there's a lot of
4:00
people talking about it who don't have the correct
4:03
information, are throwing around terms that they don't really
4:07
even understand. And again, just have this way of
4:10
looking at it that it's dangerous and going to replace
4:13
US lawyers and everything we do. And you know, I think shocker
4:18
alert, it is gonna replace a lot of things we do.
4:20
But I think for the listeners here, start adapting, start
4:25
using it, Start learning how to use it, Start learning how it's
4:28
being used against you so that you can be a better lawyer
4:32
'cause it's not going away. e-mail didn't go away.
4:34
Westlaw didn't go away. It only got better.
4:36
It only made our lives easier and more complicated at the same
4:40
time, and I think AI is doing that like every day right now.
4:45
So I'll kick it to Sherrod with kind of a, a, a pushback on what
4:51
you just said, Marshall, which is, you know, I've, I've come
4:54
across lawyers who have said things akin to, well, jeez, you
4:59
know, I, I, I expect my associates to be writing things
5:04
themselves. There's a difference between AI
5:07
drafting and an associate drafting.
5:09
And I want my work to be done by humans.
5:12
And Sherrod, what's, what are your thoughts on that?
5:16
I mean, let me arrest one thing that Marshall said.
5:19
I think the, the perception that people have about the closed
5:24
loop systems hallucinating, I, I think part of it comes from this
5:29
case that just came down at the Lacey versus state State Farm
5:33
case where two law firms, you know, were essentially using AI
5:40
and they had I think 4 out of 20 sites were hallucinated.
5:45
And they mentioned that that came out of Co counsel and West
5:49
law's precision platform. So it so there is that piece of
5:54
it. There are some isolated
5:56
incidents where that's happened. I don't know how accurate this
5:58
is because they also use ChatGPT or Gemini in that research,
6:03
which they but I think the sanctions in that case appear to
6:06
have come from the lawyers not being as forthright the first
6:10
time they were given the opportunity to to fess up to
6:13
their mistake. And but there is something to be
6:17
said about learning how to write and being able to do that.
6:20
But at the same time, this dialogue that's going back and
6:24
forth between some of the people who are resisting this so hard
6:29
reminds me of the comic where there is a, you know, these
6:34
Cavemen pushing a cart that has wheels that are square.
6:37
And then somebody with a round wheel is standing there saying,
6:40
hey, what do you think of this? And they're like, oh, we're too
6:42
busy, right. And this is the same pushback
6:46
that I got when I originally started trying cases using an
6:49
iPad was we've done this for 40 years the same way we use foam
6:54
boards and butcher paper. I, you know, my, I have a
6:57
handwriting that no one can read, you know, So to me, to use
7:01
a foam board of butcher paper is just idiotic.
7:05
So in that context, I think we have to look at it from a big
7:08
picture of all right. Why are they resisting?
7:10
Because they feel that this is how they've succeeded in the
7:14
past. But at the same time, those same
7:16
lawyers who used to go to the stack and look at the like cards
7:22
in the library to find their cases are now using Westlaw.
7:26
So why not have, you know, the opportunity or give it a chance
7:31
for the next step to occur? But that's, you know, I've had
7:34
this conversation with many lawyers who are older who are so
7:39
resistant, and they think that our abilities to think
7:42
critically is going to go away, which is not a a totally
7:46
baseless argument because there's some research out there
7:50
that says that, you know, as humans, we're going to be lazy
7:53
if we're given the opportunity to be lazy.
7:56
So on the lazy side, when somebody does decide, OK, well,
8:03
yes, I'll, I'll start using AI to help me with some of the work
8:07
that I'm doing, Should somebody take the first answer that AI is
8:11
giving them? Marshall, Sherrod, either one of
8:14
you. And and if?
8:16
If no, why not? I well, I think it depends
8:20
really actually if you're using it from like a creative
8:23
perspective on developing case themes and things like that
8:27
where you're just using it for creative purposes, you
8:31
presumably could use the first answer, but I always would say
8:35
no in the sense that you can always do better with it.
8:38
Like you can always ask it to follow up, you know, give me
8:41
more options, do more, do less however you want it.
8:44
And I'm just kind of talking about sort of the general
8:46
ChatGPT model with things like that.
8:49
But obviously I think for the most part, you should not just
8:53
accept it and then put it on paper if you're going to be
8:56
using it. But if you are using it for
8:58
cases, which is isn't recommended, at least by me, you
9:02
better check those cases to make sure that they actually do
9:05
exist, which is pretty easy to do.
9:07
Most people probably have Westlaw accounts and so someone
9:11
might find that it's really quick to research through
9:14
ChatGPT and then just confirm that the case sites are accurate
9:17
and the case actually says what you're going to cite it for.
9:21
And then I think it becomes a little bit different.
9:23
So I don't think it's as much like laziness as much as it
9:28
really allows you to do more. I think it opens doors versus
9:33
I'm sure there are people that will view it as a way to, you
9:37
know, work less and be on the beach more perhaps.
9:40
But I think there's less of those people and more people who
9:44
are viewing it as an opportunity to, you know, be a better lawyer
9:49
and help their clients of new in different ways and be more
9:52
efficient. And just 'cause you're more
9:54
efficient doesn't mean you're lazy.
9:58
Yeah. And I think a lot of times, and
10:00
Marshall can probably speak on this point a lot better than I
10:03
ever will, is lawyers get into trouble when they see their jobs
10:07
as a burden as opposed to an opportunity to help people,
10:11
right? When you think discovery, Oh my
10:13
God, why do I have to do discovery or whatever?
10:15
You're more likely to take the shortcut.
10:17
And if you look at it and say, look, this is just one component
10:21
of a bigger picture and how do I deal with it in an appropriate
10:24
way so that the outcome at the end is something that's
10:27
favorable for my client? And again, if a first year
10:32
associate walks in and hands me a, you know, a piece of work or
10:35
they e-mail it to me, I don't say we'll go ahead and file that
10:38
we have to take the next step. And why would we do it when it's
10:42
a machine that's generating it as opposed to a, a human?
10:47
It just is this the step that we have to take?
10:50
And when people are faced with time crunches, that's when they
10:55
seem to go ahead and take the shortcut.
10:57
That then backfires. So in scratch that lost my train
11:09
of thought. Sorry, rebooting.
11:23
So one of the concepts that a fellow I, I, I ride with brought
11:31
to me when we're talking about AI and, and writing and then
11:37
digesting information is what, what is known as the dead
11:40
Internet theory. The idea that content is being
11:43
created by machines for the Internet that is then being
11:46
consumed by machines on the Internet and it's application.
11:49
Then in say the courtroom briefing is 80% of briefing is
11:56
built by AI, fine-tuned by a lawyer submitted to a the
12:01
opposing side. The opposing side then uses
12:03
briefing that is 80% done by AI, and then the judge or research
12:08
attorney is taking these, putting them together, dropping
12:11
them in to be digested to figure out what the answer is and at
12:16
what point, you know, where's the line where we're not
12:19
actually lawyering anymore and it's all in essence being
12:22
predetermined by tokens? I think we're a long way from
12:29
that. But not to say that it's not
12:32
possible that we will get there. I think every court at some
12:38
point in the next few years is gonna have some type of AI doing
12:43
something in the courtroom, whether that's just helping them
12:46
manage their caseload. I mean, the Los Angeles Superior
12:49
Court judges have hundreds and hundreds of cases at any given
12:52
time. I'm sure there are plenty of
12:55
ways to make them more efficient with how they manage their cases
12:58
and get cases to trial quicker because right now it's manually
13:03
managed and that's just not the most efficient way to go about
13:06
it. There's a, a really interesting
13:09
new platform called Garfield dot AI, which is I believe it's in
13:16
the United Kingdom where they're now allowing you to basically
13:22
use this to recover this, this platform to recover deaths of
13:27
under $10,000. And it will do the whole thing
13:29
for you. You can send it'll send a letter
13:32
for you. So you up you load the invoice,
13:34
then it says, OK, first thing you need to do is send a letter.
13:37
Then it sends a letter. If there's no response to the
13:39
letter within the amount of time provided that it gives you the
13:42
next step and goes and does the whole process for you, which
13:47
allows their sort of small claims system to be that much
13:51
more efficiently handled because it's basically automated at this
13:54
point in time. And I think there will be, I
13:59
think it's not long before there are various courts here
14:03
throughout the United States that adopt certain ways for
14:05
people to get help easier and, you know, allow more people to
14:11
have access to the courts through some of these platforms.
14:15
Yeah. I think as it stands right now,
14:17
having discussed the same issue with some, you know, research
14:21
attorneys and one in particular who was a head research attorney
14:26
at one of the appellate divisions, the feedback that
14:29
I've gotten is that at this point, the courts, because of
14:32
their budgets, don't have the money to really invest in the AI
14:37
that's, you know, dedicated to legal, for example.
14:40
So right now you don't have that issue where the court is fully
14:44
reliant on that. But, you know, the scary part is
14:47
that I think they're using some of the cheaper, you know, open
14:50
models, which then has repercussions, you know, of its
14:55
own. But I think at the end of the
14:58
day, this is something that's happening.
15:00
And the question is, how do we adapt to it to be the most
15:03
effective advocates for a client, as opposed to sitting
15:07
and complaining about it and watching the wave, you know,
15:09
wash us out. Well, and and and I think you
15:13
raise a, an interesting point. If the courts don't have the
15:17
resources, then the path of least resistance is dropping.
15:21
You know, the moving papers, the opposition, the reply brief into
15:25
ChatGPT or Claude and seeing what the response is.
15:31
And if you're a young clerk at or intern at the the courthouse
15:39
where you're being asked to come up with a memo for the judge,
15:43
that seems like a a path of least resistance for
15:46
technologically savvy people. Yeah, I agree.
15:51
And also think about court reporting.
15:53
I mean, there's A at least down here, there's a crazy shortage
15:57
of court reporters to get an in person court reporter to attend
16:01
a deposition is almost impossible at some points in
16:04
time. And so it's, it's remote.
16:06
But think about just a basic AI model that you turn on and
16:10
transcribes everything the deponent says and just like a
16:14
normal deposition and can identify when it didn't hear a
16:18
particular word you put in spelling and it becomes a
16:22
certified transcript in some way.
16:24
I mean, I think that's going to happen because there just aren't
16:28
enough reporters that, you know, I think at one point in time the
16:31
Los Angeles Superior Court was offering like a huge incentive
16:34
bonus program to go to court reporting school and, and, and
16:37
become a certified reporter to be available for the courts.
16:40
And it, I don't think it was as successful as they thought it
16:43
might be. So a real easy solution to that
16:46
is to have an AI model that works just like a court
16:50
reporter. And so it's, I mean, we
16:52
obviously already have recording devices.
16:54
So we're 98% of the way there. It's just getting it to be
16:59
produced in a certified way that's reliable, that can then
17:02
be prepared into a transcript and given to a court.
17:07
I think we're at A, at least from my conversations with court
17:10
reporters, we're at a very precarious inflection point.
17:16
And and precarious inflection point is do you want to be an
17:19
apprentice buggy whip maker learning your skills as
17:24
automobiles are being rolled out on the road?
17:26
Or are you being, you know, doing high level training for a
17:30
job that is going to be probably these legislation sometime in
17:34
the next 10 years absent the court report lobby being strong
17:39
enough to hold it back, doing exactly what you're talking
17:41
about, Marshall? Well, I mean, even and also
17:44
like, I don't know about you guys, but some of the depot
17:46
companies I use now already give me an AI generated depot summary
17:51
that's free. And so right there you don't
17:55
even ask for it. They've eliminated what would
17:59
generally be, at least on my end, a paralegal task perhaps or
18:02
a first year lawyer task, which is a really important task as a
18:06
lawyer to learn right, to read a deposition and to summarize it,
18:09
to pull the key things out. Well, now you have that.
18:12
And I think they generally do a pretty good job of the depos
18:15
I've taken. You know, you can tell whether
18:17
they pull that important information.
18:19
And so no longer do I need to have someone do that work if
18:25
that deposition summary is created.
18:27
But it's also a missed opportunity to train a lawyer or
18:29
a paralegal and how to do that. And so I think that's one area
18:33
where I see just kind of going away based on those AI generated
18:37
summaries. Yeah.
18:39
And I think that raises an important point, and that is how
18:42
do we maintain the building blocks of the skills that we
18:46
need to know in order to be able to do the various jobs that the
18:50
three of us do? When you don't have that
18:54
ability, you know, you don't have that as a training
18:56
opportunity. So we just have to be creative
18:59
and how do we bring up the younger lawyers in our firm to
19:03
be able to do those things? If they can simply dump that
19:07
transcript into AI and have it spit out the the summary and not
19:14
have to go through it themselves and pick up the you know, the
19:18
nuances of how do you ask a question and you set it up so
19:22
you get a usable transcript or a sound bite at the end of the
19:26
day. And that's also where I think
19:28
motivations come into play that differ between the plaintiff
19:31
side and the defense side on, you know, on the plaintiff side,
19:35
you guys contingency work. The the time it takes to train a
19:40
lawyer to do that is not necessarily time you get paid
19:43
for versus a defense side lawyer where we bill at, you know, .1
19:48
at a time. And you know, those were ways
19:52
not only to train a lawyer, but you know what, they're necessary
19:55
things. But he also generated revenue
19:57
doing it. And now obviously, if I get a
19:59
deposition summary from AI, I'm not billing a client for that.
20:02
It's received and that's that. So, you know, arguably the
20:06
client benefits from it if it's accurate, right?
20:09
If it's not, then they didn't and it doesn't replace that.
20:12
So I think there's this give and take on depending on which side
20:15
of the aisle you sit on and how you get paid.
20:19
And it's going to really dictate how you want to train people.
20:22
Marshall, I'm curious since you're you do defense work.
20:24
My perception from my years of doing defense work is that the
20:28
bills get cut significantly as it is.
20:31
And now you have AI reviewing your bills and now you have an
20:36
AI generated work product that if you give to a first year
20:40
associate to quality control, you're probably not going to get
20:44
paid for that, for that work. And to me, that creates a lot of
20:49
pressure on the defense firms and that may reduce, you know,
20:54
some of the quality of how that translates into the
20:58
representation of the the clients, especially the ones who
21:02
are insured and the insurance carriers are paying for.
21:05
Yeah, no, I I completely agree. And obviously our bills are
21:09
heavily scrutinized by an I don't know if it's probably AI
21:13
now, but it will used to just be a computer algorithm, right.
21:15
If your, if your bills had a specific word that hit a red
21:19
flag, it was automatically going to be cut to zero and then you
21:23
would have to appeal it. I know both of you guys are just
21:25
like loving being on the plaintiff side right now.
21:28
And so, but if that's how it is, the person, the adjuster on the
21:32
file that we're working with, who knows the case, who knows
21:35
why two people had to attend a mediation doesn't even review
21:39
the bills necessarily. And so they don't know that your
21:43
time got cut from, you know, 5 hours to 0 hours because you
21:47
arguably didn't comply with the billing guidelines even though
21:49
you did. But yeah, I know I think it's
21:52
going to become a nightmare, especially when I think on that
21:56
side of the industry, the claims side, I think a lot of their
22:01
analysis and what they do is going to be taken over by AI
22:06
models who can objectively evaluate a demand letter.
22:12
I mean, there's there's even up out is out there which which for
22:16
you all, you know, basically takes information you give it
22:20
and creates this beautiful demand letter with damages and
22:23
tells you how valuable your case is, etcetera.
22:25
Well, I think on the carrier side, they're going to have
22:27
their own AI model that's going to take that AI generated miles
22:30
to your point earlier, take that AI generated demand letter and
22:34
say here's why it's wrong. Now, I think if the system's
22:38
working perfectly, it should just objectively prepare a
22:42
demand letter and then the claims AI model should
22:45
objectively evaluate it. And you'd think the case would
22:48
settle, right? Like if it's all working
22:50
perfectly, there would be a number that no one could really
22:54
complain about. Now, of course, plaintiffs
22:56
lawyers are probably going to disagree or what, however it
22:59
goes. But if if it all works
23:01
perfectly, you would think that everything just kind of lines
23:03
up. Well, that's what I there's two
23:09
points that I want to make on that.
23:11
One of them is that the, the approach that you just talked
23:14
about really erodes. When I started doing defense
23:18
work, there was a relationship, right?
23:19
I knew the adjusters, I knew the claims reps, I knew their
23:23
bosses. And that relationship really
23:26
mattered. And as time went on, it seems
23:28
like it was less and less of a factor in terms of whether or
23:31
not you were getting work for a particular company, right?
23:34
I think that's going to erode a lot of that the more AI gets
23:38
involved. And then the second piece of it
23:41
is, I think historically, carriers have had a very hard
23:44
time with the loss of, you know, human losses, right?
23:49
For instance, you know, what's emotional distress, pain,
23:51
suffering, all that. And it'll be interesting to see
23:55
how they program their, their AI to account for what jurors will
23:59
do in a courtroom. Because that's the biggest
24:02
disconnect is, well, the jurors are never going to do that.
24:05
And then all of a sudden they're looking down at a massive
24:08
verdict and they're scratching their head about how did we get
24:11
here? And that'll be interesting
24:13
because ultimately the input in terms of programming and how
24:18
that AI learns will determine how it evaluates and spits out
24:23
the final outcome in its evaluation.
24:26
Yeah. And I think it's equally true on
24:27
the plaintiff side because you know, let's say you use even up
24:30
or whatever program to create this great demand letter, but
24:34
you're a phenomenal trial lawyer, right?
24:36
And a phenomenal trial lawyer creates value in a case far
24:39
above whatever that demand letter is going to spit out.
24:43
It just doesn't know that metric.
24:45
And so I think that's where a, you certainly can't just be
24:49
pumping these letters out without looking at them because
24:51
I think on the plaintiff side, probably undervaluing your case,
24:54
depending on the type of lawyer you are.
24:56
And if it's on the claim side, you've got to really
24:59
realistically account for who the plaintiff's lawyer is,
25:02
right? That that's a person who can
25:04
really hit a home run and tried however many cases or whatever,
25:09
you know? And so I think those are things
25:10
that aren't accounted for, but probably can get overlooked if
25:15
you're just being quote lazy, I guess, and, you know, trying to
25:19
have AI do everything for you. I, I think one of the areas that
25:24
I'm really interested to see in terms of data sets and, and how
25:27
they develop, you know, in, in times past there was Colossus,
25:33
you know, evaluations of cases based on algorithmic
25:36
information. As we have large language models
25:41
and we start having the ability to gather more and more
25:44
information, to have that information analysed in
25:46
different ways, having a database that that knows what
25:50
cases are settling for, knows what verdicts are coming in at,
25:54
can account for, you know, rating a trial error in terms of
25:58
the quality on the firm. Theoretically, you know, you
26:02
almost have a limited approaching Infinity of, of, of
26:05
perfect information for valuation on cases.
26:09
In theory it won't happen, but I see that being rolled out to
26:14
insurance companies as as a data set where they start putting
26:17
values on cases and sticking to it.
26:19
Well, there's actually a defense firm that's now, you know,
26:23
coming out with an AI to prevent nuclear ver verdicts.
26:27
And, oh, you know, but they've since been hit with several
26:30
nuclear verdicts. So I'm like, maybe your AI isn't
26:33
working as well as you thought it?
26:35
Would we just need to adjust the model?
26:37
Right, so. There's.
26:40
There's that piece of it. It's like, all right, well,
26:41
you're gonna to guarantee a result.
26:43
And that's the differentiating factor, right, Is no one can
26:46
guarantee a result for their client.
26:48
But again, I always really look at verdicts in similar cases,
26:54
but I discount them because, you know, every lawyer is different,
26:57
every client is different. And if you switch out two jurors
27:01
or three jurors out of that case, maybe you have a different
27:04
result. But and the most valuable tool I
27:08
think we have is the conversations we have with
27:11
normal human beings who are looking at, you know, the the
27:15
hearing, what we're telling them and saying, you know what, that
27:17
doesn't sound right or yeah, that's a great argument.
27:20
Go ahead and make it. Well, as, as we touch on
27:25
different topics and, and we stay topical and, and we also
27:29
stay mindful of time because the commitment on these were, they
27:32
were not going to be deep dives are going to be in the, in the
27:33
30 to 40 minute range. I did want to bring up the, the
27:37
recent research from Anthropic where they were looking at and,
27:42
and these were sandboxed testing, but meaning these were
27:46
not real world evolutions, but where they had put AI agents in
27:52
place. And, and these are different
27:54
than what I think a lot of lawyers are using these days.
27:56
It's not us dropping information into Claude or Westlaw or some
28:02
other piece, but where an agent is allowed to roam free within a
28:07
a server system and and review e-mail and and cobble to gather
28:10
information. And the agents that were allowed
28:14
to their own autonomy when threatened with being shut down,
28:19
found the emails amongst people in in the data sets and use
28:24
those to blackmail the folks who were threatening to shut them
28:27
down. And, and I thought that was an
28:29
interesting thing just for us to talk about both from, from a
28:32
tort perspective as as people start the potential of deploying
28:37
these, but also in terms of, you know, are are we ready for prime
28:40
time with AI in terms of agents being deployed into our, our
28:45
systems? Probably not right I I'm.
28:55
Sorry, Sharon, I'm afraid you can't do that.
28:58
Well, I mean, you look at, for instance, Google has this thing
29:00
like the, you know, the the agent will call and make you an
29:03
appointment for, you know, a haircut or whatever, right?
29:06
I mean, and that's seems pretty neat when they demonstrated,
29:10
right? But then you have this wild card
29:12
of all right, I'm going to use your emails to suggest to your
29:15
boss that you're having an affair with one of your
29:18
colleagues. I mean, that's what the
29:19
anthropic model did, right? And that to me is scary.
29:24
And I think Miles and I talked about this.
29:26
I always say please and thank you in my prompts.
29:30
And, you know, Sam Altman, the, you know, head of open AI said
29:34
it's costing them hundreds of millions of dollars where people
29:37
are doing that. So out of curiosity, I went on
29:40
and I asked ChatGPT, do you care if I say please or thank you?
29:44
It said, well, I don't care. I do notice.
29:46
And it sets a nice tone for our conversations.
29:50
And I'm thinking that's a veiled threat if I ever heard one.
29:54
So all all things equal, Sandy, you're going to lose $100
29:57
million a year, 'cause I'm going to keep saying please and thank
30:00
you. That's yeah, I, I remember you
30:06
bring that up in a previous conversation and and it is
30:10
certainly is an an interesting piece to to to think about
30:13
Marshall. I don't have much to add on that
30:17
I'm not, excuse me, I'm not as familiar with that that
30:20
research. Understood, understood.
30:23
I think that the other area and the reason I bring up the tort
30:27
side of this is as with every technology that threatens our
30:33
future as lawyers in say the personal injury world, it
30:38
appears that there may be new areas where there may be
30:42
opportunities and, and that is, you know, coding errors,
30:46
malfeasance on behalf of agents and, and the deployment of them.
30:50
I I see areas where people certainly have the opportunity
30:53
for practice as far as that's. Concerned well what about I
30:56
believe there was a deceased victim I think it was in a
31:01
criminal case who gave a victim an impact statement.
31:07
It was an AI generated victim impact statement from the
31:10
deceased. And so, you know, I, I think
31:14
there's on the plaintiff side, a lot of opportunity for real life
31:22
reconstruction. You know, like you all hire
31:25
accident reconstructionists and, and the technology's pretty
31:27
unbelievable how, how they can do things and just make it like
31:31
a, a multi $1,000,000 motion picture.
31:34
Now just imagine adding AI tools to that and creating, you know,
31:39
the actual people who were injured and put them back in the
31:43
car. Not just like a, you know, like
31:46
a robot or whatever it is like the literal person.
31:48
Think about how effective that will be at, you know, capturing
31:54
the jury's attention and making it even more emotional because
31:59
they can see the faces and the horror that that person went
32:02
through with respect, with respect to whatever it is.
32:05
So I think, you know, someone's going to get there and then it's
32:08
going to try. Someone's going to try to get it
32:10
in, you know, into their closing argument or into their cross
32:13
with their expert and it's going to get denied.
32:15
It's going to get denied. And then some judge is going to
32:17
allow it and then the door will be come open to this type of
32:21
thing because it's going to be the new normal.
32:24
There's there's no longer going to be a need for just a faceless
32:27
person in your video, you're going to have the actual person.
32:33
And I think that's going to be very, very powerful.
32:36
Yeah. And that that's a great point
32:38
because now we're seeing AI work.
32:40
But you know, I don't know if you've seen these ads where you
32:42
give it a picture and then it animates that picture and it
32:46
looks exactly like the people who are in the picture.
32:49
So that's not out of, I mean it's not out of the reality that
32:54
you're talking about. And at the end of the day, you
32:56
look at the Google system that just came online and for like
33:00
140 bucks a month, I think you can basically create AI movies.
33:06
I mean, that's an incredible thing.
33:09
And from and of an entry standpoint, if I can get an
33:11
expert to say, yeah, this is accurate in terms of what
33:14
happened it, it's at the very least the demonstrative that the
33:17
jury gets to see. I agree.
33:21
And so I think those are, and maybe someone's even tried
33:25
already. You know, we don't know, it
33:26
hasn't been reported, but I do know there was an article where
33:30
the the victim, he, you know, gave his statement.
33:34
And imagine how powerful that is for a judge sitting there when
33:38
you're about to sentence someone who committed a crime and now
33:40
you're looking at the person who died giving a statement.
33:46
It's pretty compelling stuff. So as we as we get into short
33:52
strokes for this episode, I, I did want to circle back to the
33:57
idea of losing what I would refer to as the the wax on wax
34:01
off exercises for clerks and associates.
34:04
The idea of of how doing lots of deposition summaries helps one
34:10
understand how to take a good deposition.
34:13
Do you, Sharad, do you Marshall, have any thoughts on how you're
34:17
going to work with your young folks to continue their skills
34:23
and stay current? Meaning we don't teach people
34:26
how to use a typewriter anymore because it's really not
34:29
necessary, yet people still know how to type is.
34:34
Is there a similar analogy or a similar way that you've been
34:36
approaching the training of younger people taking into
34:39
account the these new AI tools that are are coming out?
34:44
From my view, it's still new enough.
34:47
And again, this is where plaintiff versus defense
34:50
differs. The incentive isn't quite there
34:53
for me to adopt AI across the board.
34:57
You know, I think law firms are certainly in interviews with new
35:02
associates. I because I, it's been asked to
35:04
me is how are you using AI? How have you adopted into your
35:08
practice? And 9 out of 10 times it
35:11
happens, they bring up a, a platform that I've never heard
35:14
of 'cause there are just so many, you know, and it's like,
35:17
oh, I haven't heard of that. I go and look at it and it's
35:19
interesting or it's not, but they're asking about it.
35:22
So they're already aware that it's going to be a part of their
35:25
practice. So for me, it's just still, you
35:29
know, I allow it to be used for creative purposes, developing
35:33
case themes, you know, deposition questions on an
35:37
expert that you aren't really familiar with, things like that.
35:40
So I haven't allowed it to replace the true legal work that
35:44
gets done here. Even discovery responses, we
35:48
don't, there's plenty of platforms that that streamline
35:51
that. But for our purposes, we're not
35:54
there yet. But if I did get there, I would
35:56
only allow people to use that who were experienced lawyers
36:01
because then they know what the right discovery response should
36:07
be. You know, sure, it got created
36:09
in 10 seconds, but a first year lawyer doesn't know if what was
36:12
created is actually what should be there, right?
36:15
If you got to know how to check, if you're going to check boxes
36:17
to insert objections, you need to know which objections to, to
36:21
check the boxes. So only an experienced lawyer is
36:23
going to know those things and then your safeguards are already
36:27
working for you where a younger lawyer can certainly learn how
36:30
to use it, but they really need to understand, you know,
36:33
objections or whatever it is, the content, what you put in a
36:36
response, what you don't, how it's worded, how much
36:39
information you want to put in, you know, maybe you want to be a
36:42
little more cagey. There's all kinds of strategy,
36:44
as we know, that goes into how a defense lawyer responds to
36:48
discovery. Basically just give them
36:49
nothing, you know, and, and per pound, 490 special
36:55
interrogatories. So that's that's kind of how I
36:59
do it. Thank you, Marshall.
37:02
What? What about you, Sharad?
37:04
I mean, I think what's important to keep in mind is training
37:08
people to understand the why and that helps them know the how,
37:15
right? And so why are we arguing about
37:18
this piece of evidence? Why is it important?
37:21
How do we tell the story? Why are we telling the story in
37:24
this sequence? And there's a lot of things you
37:27
can do even outside of our offices, right?
37:31
They can go to trial training programs, they can go to
37:34
storytelling classes, improv classes, acting classes.
37:39
I've read an evidence book 3 1/2 times and I'm going through it
37:43
for the, you know, I'm about to finish it, that I'm halfway
37:46
through it, right? Why?
37:47
Because I don't need to know the exact code section, but if I can
37:53
understand the basics of why that code section is there, then
37:57
I can argue out of any evidentiary issue that I need to
38:00
do. And I think like we've talked
38:02
about before, Miles, is that people need to take an active
38:05
role in their own development as well, right?
38:08
So we want to encourage our associates to be interested in
38:11
their own development, learn skills, ask questions.
38:15
And if you're going to come to me with a question, I'd like you
38:17
to be able to give me your solution 1st, and then we'll
38:20
work through it and see if that's what we should do.
38:25
That's Jenny. Well, anything else you think we
38:28
should cover today? For me, I think it's a nice
38:33
jumping off point to start. You know what's going to be an
38:36
ongoing and collaborative discussion as topics come up?
38:39
Because they just literally come up every day.
38:41
I mean, every day I get an e-mail about a new Cle for AI
38:45
something, so it's couldn't be more relevant.
38:49
Well, I'm looking forward to having this journey with the two
38:52
of you. So, Mike, go ahead.
38:54
Thanks for having us. Yeah.
38:56
And likewise, with all of you listeners out there, if you have
38:59
any questions, comments, feedback, your own thoughts on
39:03
AI and its implications, please reach out at podcast at Coopers
39:06
dot Law so we can address it. Like what we're doing?
39:09
Leave us a five star review on your podcast platform of choice
39:12
to all of you doing justice out there.
39:14
Happy hunting.