Latest / Coopers' Code / This Month in Legal AI: New Tools for Lawyers, LLM Biases, and Avoiding Brainrot

Transcript

This Month in Legal AI: New Tools for Lawyers, LLM Biases, and Avoiding Brainrot · Coopers' Code

0:00
Howdy, and welcome to the show. In this particular episode, we
0:04
are doing our monthly deep dive into the AI world.
0:07
Here on Coopers Code, we have Sherrod Milanfar.
0:10
Normally. We also have Marshall Cole.
0:12
Marshall is headed out to trial on Monday and so he's taking a
0:16
pass. He will be joining us again next
0:18
month. But without further ado, there's
0:21
been a lot going on in the world of AI.
0:23
We've got a few different things that we want to to bring up and,
0:27
and kind of spitball what's been happening in the AI world and
0:31
how it relates to our legal world.
0:33
So anything that is is is tickling your brain right now in
0:37
terms of cool new things, changes in the law, AI bias,
0:43
what have you any, anywhere you want to start?
0:46
Basically, one of the things that's striking to me in terms
0:49
of AI is how quickly things are changing and how quickly things
0:56
are evolving, right? And so I think one of the things
1:00
which might be beneficial for our listeners or viewers to talk
1:04
about is what AI is and what it's likely to become and how
1:09
that might impact us as practitioners in the law.
1:13
When you think about it, right? Right now, these are called
1:16
large language models. And to me, that's mostly a
1:19
predictive type of model, right? The model looks at what you
1:24
input and then it predicts what the word should be 1 after the
1:28
other. I know what we're seeing is that
1:31
a lot of the top AI executives are on podcasts or giving
1:37
speeches talking about super intelligence, which means that
1:41
this system is not only just predicting, but it's attempting
1:45
to learn and become a thought process that can then change the
1:53
way it responds and maybe even develop some sort of intuition
1:58
that it currently doesn't have. And it'll be interesting to see
2:02
how that plays out in the long run when the systems start
2:06
leapfrogging their current piece of learning and all of a sudden
2:11
their way of thinking make somehow super supersede what
2:15
we're able to do at this point. All throughout there as a a
2:21
competing theory point that came up, a friend of mine was talking
2:26
about something he came across the economist and and how with
2:30
all the computing power that a a a few discrete individuals.
2:36
So we're looking at Elon, Mark, Jeff, Sergei acquiring massive
2:44
amounts of compute. Are we just going to end up with
2:48
their AI world, meaning the Super intelligence that is that
2:52
is potentially evolving from it is being built by individuals
2:56
and and individuals who who have their own agenda or own bias.
3:00
That's an entirely possible scenario, right?
3:04
Is that these 345 or 10 individuals are going to have
3:08
immense power because they get to tell the AI system how it
3:12
learns and why certain things are important and why certain
3:16
things aren't. And even now what we're seeing
3:20
is that when I ask the system to generate a picture of an
3:24
attorney doing XY or Z in a deposition, oftentimes it's a
3:29
Caucasian male attorney doing those things.
3:33
It's rare when it does it on its own where it's a diverse person,
3:38
you know, of backgrounds that's not essentially, you know,
3:43
Caucasian or a certain age range that they the system sees as the
3:50
typical attorney sort of speak, right.
3:52
It's when you tell it to make the attorney or the witness more
3:56
diverse, that's when it comes up with something different.
3:59
So right off the bat you're looking at it, it's
4:02
automatically, it's looking at something and saying, all right,
4:04
it must be a white male attorney in this age range that does this
4:09
type of work. This is a, a good pivot into
4:12
some of some of the pre work that was done before we sat
4:16
down. You, you sent over a, a
4:18
fascinating article, I think by James Mixon that was from the
4:21
daily journal that title what AI learns from us and why that
4:26
could be a legal problem. And and that that starts going
4:30
into some of these biases where massive overview.
4:33
But the institutional biases of say the history of Amazon hiring
4:40
then translates into how AI selects resumes because they can
4:45
the AI sees, oh, people who look like this white men have been
4:50
hired in the past. Therefore, the resumes we call
4:54
as AI are white men. Right.
4:56
And the thing I think it's twofold, right?
4:58
It's one how the AI is programmed to learn and respond.
5:04
The other piece of it is what responsibility do we have in our
5:07
inputs to put things that might essentially level the playing
5:11
field or skew it in a way that it's less of you having blinders
5:16
on, I guess might be the best way to put it.
5:18
And so those are two things that are important.
5:21
But the question is going to be, as you talked about before we
5:24
started, is if there are studies that show that our brains are
5:28
more likely to get lazy with this kind of assist, are we
5:32
going to take the time to do that?
5:34
Or are we just going to say, you know what, just put in the basic
5:37
prompt and take whatever it says?
5:40
Which like you said, there's studies out there and on, you
5:45
know, neurological and otherwise that say the human brain has two
5:50
segments to it, right? The reptilian brain and the
5:53
higher thinking brain and the reptilian brain is the snap
5:56
judgement. And then the higher thinking
5:57
brain is more lazy, but it takes more development to get an
6:03
answer from it. And those are all very
6:05
fascinating things to think about, but it's not something we
6:09
typically do when we're doing day-to-day work, right?
6:13
So we have to be very mindful of how we approach it.
6:18
So you're, you're telling me that as you're sitting there,
6:22
you're not thinking of, of Daniel Kahneman and thinking
6:25
fast and slow and, and, and how you're responding in the moment
6:28
to, to various information input that you decide things.
6:30
Right, only when I'm buying things incessantly online that I
6:33
don't need. So well, as we, as we navigate
6:39
this, it feels like the, the kind of the next thinking point
6:45
to cover would be the MI, the recent MIT study your brain on
6:50
AI, which, which disclose certain things that that could
6:55
be concerning as one goes into meetings where one holds
6:59
strategy sessions related to the use of AI.
7:02
And so for those who haven't come across the study, again, a
7:06
high level overview, no real surprise is if you have a
7:10
thinking partner, your AI tool do all of your thinking for you,
7:14
your brain might start not working out as hard.
7:18
How do you, you Sherrod, see that impacting the work that you
7:22
do, the strategy sessions you may hold with your Co counsel,
7:26
what have you? It it's not an easy thing to be
7:31
mindful of, right? Because I think again, as you
7:35
correctly pointed out, the, you know, the book Thinking Fast and
7:38
Slow, Encantaman and a lot of the other studies show that if
7:41
our minds have the option of doing things the easy way versus
7:48
the hard way, that's what they're going to go with the
7:51
easy way. And just from my own personal
7:53
development, anytime I've been pushed out of my comfort zone,
7:57
I've been able to accomplish things that I never thought
7:59
would be possible, where if I'm sitting in my comfort zone, I
8:03
joke with my kids that your comfort zone is a dead zone,
8:06
right? You're not going to go anywhere
8:07
sitting in your comfort zone. And that's something where we
8:11
have to be very vigilant about it because using AI as a crutch
8:18
is a very different approach than using AI as, you know,
8:22
maybe a platform where you're bouncing ideas off of something.
8:28
So again, it's not an easy thing to do, but it's almost like you
8:33
got to put a posted note on your monitor if you're going to use
8:36
AI to say, hey, don't accept everything that it says and use
8:41
it as a tool, not a crutch. But it's not an easy thing to do
8:48
on a daily basis for sure. And on the accepting side, a
8:52
recent thing that came up in in our office, there was an
8:57
associate who was using AI as a background to digest some, some
9:04
codes. Are there any rules about
9:06
whether 1 needs to submit a notice of appearance in federal
9:11
court if one is going to take a deposition?
9:14
And Claude spat back an answer very emphatically with a no,
9:23
you, you don't need to do that. And it's based on Rule 30
9:29
subsection blank. So associate diligent associate
9:33
goes to that section looks at it says nothing that what Claude
9:38
says and and so associate pushing back says are.
9:41
You sure about that, Claude? And Claude says, Oh yeah, now
9:46
looking deeper, I gave you the wrong answer.
9:49
So it's it's interesting in the in the thinking partner
9:52
department, if you're riffing with it as opposed to relying on
9:57
it, you, you, the the rift may get you to an interesting place,
10:01
but the reliance has the potential of being highly
10:04
troubling. Well, but if you think about it
10:06
right, if that associate had done research, let's say on
10:09
Lexus, Westlaw or whatever platform and come to your office
10:12
and said, Nah, I don't need to do that.
10:15
You may have said, Are you sure? Yes, right.
10:19
And this is, this happens all the time with us and our
10:22
colleagues who we work with, right?
10:24
So what the question is, why wouldn't somebody say that to
10:27
the AI model? Because we know as a fact that
10:32
the first answer is usually not as reliable as the 3rd or 4th 1.
10:37
And that's where whether or not you're being lazy or you're just
10:41
saying, you know what, I don't have enough time.
10:43
I don't, I don't, I'm exhausted. Let me just get get through it
10:47
when in reality what you're buying is a whole hell of a lot
10:50
of trouble on the back end. Again, it's how do you treat it
10:55
and whether or not you just whole cloth except what it says
10:58
as opposed to Are you sure is a perfectly good prompt for AI
11:04
because it forces it to do a deep dive and actually give you
11:08
more. I I've even asked.
11:10
It's like, give me the sites of where you're doing this, where
11:13
you're coming up with this, and I've had it, you know, give me 3
11:16
sites. Two of them were accurate.
11:18
I checked it on Westlaw. One of them was not an accurate
11:23
site. And it was like, all right,
11:25
where'd you get this from? Oh, I'm so sorry I made that up.
11:28
But if the prompt also says don't hallucinate, don't give me
11:33
information that's not verifiable with a source, that
11:38
helps reduce the error rate. And it's how proactive are we?
11:42
Are we using it properly? If you tell an associate, hey,
11:46
just go do this research for me and you don't give them specific
11:49
instructions, they're going to spend 10 hours doing the
11:52
research and it's going to be totally useless to you.
11:54
Same thing with a prompt. How do you utilize it in a way
11:58
that's going to be more accurate?
12:00
And how do you check those answers?
12:02
Possibly cross-platform? I do just want to say that if
12:07
you're engaging with a product, and I'm going to call it a
12:11
product for the purposes of this conversation, and you have to
12:16
specifically ask it not to hallucinate as part of the
12:19
prompting process, that seems like a problem with the product.
12:22
Seems like it should default to not hallucinating.
12:26
Should and reality are two different things, right?
12:29
And if you look at most of the AI models at the bottom of it,
12:33
it says Claude or ChatGPT makes mistakes.
12:37
Make sure you're not accepting everything it said.
12:40
So at the very least they're giving you some sort of a
12:44
warning, right? Even though their whole thing is
12:46
this thing is the best thing since sly spread.
12:49
Yes, it has some incredible functionality, but again, as as
12:54
the end user, am I doing some due diligence to make sure
12:58
knowing that it makes mistakes, right.
13:01
It's not a secret that it makes mistakes.
13:04
I mean, when our associates walk in the door, we, they, they
13:07
don't have a label on their forehead that says humans can
13:10
make mistakes. But same thing with the AI.
13:14
It does have a propensity of being a people pleaser, and I
13:19
think that's its biggest downfall.
13:20
But again, maybe it's a good thing that there are people
13:23
pleasers considering how the intelligence is growing going
13:26
forward. I yeah, yeah, I suppose as you
13:31
very quickly crossover into the dark side of people pleasing
13:33
versus not people pleasing. So although, and, and, and, and
13:40
here we have the opportunity to go a couple different directions
13:43
in terms of the conversation where where I was thinking of
13:46
going was mistakes. And maybe I'll stick with that
13:52
and then I'll come, we can come back to the dark side of, of,
13:55
of, of AI, because there there was something interesting
13:58
recently on that too in the mistakes department.
14:01
One of the things you brought forward, Gerard, before we sat
14:04
down was a, a, a fascinating and challenging way to deal with the
14:11
potential for judicial or research attorney use of AI.
14:17
Do you want to unpack that a little bit?
14:19
Sure. There was recently some news
14:21
reports of a Federal court judge who issued a temporary
14:25
restraining order and then the firm sent a letter pointing out
14:31
to certain inaccuracies and how that order was written and
14:34
attributions that were made to cases that weren't necessarily
14:38
accurate. And there was no evidence at
14:42
this point and no accusation that AI was used.
14:45
But kind of reading into it, you know, this suspicion is that
14:51
there was a research attorney or a federal court clerk who was a
14:56
law student who somehow or post grad who did this research on an
15:02
AI tool and then sent it to the judge who then signed that
15:06
order. The court quickly corrected the
15:09
order and issued A subsequent order.
15:11
But it kind of, you know, the damage is done, right, once that
15:14
order is out there. So there is that potential.
15:19
The question is, as attorneys, how do we call it out if we
15:22
encounter it and do it in a way where you're not being insulting
15:27
to the court, but essentially protecting your clients interest
15:31
in the process? And what I will say is the firm
15:36
handled the corrections so gently that I wasn't sure why
15:40
you were sending this to me as far as a conversation about AI
15:43
was concerned. So with that, we've seen
15:48
lawyers, you know, from a while ago, I think this was a year and
15:51
a half, two years ago when the lawyer submitted briefing on the
15:55
aviation told B6 motion where the the cases were completely
15:59
made-up and the lawyer got in trouble for that.
16:03
And we've seen the expert who cited non existent items.
16:08
I think actually in an AIAPAI intellectual property
16:13
litigation, the the irony was was palpable.
16:17
But when it's coming from the bench, what what was your take
16:24
away in terms of best practices on how to help the bench reach
16:29
the right decision when one comes across what one might
16:32
think are hallucinations or errors in AI analysis that have
16:37
been infected the the order? What I thought was interesting
16:41
was just the way they phrased the the letter, right?
16:45
It wasn't an accusation. It was basically, hey, your
16:49
honor, there was some inconsistencies that we saw that
16:51
we wanted to bring to the court's attention.
16:54
It didn't ask for any specific relief.
16:57
So it was done very well in that context.
17:01
But the other part of it is often times we get orders from
17:04
the court and you read it, you get to the conclusion, then we
17:08
win, then we lose. And maybe we need to double
17:11
check whether or not these orders have any inaccuracies or
17:16
elements that may have caused a problem or, you know,
17:20
essentially the wrong results. And that's something that I
17:23
think we've been taking for granted up to this point.
17:26
And from what I've heard from judges or research attorneys is
17:30
that a lot of the courts are working on AI policies but they
17:35
don't have them in place yet, which is understandable
17:38
considering how quickly the space is moving.
17:41
You're causing me to to click over to one one of the other
17:47
things that that, and this was something I pulled that the
17:51
California Judicial Council adopted new AI rules and, and
17:55
this, this happened since the last time we recorded.
17:57
And this talks about the, the use of generative AI in the
18:04
state's court system. And, and it's not specifically
18:07
on point as far as the judicial use.
18:09
It's, it's more broadly based, but it's, it it talks about how
18:15
courts that are using AI must adopt A use policy by December
18:19
15th of 2025. There are other specific
18:23
requirements related to this. Best thing to do would be to
18:26
take a a quick gander at the the this new rules, but the the
18:33
summary version is that the task force chose not to prohibit the
18:37
use of generative AI just because it's evolving so
18:39
rapidly. And that it is one must disclose
18:43
when AI is being used to create publicly accessible content.
18:47
This next piece is a common theme that I think we've seen
18:50
all through time and that we will see again that users must
18:53
review and correct AI errors, including hallucinations.
18:57
And, and this comes back to, and I think this is your statement,
18:59
Shrod, from one of our conversations in the past that
19:03
in essence, no matter who produces it, the associate AIA
19:08
Co counsel, if you're signing it, you're responsible for it.
19:12
So you better be sure it's right.
19:14
Right. And I mean, look, we're all
19:15
human, right? There is no 100% accuracy on
19:18
anything. But the question really is what
19:22
due diligence are you doing to reduce the potential for error?
19:28
And that's something that I think from a mental standpoint
19:33
can get lost because of the cognitive tendency to go for the
19:38
easy route as opposed to continue to do the thing that
19:42
may cause you a little bit more discomfort, but may save you
19:45
time down the road. The other thing that is
19:49
interesting about that information that you just
19:54
provided is that, all right, if the courts are creating a policy
19:58
by December of 2025, what happens between now and 2025
20:03
that in terms of the AI changing?
20:06
And then are these policies updated every quarter, every six
20:11
months, every year? You know what I mean?
20:12
Because what is effective today may not be effective in six
20:18
months or a year, considering how fast things are evolving.
20:22
And that's where I think the law historically has had a hard time
20:26
keeping up with technology. The law and the legislators have
20:29
had a hard time keeping up with technology across the board.
20:32
And, and I it appears that one of the learnings there is that
20:37
having highly specific rules as opposed to highly specific rules
20:42
don't work. Guidelines and guardrails do so
20:47
when you, when you have comments about being responsible for
20:51
what's being submitted, that you know, that's a guideline
20:54
guardrail that is going to go forward I think forever more no
20:57
matter who produces, even when we have 3D generative AI counsel
21:04
showing up to argue cases, if it's your generative AI
21:09
associate, you're responsible for what that that entity is
21:13
doing, right. And as far as the for the courts
21:16
are concerned, the question then is, all right, do you have this
21:18
AI policy in place? What do you do you just give
21:21
everybody access to an AI model and say have at it?
21:24
Or is there training that comes along with that to help
21:28
implement and make sure that those guardrails have some
21:32
cement holding them in place, Right.
21:35
So that's another issue from a budgetary standpoint that the
21:38
courts are going to have to face.
21:40
This goes back to your comment and and bring forward the
21:46
corrective communication to the court.
21:49
As courts are allowed more access to these tools, 1 is
21:55
going to need to read the analysis and and perhaps drop it
22:01
in zero and AI to see if there's a hallucination there, at least
22:05
as as a first round of of of of of fighting the issue, right?
22:09
Right, yeah. And, and the the reality of it
22:12
is if people are stuck with, you know, we've always used foam
22:16
boards and paper, there are always susceptible to making
22:21
mistakes once they jump into something like this.
22:24
And it's just the matter of go slow and be methodical and
22:30
that'll help you reduce the potential for a problem.
22:35
If you're enjoying what we're doing here with Cooper's code,
22:37
please help us extend our reach by giving us a five star review
22:39
on Apple Podcasts and consider sharing an episode with a
22:42
friend. Additionally, we want to know
22:44
who you want to hear on the show.
22:46
Have an interview suggestion? e-mail us at podcast at Cooper's
22:48
dot law. So I, I mentioned that there,
22:52
there was a, a dark direction we could go with the AI piece.
22:57
And I'll share with you one, one of again since the last time we
23:00
recorded one of the papers that came forward is a paper by
23:06
Kenneth Payne and Baptiste Alou Cross on strategic intelligence
23:10
in large language models. And, and what this was, was they
23:14
had different large language models play a variety of, of, of
23:19
game theory in essence, Prisoner's dilemma for those who
23:23
are familiar with with that particular piece.
23:26
And what they noticed is how different models took on
23:31
different personalities. Newhall's Gemini, I guess was
23:35
considered particularly ruthless in terms of exploiting opponents
23:41
and retaliating. So the take away here is not all
23:46
AI is created equal. And I I think that's something
23:49
that people have have started recognizing why some people use
23:53
one. Why, why some people will run
23:54
the same thing through multiple AI sets.
23:57
I know you've been someone who proponent might not be the right
24:04
word and I don't want to put words in your mouth.
24:06
You, you were the one who brought Claude to my attention
24:09
as being hip and talk to compliant, which for those of us
24:12
who are doing work where there's confidential client information
24:17
that we don't want showing up anywhere else is very important.
24:21
With that as backdrop, have you been playing with other models?
24:24
Have you seen different answers from different models?
24:26
Are there other models that you like for different things and
24:29
and why? So one thing about Claude is
24:31
that it before it was very clear on their website that this is
24:35
HIPAA and Sock 2 compliant. Now it's a, it's a little bit
24:39
more murky in terms of their paperwork.
24:41
So it's not as clear as it used to be.
24:44
So that's something to consider. And possibly, you know, putting
24:48
some safeguards around what you put in there or how you put the
24:50
information in there is important.
24:53
As far as other models are concerned, I've, you know,
24:56
tried, you know, ChatGPT and Gemini and other models.
25:01
And the real question is going to be whether or not, let's say,
25:05
Claude and ChatGPT are going to have the same horsepower as
25:08
Gemini has, considering the computing power that Google has,
25:11
right? I mean, let's be realistic,
25:13
their servers and their computer systems are just far more
25:18
powerful, resourceful and information full for lack of a
25:23
better description than any of the other models.
25:26
And it's going to be interesting to see how that's going to play
25:29
out in the long run. But look, there is definitely a
25:35
dark side like you said to all of this because even just
25:38
looking at it as human beings, the amount of energy the systems
25:41
are sucking up and you know, fossil fuels aren't going to
25:46
save us. You know, with the amount of
25:47
energy that we're using for running searches on these and
25:51
people just kind of do it be and don't think about it.
25:54
But that's definitely going to be a concern as we go along.
25:57
And how these companies are going to handle that issue is
26:02
going to be interesting to see. Fossil fuels water consumption
26:06
is the other one that's been in the news a lot recently in
26:08
terms, I think there's a data center outside of Georgia that
26:12
is consuming 500,000 gallons a day and, and, and impacting that
26:17
community's water supply and the quality of that water.
26:21
And and that's just that's just one data center that opened up
26:25
within the last little while. Right, right.
26:28
And I think Meta just announced the data center that they're
26:31
building. It's essentially a, an AI server
26:35
farm and it's going to be like close to the size of Manhattan
26:38
or something like that. And it, it's incredible what
26:42
resources they're putting in. But the question is going to be
26:44
to what end? And we're kind of looking at
26:48
this very small sliver of AI in the legal field.
26:53
But I does have wider implications that we I don't
26:57
think anybody really knows how it's going to play out.
27:01
It's all going to be fine. We're going to move to Mars and
27:07
actually in in a coming full circle.
27:10
You know, Elon Musk's Grok Grok AI it is is a nod to stranger in
27:17
a strange land where it's a Martian.
27:19
So perhaps that is where we need to go as we destroy this planet
27:22
with server farms. Right, right, right.
27:25
Yeah, since they really have that figured out and dialed in
27:27
already, so. In the more uplifting,
27:33
interesting, cool department, we talked for a second before we
27:37
started about what I've described as the category of
27:41
cool new things. So, and, and as as background,
27:45
one of the the coolest new things in the last year when I
27:49
do an AI presentation that I show is the Google LM drop
27:54
content in and within a few minutes, it turns it into a
27:57
talking head podcast where it's seamless.
28:00
You, you, you can't tell that it's not human.
28:02
That that is something that you showed me.
28:04
I actually use it as a regular research tool.
28:06
When I don't have time to read all the things, I will go ahead
28:12
and drop the content in how to turn it into a, a, a, that
28:17
podcast format so that I can listen to it while I'm doing
28:19
chores around the house. It's just phenomenal.
28:22
So thank you for that. You're welcome.
28:25
But you are the one who seems to keep bringing cool new things
28:29
forward. And so that's that's a launchpad
28:31
for what cool new thing do you have this week, Sharad?
28:34
Let's stay on Google Notebook LM that they've added some
28:40
additional features to that. And one of them is you put, let
28:42
let's say you want to put a case in there and it will give you
28:45
the overview of the case. And it's always fantastic to
28:49
listen to it and then read the case because then it becomes
28:52
much more sticky. The other thing that he can do
28:56
is you can interact with it, right?
28:58
Ask it questions that he can answer.
29:01
But it also has a feature that creates mind maps within that
29:07
particular chat. So it can literally breakdown
29:11
who the parties are, what are the legal issues, what cases
29:14
were cited. And so that in my opinion, is a
29:16
fascinating feature because it's literally dicing up the
29:20
information that you're putting in and giving it back to you and
29:23
bite sized pieces. That's very much digestible, as
29:29
opposed to you slogging through 30 pages first to try to figure
29:33
out if it actually is useful or not.
29:36
And for use cases, are you using this with say single like Cal
29:42
app case or are you using this with like a defenses summary
29:47
judgement motion, the how have you played with that?
29:50
So the way I found it to be most useful to me is let's say there
29:55
was a new case that came down from the California Supreme
29:57
Court basically saying that the statue of limitations doesn't
30:00
apply from the micro statute of one year to a person who was
30:05
injured when an ambulance crashed into them.
30:08
It's actually the two year statute which is the right
30:11
result. So that case came down and I
30:14
just took it and I dumped it into the Google notebook LM and
30:21
it did a podcast for me. And then it broke down the
30:25
various issues in the in the mind map section.
30:31
And now I started reading the case and it's like, Oh yeah, I
30:34
know, I remember this. They talked about that on the
30:37
podcast. One of the things that, that we
30:40
learn and, and I think is important for trial lawyers in
30:44
terms of communicating information was that people have
30:47
to see here, feel process information typically 7 times
30:53
before it actually sticks. And So what I'm hearing from you
30:57
is, is this is a great way to break the information down into
31:01
component parts and, and just make it stickier, more
31:03
digestible. I think that was your word
31:05
actually sticky. Absolutely.
31:06
I mean, look, I'm a huge proponent of audiobooks.
31:09
I used to have a very long commute, but when it came to
31:13
going to work, So I've gone through probably 380 maybe
31:17
audiobooks in my lifetime. And it's an easy way for me to
31:23
learn 1st. And then if I read it, then it's
31:27
like, Oh yeah, my mind now makes the association.
31:31
So it to me, it's a no brainer where if I'm taking that case,
31:35
I'm putting it in there. But sometimes I'll look at it
31:37
like in in the audio version of that case that I was just
31:41
talking about, they kept mispronouncing Micra, right?
31:47
So that was a clear thing that this is not perfection.
31:51
I mean, you know, so it it was a good reminder, but at the same
31:55
time, it provides you a basic understanding of the case.
31:59
And then you can read the case and flag it and highlight it and
32:02
do whatever you do to make it sticky for you.
32:05
But to me, if I have, you know, 15 minutes in the car or if I'm
32:08
on a treadmill or something, it's a no brainer to listen to
32:11
that because it helps me learn better than to just jumping
32:16
right in onto the case and trying to digest it and maybe
32:20
having to read it two or three times to fully grasp exactly
32:24
what's going on with it if I want to truly be proficient in
32:28
how I talk about that case. Fascinating.
32:32
So I, I took us down this path. What I was trying to draw out
32:38
was the some of the video work that you've been playing around
32:42
with, with, with Google and and and and cool new things there.
32:48
Sure. So Google has the VO3VO3 model
32:54
that they have and basically you can go in and put information in
32:58
there and it'll essentially render a video and they're all 8
33:02
seconds videos. So you can use a particular
33:06
platform, I think it's called Flow, to have eight second
33:10
videos back-to-back if you need to.
33:13
But certainly it's something new where you can put in deposition
33:16
transcript, like a blurb of a question or two and it'll render
33:21
the scene and the witness and it'll allow you to basically
33:24
play that in a video. And it does it in pretty short
33:30
order, like 3 to 5 minutes. I think one of the challenges
33:33
and, and, and one of the benefits of, of doing this
33:35
program and, and sharing with people is it is hard to keep up
33:40
with all of the cool new things that we may be able to use
33:45
either as learners or as demonstrators in the cases that
33:49
we're handling. Absolutely.
33:51
And that's why I love having these conversations with you and
33:53
Marshall and other people. And, you know, hopefully people
33:56
who are listening or watching this, if they're doing something
33:59
cool, we'd love to hear about it because all of this stuff is
34:03
only limited by our imagination. And as trial lawyers, if you
34:08
don't have an imagination, you better not be trying cases
34:12
because nobody wants to hear boring garbage coming out of our
34:16
mouths, right? So to me, if I get up in a trial
34:21
and I'm not doing something that essentially captures the
34:24
audience's attention, it's not a good trial.
34:27
And this applies to AI as well as well.
34:31
What do you think should be, can I do this?
34:33
OK, let's try it as ask it and find out.
34:36
And sometimes it doesn't do it very well and you have to tweak
34:39
your prompts, but it does provide you a lot of
34:43
opportunity. Use your imagination to see what
34:46
the possibilities are. Yeah, I really appreciate all
34:49
the the things that you brought being gathered before.
34:52
So we had a a meaningful conversation.
34:55
I'm sorry Marshall couldn't join us.
34:56
I I wish him, he and his client that the best of luck in in
35:00
their trial and look forward to having them join us when we hit
35:03
this session again next month. Yeah.
35:07
Well, thanks for organizing this and running a great
35:10
conversation. I really appreciate it.
35:12
Appreciate having you on. It's always, always a pleasure,
35:15
Sherrod. Thank you.
35:16
As Sherrod said, if you have anything to share, please feel
35:19
free to either drop it in the comments or reach out to at
35:22
podcast at Coopers dot law. If you have things that you want
35:25
us to cover or do a deep dive into, we we'd love to
35:27
incorporate things into the show.
35:29
So for all of you who are listening and enjoying this,
35:33
please give us a five star review on your favorite podcast
35:36
platform of choice. Share it with a friend.
35:39
And to all of you doing justice out there, happy hunting.