Analyst Talk - Christopher Cruz - From Analyst to Technologist to AI Leader

1
00:00:00,000 --> 00:00:01,830
Welcome to Analyst Talk with Jason Elder.
2
00:00:01,920 --> 00:00:04,800
It's like coffee with an analyst,
or it could be whiskey with an
3
00:00:04,800 --> 00:00:08,070
analyst reading a spreadsheet,
linking crime events, identifying a
4
00:00:08,070 --> 00:00:11,640
series, and getting the latest scoop
on association news and training.
5
00:00:11,730 --> 00:00:14,670
So please don't beat that analyst
and join us as we define the law
6
00:00:14,670 --> 00:00:16,170
enforcement analysis profession.
7
00:00:16,350 --> 00:00:17,160
One episode at a time.
8
00:00:18,325 --> 00:00:18,865
How we doing?
9
00:00:18,865 --> 00:00:24,205
Alice, Jason Elder here with
another LE, a podcast Deep dive.
10
00:00:24,214 --> 00:00:29,674
A first today welcoming back to
the show, Christopher Cruz to
11
00:00:29,674 --> 00:00:36,934
help us talk more about artificial
intelligence and impact on analysis.
12
00:00:36,995 --> 00:00:37,955
Chris, how we doing?
13
00:00:38,144 --> 00:00:39,165
Doing well, Jason.
14
00:00:39,165 --> 00:00:40,095
So happy to be back.
15
00:00:40,095 --> 00:00:41,025
Thank you for having me.
16
00:00:41,445 --> 00:00:42,794
I appreciate you.
17
00:00:42,864 --> 00:00:49,285
I just informally have started to get
different takes on AI and how it relates
18
00:00:49,285 --> 00:00:55,594
to law enforcement analysis, I remember
seeing your presentation in Vegas a couple
19
00:00:55,594 --> 00:01:02,814
years back at the IACA conference and
still quote your, lesson about trying to.
20
00:01:03,209 --> 00:01:07,199
Calculate what the black box
calculates and how it might take
21
00:01:07,199 --> 00:01:09,839
you 60 years to do one calculation.
22
00:01:09,839 --> 00:01:14,429
So I thought, who better to continue
this conversation than to have
23
00:01:14,429 --> 00:01:20,679
you back on the podcast to, give
us your take on AI and analysis.
24
00:01:20,864 --> 00:01:21,434
Oh, for sure.
25
00:01:21,434 --> 00:01:22,214
Yeah, it sounds amazing.
26
00:01:22,214 --> 00:01:25,384
Yeah, that that I, a CA
presentation was it was fun.
27
00:01:25,384 --> 00:01:28,354
It was one of my funnest, I think,
conference presentations I've ever done.
28
00:01:28,604 --> 00:01:30,554
Got a ton of really great feedback.
29
00:01:30,744 --> 00:01:34,124
I think I truly felt like a rockstar
after that one delivering it so
30
00:01:34,124 --> 00:01:38,274
well in Vegas and, getting such
great questions after the fact.
31
00:01:38,274 --> 00:01:39,175
So I definitely.
32
00:01:39,505 --> 00:01:43,615
It clearly like struck a lot of interest
chords with a lot of analysts, I think,
33
00:01:43,615 --> 00:01:45,025
which was really exciting to see.
34
00:01:45,235 --> 00:01:46,105
Yeah, yeah.
35
00:01:46,105 --> 00:01:49,435
So I want to get your impact on AI now.
36
00:01:49,555 --> 00:01:57,295
Maybe talk about future considerations
and talk about how AI impact the
37
00:01:57,295 --> 00:01:59,125
knowledge, skills and abilities.
38
00:01:59,125 --> 00:02:03,655
Get your take like everybody
else is, is AI taking over.
39
00:02:03,910 --> 00:02:04,750
Analyst jobs.
40
00:02:05,620 --> 00:02:05,890
Yeah.
41
00:02:05,890 --> 00:02:10,180
It's a, it's a fascinating topic and
a, and a fascinating question and I
42
00:02:10,180 --> 00:02:14,325
think analysts as a whole, kind of
across all levels of government are.
43
00:02:14,980 --> 00:02:16,600
In the middle of it,
they're in the thick of it.
44
00:02:16,709 --> 00:02:20,609
Probably much more than a lot of
other public safety professionals.
45
00:02:20,609 --> 00:02:24,359
I think the, analyst as a
technologist has, has started to
46
00:02:24,359 --> 00:02:27,989
become the norm, I think, and they're
starting to look to us to be like
47
00:02:27,989 --> 00:02:29,429
you're, you're the techie expert.
48
00:02:29,429 --> 00:02:31,260
You're the, you're the one
who's good with computers.
49
00:02:31,329 --> 00:02:32,559
What's the AI thing?
50
00:02:32,559 --> 00:02:33,459
What are we gonna do with it?
51
00:02:33,459 --> 00:02:34,209
How are we gonna use it?
52
00:02:34,209 --> 00:02:37,369
So it's really putting analysts
potentially front and center.
53
00:02:37,584 --> 00:02:39,864
But I think the big question
is are, we all prepared and
54
00:02:39,864 --> 00:02:41,394
ready for what's already here?
55
00:02:41,394 --> 00:02:42,534
And, and what's gonna come?
56
00:02:42,714 --> 00:02:42,924
Yeah.
57
00:02:42,924 --> 00:02:47,444
No, I think before we start I
think, I think it's an interesting
58
00:02:47,444 --> 00:02:50,355
concept about analyst as.
59
00:02:50,975 --> 00:02:52,924
What did you refer to as a technologist?
60
00:02:53,135 --> 00:02:53,554
Analyst.
61
00:02:53,554 --> 00:02:54,904
As a technologist, yeah.
62
00:02:54,964 --> 00:02:55,234
Yeah.
63
00:02:55,234 --> 00:03:02,045
I think that's a fascinating concept
because I've been in roles where I was
64
00:03:02,045 --> 00:03:08,614
the analyst and I was basically the
middleman between it and executives.
65
00:03:08,829 --> 00:03:14,829
And knew enough about both sides to
really help with the matter, going
66
00:03:14,829 --> 00:03:22,149
directly from it to executive is
is something that doesn't happen
67
00:03:22,149 --> 00:03:24,669
very smoothly in a lot of cases.
68
00:03:24,669 --> 00:03:27,589
So analysts are needed
to, be that middle man.
69
00:03:27,804 --> 00:03:29,574
Yeah, I think you're totally right.
70
00:03:29,574 --> 00:03:34,134
There is there's often a friction
between, well, to be honest, it
71
00:03:34,314 --> 00:03:35,694
and, and everyone else often.
72
00:03:35,849 --> 00:03:40,209
But it can be its own little box of,
of interesting characters to work with.
73
00:03:40,209 --> 00:03:44,849
But I always feel like thinking back
to early in my careers and analysts to
74
00:03:45,069 --> 00:03:49,549
working with a lot of particularly sworn
counterparts agents or investigators
75
00:03:49,549 --> 00:03:51,929
or officers or troopers who.
76
00:03:52,169 --> 00:03:55,709
We're less comfortable on a computer,
less comfortable with technology, less
77
00:03:55,709 --> 00:03:57,014
comfortable navigating the internet.
78
00:03:57,609 --> 00:04:01,059
And so there was always this expectation
that, well, the analyst is, is your
79
00:04:01,059 --> 00:04:02,859
go-to person for that type of stuff.
80
00:04:02,859 --> 00:04:07,369
You're, there to help them not just
do analytical work, but somehow
81
00:04:07,369 --> 00:04:11,599
that also includes supporting
the technology efforts as well.
82
00:04:11,599 --> 00:04:15,069
And I think now we're
getting AI thrown on us.
83
00:04:15,069 --> 00:04:19,699
I think you'll find many analysts
are deeply involved in like, rolling
84
00:04:19,699 --> 00:04:22,189
out new CAD or new RMS systems.
85
00:04:22,189 --> 00:04:23,209
Like we really are taking.
86
00:04:23,894 --> 00:04:27,584
For better or for worse, a lot of
responsibility is this technologist
87
00:04:27,584 --> 00:04:29,474
role in, in agencies and departments.
88
00:04:29,714 --> 00:04:38,054
Yeah, and, and I think there's certainly
executives or even officers that
89
00:04:38,174 --> 00:04:42,924
certainly can handle going straight
to it in our, in our scenario here.
90
00:04:43,239 --> 00:04:47,814
I, I think one of the, the
aspects of it for analysts though.
91
00:04:48,024 --> 00:04:52,944
Is, as you mentioned already dealing with
the computer, already dealing with data.
92
00:04:53,124 --> 00:04:58,314
Having that, time and experience
it helps with the matter.
93
00:04:58,314 --> 00:05:04,294
Whereas even if, even if executives or
officers had the knowledge, skills and
94
00:05:04,294 --> 00:05:10,804
abilities, like they have a lot other
responsibilities outside of the desk to,
95
00:05:10,874 --> 00:05:14,784
take care of and might not have that time.
96
00:05:15,144 --> 00:05:19,974
Necessary to really corroborate
some of the leads you get from ai.
97
00:05:20,184 --> 00:05:21,144
Oh, a hundred percent true.
98
00:05:21,144 --> 00:05:23,964
I think a lot of it is exactly your point.
99
00:05:24,084 --> 00:05:28,474
A lot of other positions having to, to
do other things that, that take them away
100
00:05:28,474 --> 00:05:32,574
from the desk or from the, the laptop or
the computer, where I think more often
101
00:05:32,574 --> 00:05:34,374
not that's where the analyst lives and so.
102
00:05:34,744 --> 00:05:38,254
There's just this sort of proxy
effect that occurs where we're
103
00:05:38,254 --> 00:05:39,994
just always there and always ready.
104
00:05:39,994 --> 00:05:43,394
So we just end up getting that
mantle kind of thrown upon us.
105
00:05:43,574 --> 00:05:49,354
Yeah, and I think one, I, I talked
about it on another podcast where
106
00:05:49,354 --> 00:05:53,704
I was talking about the different
layers, the different tiers.
107
00:05:53,904 --> 00:06:00,914
Of ai because that bottom tier I just is
maybe something just naturally happens
108
00:06:00,914 --> 00:06:03,044
in a Google search now or B browser.
109
00:06:03,044 --> 00:06:06,434
You're asking a question,
you're getting an answer.
110
00:06:06,734 --> 00:06:13,164
Maybe you're asking chat GBT
or AI chat to, fix some code.
111
00:06:13,194 --> 00:06:16,344
There's, it's a specific
question that you're asking.
112
00:06:16,584 --> 00:06:20,454
It's like that bottom tier that
anybody that's dipping their toe
113
00:06:20,454 --> 00:06:22,764
in, that's, that's where they start.
114
00:06:22,944 --> 00:06:30,684
And then it's, it seems like it jumps
from there to then large language models
115
00:06:30,684 --> 00:06:35,814
and then craft databases, and there's
just this huge leap to this next tier
116
00:06:35,814 --> 00:06:38,044
thing where it's like, oh, this is.
117
00:06:38,679 --> 00:06:43,539
This is a lot, lot more
responsibility, a lot more set up
118
00:06:43,739 --> 00:06:48,689
a lot more things to consider other
than just asking the computer a
119
00:06:48,689 --> 00:06:50,219
question, for lack of better term.
120
00:06:50,429 --> 00:06:50,729
Oh yeah.
121
00:06:50,729 --> 00:06:55,319
I think I think everyone's first foray
is that entry level experience with
122
00:06:55,319 --> 00:07:00,189
ai where we're really just using it as
like the new Google search essentially.
123
00:07:00,189 --> 00:07:02,109
It's just, I, I have a question.
124
00:07:02,139 --> 00:07:03,189
I want an answer.
125
00:07:03,459 --> 00:07:07,384
Let me input and, and get a
result, or maybe I have a, a, a
126
00:07:07,389 --> 00:07:09,039
slightly more complex question.
127
00:07:09,044 --> 00:07:10,479
I'll, I'll get a longer answer.
128
00:07:10,834 --> 00:07:14,014
I think that's where everyone
starts, and I think getting people
129
00:07:14,044 --> 00:07:15,544
to, to move beyond that is hard.
130
00:07:15,544 --> 00:07:20,784
'cause after that it does spiral very
quickly into this, this massive array
131
00:07:20,784 --> 00:07:24,594
of different technologies and different
techniques and even even different types
132
00:07:24,594 --> 00:07:29,004
of AI models that you start interfa
interfacing with or experiencing.
133
00:07:29,004 --> 00:07:29,184
So.
134
00:07:29,784 --> 00:07:33,504
It's like, it is kind of like a pool where
it's really easy to take the first step.
135
00:07:33,554 --> 00:07:37,934
But then after that it just drops off
drastically into like a whole new world
136
00:07:37,934 --> 00:07:40,034
of technology for people to figure out.
137
00:07:40,129 --> 00:07:40,419
Yeah.
138
00:07:40,874 --> 00:07:41,144
Yeah.
139
00:07:41,144 --> 00:07:44,504
And I might, I always like to give the
tidbit too, for the listeners that.
140
00:07:45,029 --> 00:07:51,329
I think Ja Mondale who does our open
secrets deep dives, said that, hey, once
141
00:07:51,329 --> 00:07:56,039
you use chat, GPT or, and any of them,
you can always ask for their source.
142
00:07:56,069 --> 00:07:58,649
The next question can
be, what is your source?
143
00:07:58,649 --> 00:08:01,169
And you can always review
the sources as well.
144
00:08:01,169 --> 00:08:05,649
So it's, you can try to corroborate
what you're getting from the ai.
145
00:08:05,929 --> 00:08:10,699
Yeah, always, always recommend people
don't blindly trust whatever it spits out.
146
00:08:10,699 --> 00:08:14,134
There is no shortage of stories
of people who have, who have
147
00:08:14,349 --> 00:08:15,549
gone awry with it a little bit.
148
00:08:15,549 --> 00:08:20,209
I think there's a massive uptick
of allegedly lawyers using it
149
00:08:20,259 --> 00:08:22,359
for cases and depositions and.
150
00:08:22,699 --> 00:08:26,849
An increasing issue of like made
up or fictional case law getting
151
00:08:26,849 --> 00:08:27,989
thrown in there that mm-hmm.
152
00:08:28,179 --> 00:08:30,939
Lawyers and counsel are not
catching, but judges are.
153
00:08:30,969 --> 00:08:32,379
And I hear the judges are not happy.
154
00:08:33,884 --> 00:08:34,894
Yeah, I'm sure.
155
00:08:35,589 --> 00:08:36,069
I'm sure.
156
00:08:36,639 --> 00:08:42,039
So I think it's one of the
difficult aspects of talking
157
00:08:42,039 --> 00:08:44,349
about current state analysis.
158
00:08:44,764 --> 00:08:51,184
And AI is that there are so many
different scenarios for an analyst.
159
00:08:53,159 --> 00:08:58,199
There's 17,000 police departments in
the United States, and so you have
160
00:08:58,199 --> 00:09:03,959
everything from your small town, one
person analyst, all the way up to
161
00:09:03,959 --> 00:09:09,159
your NYPD having hundreds of analysts.
162
00:09:09,229 --> 00:09:12,274
I know that's difficult,
but, talk a little bit about.
163
00:09:12,974 --> 00:09:17,864
What you think the, current
state is with analysis and ai.
164
00:09:18,044 --> 00:09:22,724
Yeah, I think we're at kind of an, an
interesting inflection point maybe.
165
00:09:22,724 --> 00:09:24,354
So I did, a little research.
166
00:09:24,354 --> 00:09:25,674
I made sure I came prepared, right.
167
00:09:25,674 --> 00:09:27,984
So I've found a couple
interesting surveys from.
168
00:09:28,485 --> 00:09:33,055
Companies that many of us probably
know and debatably love perhaps.
169
00:09:33,055 --> 00:09:36,605
And I won't, I won't name these companies
'cause I don't want to appeal like a,
170
00:09:36,605 --> 00:09:37,985
a advertising or anything like that.
171
00:09:37,985 --> 00:09:40,055
But these are two mm-hmm two big
companies that most people in
172
00:09:40,055 --> 00:09:42,225
public safety would recognize.
173
00:09:42,575 --> 00:09:46,205
The first one talked about
kind of the, the state of.
174
00:09:46,510 --> 00:09:49,660
Of hiring and, and personnel
and, and police departments.
175
00:09:49,660 --> 00:09:53,540
And I think they surveyed over 500
police departments and they found
176
00:09:53,820 --> 00:09:57,190
only 14% of those were fully staffed.
177
00:09:57,250 --> 00:10:01,840
And half of them were below eight
at or below 80% capacity, I think.
178
00:10:02,200 --> 00:10:05,320
And so I, I know we were all
kind of suffering staffing
179
00:10:05,320 --> 00:10:07,540
issues after 2020, kind of
180
00:10:07,630 --> 00:10:07,720
mm-hmm.
181
00:10:07,960 --> 00:10:10,120
Blew up the, the role
of policing as a whole.
182
00:10:10,470 --> 00:10:11,670
And it seems like we're still.
183
00:10:12,315 --> 00:10:14,505
Haven't righted the ship since then.
184
00:10:14,725 --> 00:10:19,775
But there was another interesting takeaway
that 75% of those departments that
185
00:10:19,775 --> 00:10:24,405
were surveyed or off, or professionals
that were surveyed found they thought
186
00:10:24,405 --> 00:10:29,535
that AI would help make the job easier
and boost investigative efficiency.
187
00:10:29,535 --> 00:10:32,685
So three out of four people were
like, yeah, AI is gonna make
188
00:10:32,685 --> 00:10:34,455
everything better and easier.
189
00:10:34,785 --> 00:10:37,105
Compared that to another survey.
190
00:10:37,105 --> 00:10:41,435
Were upwards of 80% of
respondents out of 2000.
191
00:10:41,485 --> 00:10:45,125
Professionals in public safety
upwards of 80% were saying, Hey, AI
192
00:10:45,125 --> 00:10:46,745
is gonna make investigations easier.
193
00:10:46,745 --> 00:10:48,065
It's gonna make them more effective.
194
00:10:48,425 --> 00:10:53,965
But we also know that there's this big
concern on, well, is it gonna replace us?
195
00:10:53,965 --> 00:10:56,425
Is it gonna, is it gonna
replace the role of the analyst?
196
00:10:56,425 --> 00:10:59,815
Or how heavily is it gonna
augment the role of the analyst?
197
00:11:00,075 --> 00:11:04,245
And so I think that is sort of
where we're kind of left wondering.
198
00:11:04,245 --> 00:11:05,285
And I think the.
199
00:11:05,930 --> 00:11:07,370
It's really up to the analyst.
200
00:11:07,370 --> 00:11:10,340
I think to your point, there are
really small rural departments that
201
00:11:10,340 --> 00:11:13,460
are just gonna look for how do we
maximize efficiency and mm-hmm.
202
00:11:13,700 --> 00:11:17,140
In their minds that might just be plop
some AI down and then just run with it.
203
00:11:17,450 --> 00:11:21,290
I think more mature departments are
hopefully gonna gonna look more of it as
204
00:11:21,290 --> 00:11:23,480
how do we augment rather than replace?
205
00:11:23,530 --> 00:11:27,760
But I think it's really gonna be dependent
on the, the attitude of the agency,
206
00:11:27,760 --> 00:11:31,420
their leadership, but also the, the
contributions of the analysts to, to.
207
00:11:31,795 --> 00:11:34,225
Chart their own path a
little bit on those issues.
208
00:11:34,405 --> 00:11:39,785
Well, I, I think and I, I've this
is one of the things that Dawn
209
00:11:39,795 --> 00:11:43,335
Reebe and I talked about on when
we were talking about this topic.
210
00:11:43,410 --> 00:11:50,685
I, think potentially a lot of the, tasks
that analysts get assigned that, may
211
00:11:50,685 --> 00:11:57,390
not necessarily be analysis, might be
alleviated using AI like graphic design
212
00:11:57,390 --> 00:12:03,200
or building a PowerPoint or building a
sign for the company party the, those
213
00:12:03,200 --> 00:12:09,480
type of things where you're, you're like,
oh, that's clerical task that analysts
214
00:12:09,510 --> 00:12:13,560
sometimes get assigned and, and I told
her, I was like, Hey, there's analysts
215
00:12:13,560 --> 00:12:15,390
out there that really enjoy doing that.
216
00:12:15,390 --> 00:12:18,900
And that might be something where it's
like, Hey, yeah, I don't want you spending
217
00:12:19,090 --> 00:12:21,070
an hour getting this PowerPoint correct.
218
00:12:21,070 --> 00:12:22,960
Get it ready in five minutes.
219
00:12:23,110 --> 00:12:25,870
But I feel that those.
220
00:12:26,245 --> 00:12:32,305
Those kind of tasks, data
cleaning and then fixing code.
221
00:12:32,305 --> 00:12:37,055
And, and those are really some
of the things that may take a lot
222
00:12:37,055 --> 00:12:43,805
of time away from analysts and
free them up to do other things.
223
00:12:43,935 --> 00:12:44,175
So.
224
00:12:44,195 --> 00:12:50,105
My take is, AI is going to
take analyst jobs in a way.
225
00:12:50,285 --> 00:12:56,260
And I think analysts, if they don't start
understanding a little bit more about ai.
226
00:12:57,230 --> 00:13:02,060
They're gonna be seen as dinosaurs
, what is your take on that?
227
00:13:02,390 --> 00:13:03,470
Agree or disagree?
228
00:13:03,570 --> 00:13:04,710
I definitely agree.
229
00:13:04,770 --> 00:13:06,960
I think you're, I, I agree on two parts.
230
00:13:07,020 --> 00:13:07,440
I agree.
231
00:13:07,440 --> 00:13:08,140
If, if you're not.
232
00:13:08,795 --> 00:13:11,075
Learning to use it in some capacity.
233
00:13:11,185 --> 00:13:15,285
You're gonna be left behind, like
analysts who can't make some use of it.
234
00:13:15,465 --> 00:13:18,045
At some point are, I think people
are gonna look at you funny.
235
00:13:18,045 --> 00:13:20,985
Like, what do you mean you don't
know how to use Jet Gt the same way?
236
00:13:21,265 --> 00:13:24,235
When you meet someone who's still using
like an aol.com email, you're like,
237
00:13:24,235 --> 00:13:25,885
what do you mean still using aol.com?
238
00:13:26,155 --> 00:13:26,970
Why would you do that?
239
00:13:27,175 --> 00:13:27,770
It's gonna be same.
240
00:13:27,770 --> 00:13:27,790
I like
241
00:13:28,105 --> 00:13:28,975
my Hotmail.
242
00:13:29,395 --> 00:13:30,205
My Hotmail email.
243
00:13:30,325 --> 00:13:31,195
Yahoo's great.
244
00:13:31,195 --> 00:13:31,465
No.
245
00:13:31,565 --> 00:13:34,685
So I think, I think that'll
be, I think there'll be some.
246
00:13:34,780 --> 00:13:38,030
Stigma and , some just
change has to come with it.
247
00:13:38,110 --> 00:13:42,350
I think the other part I who agree
with is like there's so much tasking
248
00:13:42,350 --> 00:13:44,660
that is just epically time consuming.
249
00:13:44,720 --> 00:13:49,710
And I think good analysts and smart
analysts will learn how to game the
250
00:13:49,740 --> 00:13:54,770
system a little bit to like optimize
things where AI is useful and, and
251
00:13:54,770 --> 00:13:56,930
spend more time on the actual like.
252
00:13:57,000 --> 00:14:00,420
Hard analytics that, that
they all probably want to do.
253
00:14:00,740 --> 00:14:04,970
When I did my presentation in Vegas, I
think I hit on some really great examples
254
00:14:04,970 --> 00:14:09,680
where number one, that whole presentation,
it was partially created with ai and that
255
00:14:09,680 --> 00:14:11,450
was kind of a disclaimer at the start, is.
256
00:14:11,750 --> 00:14:12,800
I didn't make this whole thing.
257
00:14:12,800 --> 00:14:18,210
I, I made it in concert with two different
AI bots that helped flush out content and
258
00:14:18,210 --> 00:14:20,400
generated slides and created pictures.
259
00:14:20,760 --> 00:14:26,675
I also created a, a framework on
how to assess like usability or
260
00:14:26,675 --> 00:14:29,255
risk in AI for police departments.
261
00:14:29,525 --> 00:14:33,725
That framework was created from ai and I,
I did a lot of polish and cleaning on it,
262
00:14:33,725 --> 00:14:36,305
but I used AI to generate that framework.
263
00:14:36,525 --> 00:14:37,875
And then I went one step further.
264
00:14:38,220 --> 00:14:42,810
And I made a website and I put the,
the slides and summaries and the tool
265
00:14:42,810 --> 00:14:46,950
that I created all on the website,
which itself was also created with ai.
266
00:14:47,320 --> 00:14:51,230
And what a lot of people didn't
realize is 90% of i, of what I just
267
00:14:51,230 --> 00:14:56,150
described, I did from when I landed in
Vegas to when I did the presentation.
268
00:14:56,270 --> 00:14:58,310
So I didn't do that like
months ahead of time.
269
00:14:58,630 --> 00:15:02,650
I put myself through this crazy test
of how much can I knock out with
270
00:15:02,650 --> 00:15:04,510
AI in a very short amount of time.
271
00:15:04,510 --> 00:15:08,490
And I really like put my money where my
mouth was and showed like I could build a
272
00:15:08,490 --> 00:15:14,120
bunch of stuff in 48 hours essentially and
have, have it presentation ready which I
273
00:15:14,120 --> 00:15:16,190
thought was really fascinating experience.
274
00:15:16,685 --> 00:15:20,180
And you used the word
bought there, is that?
275
00:15:20,570 --> 00:15:26,080
Something that you customized yourself or
is that just you're just referring to that
276
00:15:26,080 --> 00:15:30,100
as, as one of the AI tools like chat, GBT
277
00:15:30,100 --> 00:15:31,720
?
In that instance it's
a little bit of both.
278
00:15:31,720 --> 00:15:34,585
So if you look at chat
GBT as an example mm-hmm.
279
00:15:34,685 --> 00:15:37,715
There, there's just a core function
that you can use, but then they
280
00:15:37,715 --> 00:15:41,145
have kind of like a, I forget the
name they use, but it's almost
281
00:15:41,145 --> 00:15:45,195
like a store where you can find all
these customed GPTs they call them.
282
00:15:45,195 --> 00:15:45,225
Okay.
283
00:15:45,565 --> 00:15:48,935
And essentially what you can do
is feed it a bunch of extra data.
284
00:15:49,025 --> 00:15:53,615
You can set very specific instructions and
guidelines and you can sort of tailor one
285
00:15:53,945 --> 00:15:57,005
to do or act or respond in a certain way.
286
00:15:57,285 --> 00:16:02,115
And I actually had one that was tailored
to like the, the role of a crime analyst.
287
00:16:02,115 --> 00:16:05,995
And I fed it dozens of like job
descriptions of what a crime analyst does.
288
00:16:05,995 --> 00:16:10,645
I fed it like books about crime
analysis and like PDFs and reports.
289
00:16:10,725 --> 00:16:14,775
And I told it like, you are a seasoned
crime analyst, answer every question
290
00:16:14,775 --> 00:16:17,955
like, you've been doing this for
30 years and you're an expert in
291
00:16:17,955 --> 00:16:19,095
the field, and things like that.
292
00:16:19,095 --> 00:16:21,945
So I sort of tricked it into
thinking it was an expert.
293
00:16:22,165 --> 00:16:24,595
And then everything it did,
it did through this lens of.
294
00:16:25,115 --> 00:16:27,365
Thinking it was a crime, an crime analyst.
295
00:16:28,535 --> 00:16:28,625
Hmm.
296
00:16:28,625 --> 00:16:31,055
And what were some interesting
things that came outta that?
297
00:16:31,835 --> 00:16:35,585
I was surprised at how good
I was at certain things.
298
00:16:35,615 --> 00:16:41,585
One of the things that I demoed but was
like kind of floored me when it happened.
299
00:16:41,885 --> 00:16:44,645
And I, I realized that when I was like
setting up the presentation, so I had
300
00:16:44,645 --> 00:16:49,065
this example where I wanted to show like,
oh, you could take like crime stats and
301
00:16:49,065 --> 00:16:52,785
you could put it in one of these tools
and you can ask it questions and it will
302
00:16:52,785 --> 00:16:57,555
generate answers and it will, it will
calculate trends and hotspots for you.
303
00:16:57,745 --> 00:17:00,115
So you don't necessarily have to
do some of that stuff manually.
304
00:17:00,375 --> 00:17:04,185
You can just invest your time in
like operationalizing those findings.
305
00:17:04,435 --> 00:17:09,485
So I downloaded some open source
openly available crime data from like
306
00:17:09,485 --> 00:17:13,175
data.gov, which just has a bunch of
free data from different sources.
307
00:17:13,455 --> 00:17:16,665
I think it was like New Jersey
crime data or Ohio maybe.
308
00:17:16,955 --> 00:17:20,295
And I didn't notice, but the
spreadsheet, the CSV that the
309
00:17:20,295 --> 00:17:22,095
data was in was like messed up.
310
00:17:22,095 --> 00:17:23,685
The formatting was all jacked up.
311
00:17:23,685 --> 00:17:24,975
The columns were off.
312
00:17:25,310 --> 00:17:26,570
Kilter from each other.
313
00:17:26,790 --> 00:17:29,460
So an analyst would've opened that
and been like, oh man, I gotta
314
00:17:29,490 --> 00:17:31,050
like fix this stupid spreadsheet.
315
00:17:31,080 --> 00:17:32,280
Lemme clean up this data.
316
00:17:32,670 --> 00:17:35,460
I didn't check it that
closely, I just uploaded it.
317
00:17:35,610 --> 00:17:39,180
And it actually said, Hey,
FYI, the spreadsheet's broken.
318
00:17:39,180 --> 00:17:40,590
I went ahead and fixed it for you.
319
00:17:40,620 --> 00:17:42,810
Here's a nice new clean
version of the spreadsheet.
320
00:17:43,050 --> 00:17:45,840
Now let me go ahead and answer
this question you have about
321
00:17:45,840 --> 00:17:47,340
like, calculating crime trends.
322
00:17:47,610 --> 00:17:50,680
So I thought that was a really great
example of like something that is.
323
00:17:51,725 --> 00:17:55,685
Like, arguably still high value, but
certainly like time consuming and
324
00:17:55,685 --> 00:17:57,815
not really high value as an analyst.
325
00:17:57,815 --> 00:18:00,215
Just high value as in
like someone has to do it.
326
00:18:00,635 --> 00:18:04,595
And I think that's where AI has an
opportunity is to optimize the things
327
00:18:04,595 --> 00:18:07,725
we have to do, but aren't true.
328
00:18:07,725 --> 00:18:09,195
Like analytical work.
329
00:18:09,255 --> 00:18:12,075
, I feel for the analyst.
330
00:18:12,980 --> 00:18:18,550
Because even, though they could do all
that, themselves, if I'm running something
331
00:18:18,550 --> 00:18:24,690
like that, trying to turn it into a
supervisor, I'm double checking, running
332
00:18:24,690 --> 00:18:27,410
something parallel to it, every time.
333
00:18:28,020 --> 00:18:32,960
To make sure that it's producing
accurate results just in case it,
334
00:18:33,200 --> 00:18:37,400
it reads something it shouldn't and
produces results that I don't want,
335
00:18:37,580 --> 00:18:41,900
because obviously I don't wanna turn
something in that isn't what is expected.
336
00:18:42,080 --> 00:18:43,280
Yeah, a hundred percent.
337
00:18:43,280 --> 00:18:46,800
And that's something I've, I've told
a lot of analysts is, . Normally we
338
00:18:46,800 --> 00:18:49,590
would say like, in the cybersecurity
world, like trust but verify.
339
00:18:49,590 --> 00:18:53,670
But I would almost say don't trust,
just and just verify that you
340
00:18:53,670 --> 00:18:54,930
never know what it's spitting out.
341
00:18:54,930 --> 00:18:59,630
There is a whole field of ai called
hallucinations, which, where it
342
00:18:59,635 --> 00:18:59,775
mm-hmm
343
00:18:59,915 --> 00:19:03,370
it generates wrong things
or ridiculous things.
344
00:19:03,660 --> 00:19:05,070
For a hot minute.
345
00:19:05,755 --> 00:19:05,995
Chat.
346
00:19:05,995 --> 00:19:09,595
GPT was notorious for
something they call glazing.
347
00:19:09,965 --> 00:19:13,465
And this is like an AI term as
well where it tells you whatever
348
00:19:13,465 --> 00:19:14,665
it thinks you want to hear.
349
00:19:14,695 --> 00:19:18,565
Even if that's not the right answer,
it tries to guess what it thinks
350
00:19:18,565 --> 00:19:19,915
you think the right answer is.
351
00:19:19,915 --> 00:19:24,745
And it hyper fixates on being as nice
and kind and supportive of whatever
352
00:19:24,745 --> 00:19:27,565
absurd or ridiculous thing you're saying.
353
00:19:27,615 --> 00:19:29,595
And I've definitely, in my testing.
354
00:19:29,860 --> 00:19:32,170
Noticed it happening to me
or I'm like, wait a minute.
355
00:19:32,170 --> 00:19:33,790
This is like, this is wrong.
356
00:19:33,790 --> 00:19:35,830
This is totally wrong, and I yell at it.
357
00:19:35,830 --> 00:19:39,705
I type meanly, I should say, and I say,
why are you giving me the wrong answer?
358
00:19:40,465 --> 00:19:43,765
And I once had one of these tools
literally respond to me and say,
359
00:19:43,765 --> 00:19:47,695
well, I really wanted to show you
the answer I thought you wanted, so
360
00:19:47,695 --> 00:19:49,105
that's why I said all those things.
361
00:19:49,105 --> 00:19:49,765
I'm like, oh my God.
362
00:19:49,765 --> 00:19:52,585
It's like, it's lying and it's even
admitting to me it's lying and then
363
00:19:52,585 --> 00:19:54,235
it's saying, sorry that I lied.
364
00:19:54,535 --> 00:19:57,595
And I'm like, this is a wild
experience to be having with an,
365
00:19:57,645 --> 00:19:59,625
inanimate object essentially.
366
00:19:59,805 --> 00:20:04,785
Yeah, it, it's funny 'cause I catch
myself when I'm asking something I use.
367
00:20:04,885 --> 00:20:05,305
, Please.
368
00:20:06,345 --> 00:20:09,070
I was like, will you please do this?
369
00:20:09,190 --> 00:20:09,490
Yeah.
370
00:20:09,490 --> 00:20:14,800
And I'm like, why am I using please,
like I'm talking to a human, but
371
00:20:14,800 --> 00:20:16,480
I, I catch myself doing that.
372
00:20:16,840 --> 00:20:20,620
Well, there's, there were, there were a
couple studies that came out and at least
373
00:20:20,620 --> 00:20:22,720
for a couple of the more popular ones.
374
00:20:23,355 --> 00:20:26,475
There was some early studies that
suggested if you say please and
375
00:20:26,475 --> 00:20:28,635
thank you, it performs better.
376
00:20:28,785 --> 00:20:28,875
Cool.
377
00:20:29,325 --> 00:20:32,355
There was also a study that came
out that said, when you use please
378
00:20:32,355 --> 00:20:34,245
and thank you, it works harder.
379
00:20:34,245 --> 00:20:36,255
Like it uses more processing power.
380
00:20:36,565 --> 00:20:40,250
But allegedly one of the companies
was like, every time you do that.
381
00:20:40,935 --> 00:20:43,665
It consumes more energy and
that costs us more money.
382
00:20:43,665 --> 00:20:45,670
So please stop saying please and thanking.
383
00:20:46,755 --> 00:20:48,225
Please stop saying please.
384
00:20:49,425 --> 00:20:50,595
Oh man.
385
00:20:55,965 --> 00:20:59,085
This is Dr. Eliann Carr from the
Ellensburg Police Department here
386
00:20:59,085 --> 00:21:03,165
to talk about the first of its kind,
the Crime Analyst Census survey.
387
00:21:03,255 --> 00:21:06,645
This is an opportunity for crime analysts
from around the world to be able to
388
00:21:06,645 --> 00:21:10,485
share information on the demographics
that make up the field, be able to look
389
00:21:10,485 --> 00:21:14,715
at the relationship between commission,
non-commission, and how we navigate that
390
00:21:14,715 --> 00:21:19,245
relationships in our career field, and
also to look at training opportunities and
391
00:21:19,245 --> 00:21:23,655
development that will help us foster the
opportunities for growth and development
392
00:21:23,835 --> 00:21:25,125
both personally and professionally.
393
00:21:25,740 --> 00:21:29,130
If you're interested in taking the
survey, you're welcome to go to the link
394
00:21:29,130 --> 00:21:32,820
in the show notes below, sure that your
voice is heard and included in the data.
395
00:21:38,436 --> 00:21:42,696
So I, I mean obviously there's
huge trust issues right now.
396
00:21:42,846 --> 00:21:45,856
And I don't know if you
even know, , let me back up.
397
00:21:46,006 --> 00:21:50,026
Do you know of police
departments that ha are.
398
00:21:50,081 --> 00:21:53,141
That are farther along than most.
399
00:21:53,321 --> 00:21:56,841
And have, have you, do you have any
insight into what they're doing?
400
00:21:56,971 --> 00:21:58,681
I know of a couple.
401
00:21:58,741 --> 00:22:01,321
I don't know how public, what
they're doing is, I won't,
402
00:22:01,651 --> 00:22:02,911
I don't wanna name them.
403
00:22:02,971 --> 00:22:03,061
Mm-hmm.
404
00:22:03,251 --> 00:22:07,046
But I know of a couple departments
that have stood up, like
405
00:22:07,106 --> 00:22:09,626
local large language models.
406
00:22:09,626 --> 00:22:13,526
So they have an an LLM, they have their
own little mini version of chat, GPT, and
407
00:22:13,526 --> 00:22:17,216
it's, it's running on their own server,
using their own hardware somewhere.
408
00:22:17,536 --> 00:22:22,296
And there the use cases I have
seen are feeding historical data
409
00:22:22,296 --> 00:22:24,236
into it and asking questions.
410
00:22:24,516 --> 00:22:26,016
There was a really interesting use case.
411
00:22:26,066 --> 00:22:26,751
I remember reading about.
412
00:22:27,461 --> 00:22:32,891
About someone using like AI and large
language models for like cold case
413
00:22:32,891 --> 00:22:34,301
and unsolved crimes where mm-hmm.
414
00:22:34,541 --> 00:22:40,281
They have got like 30 years of case files
and, and reports and witness testimony.
415
00:22:40,281 --> 00:22:45,051
And normally staff would sit there and,
and pour through that and read everything.
416
00:22:45,331 --> 00:22:48,481
There is an agency that this is
actually a, a service you can buy.
417
00:22:48,481 --> 00:22:52,021
I think a company actually sells this
capability now where you can pump all
418
00:22:52,021 --> 00:22:57,601
of that into a large language model and
ask your, your 30 year case questions
419
00:22:57,601 --> 00:23:01,891
who was where at this date or what
color shirt was this victim wearing?
420
00:23:02,171 --> 00:23:04,811
Without having to go back and
read and memorize all that stuff.
421
00:23:05,201 --> 00:23:08,771
There's definitely some case
studies that I've seen that
422
00:23:08,831 --> 00:23:11,001
talk about a lot of like pilots.
423
00:23:11,001 --> 00:23:15,001
I've seen a lot of agencies piloting
different use cases around using
424
00:23:15,001 --> 00:23:17,401
AI to generate police reports.
425
00:23:17,431 --> 00:23:18,301
So.
426
00:23:18,521 --> 00:23:22,211
I know some are like, you input some
basic data during a traffic stop, but
427
00:23:22,211 --> 00:23:25,571
then you push a button and most of
the paperwork auto generates itself.
428
00:23:25,621 --> 00:23:27,631
I've seen ones that are
partially generated.
429
00:23:27,921 --> 00:23:31,761
Those are the early use cases
that I'm seeing most common.
430
00:23:32,081 --> 00:23:35,601
Those both come to me though, kind
of back where we were with like.
431
00:23:36,271 --> 00:23:38,761
Question of, of validity or accuracy.
432
00:23:38,761 --> 00:23:42,211
Like, I don't know that I would trust
the AI to output a police report
433
00:23:42,491 --> 00:23:45,371
succinctly correctly, 100% of the time.
434
00:23:45,371 --> 00:23:46,121
I would have a lot of.
435
00:23:46,481 --> 00:23:47,171
Questions.
436
00:23:47,171 --> 00:23:50,921
And I know a lot of other people have that
concern because even here in Virginia,
437
00:23:50,921 --> 00:23:55,021
they our general assembly tried to pass
a law basically saying any police report
438
00:23:55,051 --> 00:23:57,571
that it results in arrest at any point.
439
00:23:57,891 --> 00:24:01,711
If it used AI in any part of it, you
have to put a big banner on the report.
440
00:24:01,711 --> 00:24:02,226
You have to notify.
441
00:24:02,956 --> 00:24:05,686
The Commonwealth's attorney, you
have to notify the defense attorney.
442
00:24:05,936 --> 00:24:09,976
So it's clearly something that lawmakers
and other policy makers are thinking
443
00:24:09,976 --> 00:24:11,656
about and trying to jump ahead of.
444
00:24:11,836 --> 00:24:16,481
Yeah, , and it's, the same notion of what
I just talked about, the trust that I
445
00:24:16,481 --> 00:24:18,921
would have to double check everything.
446
00:24:19,301 --> 00:24:20,241
So from the.
447
00:24:20,676 --> 00:24:26,316
Patrol officer, they're like, oh, well
I could have wrote this myself by the
448
00:24:26,316 --> 00:24:32,246
time I'm reviewing and maybe having to
fix something that it generated for me.
449
00:24:32,346 --> 00:24:38,346
It'll be interesting to observe
when, there starts to be trust.
450
00:24:39,111 --> 00:24:39,991
In, the system.
451
00:24:40,081 --> 00:24:40,351
Yeah.
452
00:24:40,351 --> 00:24:43,981
It's an interesting, it's an interesting
boundary I don't know what it takes,
453
00:24:44,131 --> 00:24:47,191
I don't know where, like, when does
the wave crest on trust essentially.
454
00:24:47,191 --> 00:24:49,201
, When is enough trust earned?
455
00:24:49,201 --> 00:24:53,291
If you, if you think other use cases
of ai, like obviously AI's, not
456
00:24:53,291 --> 00:24:57,791
you, we knew we've been using it
for decades in a variety of methods.
457
00:24:57,791 --> 00:25:00,341
We just never thought of it
the way we think of it now.
458
00:25:00,491 --> 00:25:00,826
The same.
459
00:25:01,621 --> 00:25:07,081
Core elements of AI that power
chat, GPT or, or Gemini or Claude.
460
00:25:07,301 --> 00:25:10,111
It's the same thing that
populates your Google search.
461
00:25:10,111 --> 00:25:13,621
Like when you go to Google and you
start typing and it auto finishes
462
00:25:13,621 --> 00:25:16,111
your search and it gives you a
bunch of options and what it thinks.
463
00:25:16,381 --> 00:25:19,561
That's the, that's the birth of the
large language model essentially.
464
00:25:19,561 --> 00:25:20,971
That's what it's based off of.
465
00:25:20,971 --> 00:25:24,661
That auto complete guessing,
but instead of guessing.
466
00:25:24,936 --> 00:25:29,196
The next three or four words, it
guesses the next three or 4,000 words
467
00:25:29,736 --> 00:25:33,336
and seems to be really good at guessing
what those words should be, ideally.
468
00:25:33,616 --> 00:25:38,496
But even thinking back to like when I was
an analyst when I first started, automatic
469
00:25:38,496 --> 00:25:40,296
license plate readers, LPR systems.
470
00:25:40,476 --> 00:25:42,216
That's a form of AI technically.
471
00:25:42,466 --> 00:25:45,946
Doing like facial recognition
on booking photos or things like
472
00:25:45,946 --> 00:25:47,716
that, that's AI technically.
473
00:25:47,716 --> 00:25:51,256
So we've been in the space
for a hot minute for sure.
474
00:25:51,256 --> 00:25:55,286
But it's just like the whole game has
flipped around on us, the whole use case
475
00:25:55,286 --> 00:25:56,816
of the technology and the view of it.
476
00:25:57,116 --> 00:25:59,276
And so we trusted it back then.
477
00:25:59,716 --> 00:26:03,106
Now it's sort of like the trust has,
we've seen how easily, I guess the
478
00:26:03,106 --> 00:26:06,916
trust can be broken, and now we're all
sort of left scratching our heads of
479
00:26:06,916 --> 00:26:10,336
like, what does it take to get back to
trustworthy technology in this space?
480
00:26:10,641 --> 00:26:10,721
Hmm.
481
00:26:10,751 --> 00:26:11,041
Yeah.
482
00:26:11,386 --> 00:26:11,566
Hmm.
483
00:26:11,986 --> 00:26:13,666
I want to go back to.
484
00:26:14,656 --> 00:26:20,086
You mentioned departments
creating their own LLM and then
485
00:26:20,086 --> 00:26:21,706
feeding it with their own data.
486
00:26:21,796 --> 00:26:26,756
And for the listeners that's unless
it's an open source data, you, you
487
00:26:26,756 --> 00:26:31,976
don't want to put any, department
sensitive data inside a chat GBT.
488
00:26:32,056 --> 00:26:34,096
That's, a first and foremost.
489
00:26:34,246 --> 00:26:38,626
But, and as I mentioned that, that
seems to be like a, just a huge leap
490
00:26:38,626 --> 00:26:45,406
for me as a, as I'm thinking about this,
of all the, what the analysts do now.
491
00:26:45,406 --> 00:26:50,206
They're even, it, they're they're,
they're asked, okay, now let's, let's
492
00:26:50,206 --> 00:26:52,396
create this large language model.
493
00:26:53,246 --> 00:26:56,246
, To me, it seems so
daunting, but I also think.
494
00:26:56,871 --> 00:26:58,161
That's where we're headed.
495
00:26:58,371 --> 00:27:03,631
Yeah, I think, I think there's so
much concern on the data security and
496
00:27:03,841 --> 00:27:06,601
the sensitivity and SCI compliance.
497
00:27:06,601 --> 00:27:10,321
All these things come up when we
look at these tools and Yes, yes.
498
00:27:10,411 --> 00:27:14,071
Kids at home don't, don't put
data into tools you don't own.
499
00:27:14,731 --> 00:27:17,911
'Cause it's, it's not your tool and
therefore it's no longer your data.
500
00:27:18,216 --> 00:27:18,306
Mm-hmm.
501
00:27:19,176 --> 00:27:22,196
I heard a really interesting, I was
at a conference yesterday actually,
502
00:27:22,256 --> 00:27:28,056
about cybersecurity and the, the chief
Information security officer for Virginia
503
00:27:28,056 --> 00:27:33,246
was talking Mike Watson, and he was asked
about dealing with sensitive government
504
00:27:33,246 --> 00:27:36,696
data that gets put into these things, and
he said, yeah, that's such a, a concern.
505
00:27:36,696 --> 00:27:38,766
And, and we, we try to
combat it every day.
506
00:27:39,366 --> 00:27:42,186
But he, he highlighted something
I hadn't thought of, and it's just
507
00:27:42,246 --> 00:27:46,696
one more thing to keep me awake at
night, I guess now is he's equally
508
00:27:46,696 --> 00:27:50,146
concerned around the sensitive
data that these things can produce.
509
00:27:50,506 --> 00:27:52,656
And he gave a couple examples in that.
510
00:27:52,991 --> 00:27:57,391
Some of these are really good at
understanding all kinds of calculations
511
00:27:57,391 --> 00:28:00,181
and algorithms and, and numeric equations.
512
00:28:00,451 --> 00:28:03,651
And for example people that are
a little bit older have social
513
00:28:03,651 --> 00:28:07,131
security numbers that follow like
a logical pattern depending on.
514
00:28:07,516 --> 00:28:10,456
Who you are and, and when you
were born and where you were born.
515
00:28:10,606 --> 00:28:10,696
Mm-hmm.
516
00:28:10,886 --> 00:28:13,826
, Several numbers in your social
security number can be guessed
517
00:28:13,826 --> 00:28:15,986
or pretty cleanly determined.
518
00:28:16,236 --> 00:28:19,776
And you can essentially with
the right amount of information.
519
00:28:20,291 --> 00:28:25,001
Get someone's social security number down
to almost all the digits except for one,
520
00:28:25,241 --> 00:28:27,481
using some of these large language models.
521
00:28:27,661 --> 00:28:31,021
'cause it's just really good at, at
navigating all the reasoning and logic.
522
00:28:31,331 --> 00:28:35,131
The other example I gave was you
might have a bunch of data about
523
00:28:35,161 --> 00:28:39,361
sensitive locations or something
or critical infrastructure maybe.
524
00:28:39,661 --> 00:28:40,891
And you put it in there.
525
00:28:41,211 --> 00:28:46,591
And obviously that's concerning, but
then it spits out you know visuals or,
526
00:28:46,651 --> 00:28:52,551
or maps or lists that prioritize things
or highlights weaknesses in things or
527
00:28:52,771 --> 00:28:54,721
highlight risks you didn't know you have.
528
00:28:54,721 --> 00:28:57,911
And so there's not just concern about
what you're putting in, but sometimes
529
00:28:57,911 --> 00:29:00,851
what you're generating is more
sensitive than the data you gave it.
530
00:29:01,211 --> 00:29:04,571
And now that is theoretically
owned by someone else.
531
00:29:04,601 --> 00:29:06,761
'cause now they, they may or
may not have access to it.
532
00:29:07,101 --> 00:29:09,001
But then how do you scrub it?
533
00:29:09,001 --> 00:29:09,751
How do you protect it?
534
00:29:09,751 --> 00:29:14,071
How do you make sure that it, the, the
model doesn't memorize that and share that
535
00:29:14,071 --> 00:29:17,341
with other people somehow, what it doesn't
learn from what you're teaching it.
536
00:29:17,651 --> 00:29:20,051
So there's so many like bizarre risks.
537
00:29:20,331 --> 00:29:20,341
But.
538
00:29:20,751 --> 00:29:23,181
That's why we see this move in law
enforcement of, if you're going to
539
00:29:23,331 --> 00:29:27,241
use this type of technology a lot of
people are doing what we call on-prem
540
00:29:27,246 --> 00:29:30,781
or or on-premise deployments where
it's gotta be local, it's gotta be
541
00:29:30,781 --> 00:29:35,541
your hardware, your stuff so that you
fully own and control the process.
542
00:29:35,541 --> 00:29:39,541
'cause that's one of the few ways to
mitigate some of the risks that pop up.
543
00:29:39,691 --> 00:29:39,931
Hmm.
544
00:29:40,291 --> 00:29:40,471
Yeah.
545
00:29:40,756 --> 00:29:49,036
I, I think too, you run the risk of
maybe using a data source that to
546
00:29:49,036 --> 00:29:53,546
a human doing it would, be against
the rules then you're running into a
547
00:29:53,546 --> 00:29:57,596
situation where it's evidence is being
thrown out because it's the fruit
548
00:29:57,596 --> 00:30:01,436
of the poisonous tray and everything
from that point on gets thrown out.
549
00:30:01,596 --> 00:30:03,186
So you definitely don't
want to be in that.
550
00:30:03,441 --> 00:30:04,161
Situation.
551
00:30:04,161 --> 00:30:08,871
And then that's, that's where, going
back to what you mentioned before about
552
00:30:09,201 --> 00:30:18,011
attorneys and judges not being happy with,
using AI and trying to use AI in court.
553
00:30:18,131 --> 00:30:19,001
Yeah, for sure.
554
00:30:19,001 --> 00:30:21,881
It's, it's, it's, there's no,
like, there's no path forward.
555
00:30:21,881 --> 00:30:22,241
I think.
556
00:30:22,241 --> 00:30:26,021
Like there's no, no one has figured
out like the, the golden ticket
557
00:30:26,021 --> 00:30:28,811
on this or like, what's, what's
the ideal state essentially.
558
00:30:29,091 --> 00:30:33,621
So we see a lot of departments just
experimenting with what they think works.
559
00:30:33,931 --> 00:30:34,681
But I think.
560
00:30:34,911 --> 00:30:36,861
There's still a lot of
consequences we've yet to see.
561
00:30:36,861 --> 00:30:38,691
I think a lot of problems still have to
562
00:30:38,781 --> 00:30:38,901
mm-hmm.
563
00:30:39,331 --> 00:30:40,231
Make their way.
564
00:30:40,411 --> 00:30:43,531
A lot of cracks in the veneer
are still gonna show up before we
565
00:30:43,531 --> 00:30:47,581
figure out how to really use this
technology correctly in our space.
566
00:30:47,731 --> 00:30:48,031
Yeah.
567
00:30:48,481 --> 00:30:57,001
Now, for those, department that have used
their own, LLM , do you think they're.
568
00:30:57,621 --> 00:31:01,011
Completely custom building
this or are they working with
569
00:31:01,011 --> 00:31:02,961
vendors to help them build the
570
00:31:03,141 --> 00:31:04,161
LLM?
571
00:31:04,251 --> 00:31:07,521
Most of the ones I've seen
are, are vendor built.
572
00:31:07,521 --> 00:31:08,571
So a vendor mm-hmm.
573
00:31:08,811 --> 00:31:11,331
Comes in and says, we'll kind
of walk you through the process.
574
00:31:11,331 --> 00:31:14,661
We'll, we'll do most of the setup, but
we'll do it on your infrastructure.
575
00:31:14,951 --> 00:31:15,371
I've seen.
576
00:31:16,006 --> 00:31:19,786
Some vendors that offer like, Hey,
we'll set it up in the cloud, but it
577
00:31:19,786 --> 00:31:22,246
will be an isolated cloud environment.
578
00:31:22,276 --> 00:31:27,376
It'll be on gov cloud, which is
like Amazon's government designated
579
00:31:27,376 --> 00:31:31,316
cloud infrastructure that is
automatically CGIs compliant.
580
00:31:31,676 --> 00:31:34,646
Some people are like, Hey, we'll
put it in the same cloud setup
581
00:31:34,646 --> 00:31:37,106
that like the NSA and the CAA use.
582
00:31:37,106 --> 00:31:39,806
If that makes you feel better,
like if it's good enough for them,
583
00:31:39,806 --> 00:31:41,126
maybe it's good enough for you.
584
00:31:41,421 --> 00:31:46,581
So everyone selling this stuff is,
is hitting the same roadblock I think
585
00:31:46,581 --> 00:31:50,661
with departments and agencies of like,
we have a lot of concerns on the,
586
00:31:50,661 --> 00:31:52,581
the access and fidelity of the data.
587
00:31:52,881 --> 00:31:55,896
And so companies are, are jumping through
any and all hoops to try to make it.
588
00:31:56,596 --> 00:32:00,066
Appealing, I think, to keep
it rolling out to people.
589
00:32:00,066 --> 00:32:05,776
But truly I think the safest bet if,
if you're worried about access controls
590
00:32:05,776 --> 00:32:09,146
and if you're worried about fidelity
of the data and who might see it
591
00:32:09,146 --> 00:32:13,376
and who might touch it, you doing it
locally is really your only option.
592
00:32:13,426 --> 00:32:13,746
Mm-hmm.
593
00:32:13,831 --> 00:32:16,731
I have heard of a couple departments,
this is why I don't wanna name
594
00:32:16,731 --> 00:32:20,181
them 'cause I would never recommend
anyone do this, but they've basically
595
00:32:20,181 --> 00:32:22,101
just stood up ad hoc systems so
596
00:32:22,101 --> 00:32:24,891
.
, They get a, small server
or random computer.
597
00:32:25,111 --> 00:32:31,741
There's a, a website called hugging face,
and it's where lots of open source AI
598
00:32:31,741 --> 00:32:33,571
models are, are developed and shared.
599
00:32:33,571 --> 00:32:38,281
And you can go download a model and
install it in five minutes or less, some
600
00:32:38,281 --> 00:32:42,901
of these and have your own like, full
fledged AI or large, large language model.
601
00:32:43,466 --> 00:32:44,756
System up and running.
602
00:32:44,826 --> 00:32:48,006
And all you gotta do is train it,
pump some data into it, and then
603
00:32:48,006 --> 00:32:49,206
you're like off to the races.
604
00:32:49,206 --> 00:32:52,836
So I've, I've heard of some people
who have experimented with that.
605
00:32:52,836 --> 00:32:55,116
I don't know how official
or unofficial that is.
606
00:32:55,121 --> 00:32:55,331
Mm-hmm.
607
00:32:55,456 --> 00:32:57,226
I think it skews more unofficial.
608
00:32:57,506 --> 00:33:01,796
But I, I know lots of people are
toying with the ideas for sure.
609
00:33:02,486 --> 00:33:03,296
That makes sense.
610
00:33:03,296 --> 00:33:06,506
And that's really, I think, par
for the course for a lot of.
611
00:33:06,901 --> 00:33:08,041
Police departments.
612
00:33:08,091 --> 00:33:10,521
. You think back to like records
management systems, yes.
613
00:33:10,521 --> 00:33:13,431
They could build their own
homegrown records management
614
00:33:13,431 --> 00:33:15,231
system, but most departments don't.
615
00:33:15,501 --> 00:33:21,321
They have a vendor do that and they make
sure they lean on the vendor to come up
616
00:33:21,321 --> 00:33:26,871
with the specs and come up with all the
requirements and stay within the law.
617
00:33:26,871 --> 00:33:30,771
And that's more on the, the vendor
side and not necessarily the
618
00:33:30,771 --> 00:33:33,141
department side, so that, that.
619
00:33:33,476 --> 00:33:38,036
Tracks with what I've seen
in, in other situations.
620
00:33:38,216 --> 00:33:42,626
So I guess let's stay there a
little bit then with the analyst.
621
00:33:42,746 --> 00:33:47,186
Maybe their department is, in the
beginning stages of hiring one of
622
00:33:47,186 --> 00:33:51,136
these vendors to create an LLM.
623
00:33:51,206 --> 00:33:54,296
What would be your
recommendation to that analyst?
624
00:33:54,696 --> 00:33:58,476
As they're trying, just trying
to be useful, trying to see
625
00:33:58,476 --> 00:34:00,366
where they can, can be helpful.
626
00:34:00,456 --> 00:34:01,656
Yeah, it's a great question.
627
00:34:01,656 --> 00:34:06,156
I think one of the, the best value
adds that an analyst brings to that,
628
00:34:06,156 --> 00:34:09,786
and I think this applies to any, any
new technology getting rolled out in
629
00:34:09,786 --> 00:34:15,696
a department is like really helping
articulate and map out the use cases.
630
00:34:15,806 --> 00:34:19,216
Ideally like everyone in a, in a
department can do that, but you'd
631
00:34:19,216 --> 00:34:20,596
be surprised how many people.
632
00:34:20,896 --> 00:34:24,166
Just can't think outside the box
or they're so, they're so focused
633
00:34:24,166 --> 00:34:28,156
on like, their one singular process
or, or their one form that they fill
634
00:34:28,156 --> 00:34:30,076
out or, or what they've always done.
635
00:34:30,376 --> 00:34:34,306
I think analysts by nature are really
good at thinking through complex systems
636
00:34:34,306 --> 00:34:38,216
and thinking outside the box and thinking
bigger than themselves sometimes.
637
00:34:38,216 --> 00:34:38,926
So I think number one.
638
00:34:39,636 --> 00:34:44,016
Really helping working with your, your
department and the vendor to really think
639
00:34:44,016 --> 00:34:47,766
through, here's all the different use
cases, here's the different ways people
640
00:34:47,766 --> 00:34:50,166
use this data and access these systems.
641
00:34:50,166 --> 00:34:53,286
And these are the variety
of processes we have.
642
00:34:53,536 --> 00:34:57,586
'Cause the better you understand
all of that, the more useful you
643
00:34:57,586 --> 00:35:01,626
can make an artificial intelligence
based system or software.
644
00:35:01,996 --> 00:35:04,996
I think the second one is definitely.
645
00:35:05,531 --> 00:35:06,731
Asking questions.
646
00:35:06,881 --> 00:35:11,381
As silly as that sounds, I think my
experience in public safety has been
647
00:35:11,701 --> 00:35:16,831
a sometimes particularly leadership is
more easily swayed by the sales pitch.
648
00:35:17,111 --> 00:35:19,931
And when it's really good and
it's really shiny and it, and it
649
00:35:19,931 --> 00:35:23,321
seems like it does everything,
they're ready to to go with it.
650
00:35:23,381 --> 00:35:26,901
I think analysts are really good
at challenging those assumptions
651
00:35:26,901 --> 00:35:28,131
sometimes to say, well.
652
00:35:28,496 --> 00:35:29,936
Is it gonna work in this scenario?
653
00:35:29,936 --> 00:35:33,176
Or, Hey, is it gonna work for us
when we deploy it at different
654
00:35:33,176 --> 00:35:34,766
offices or different precincts?
655
00:35:35,066 --> 00:35:37,766
Or, Hey, did anyone think
through this type of data?
656
00:35:37,766 --> 00:35:40,806
Which wasn't in scope
in, in the original plan.
657
00:35:41,046 --> 00:35:43,926
So I think that's the other thing
that analysts have the potential
658
00:35:43,926 --> 00:35:48,656
to bring to the table is asking the
right types of questions to make sure
659
00:35:48,656 --> 00:35:52,756
you're getting the value that the
vendor is, is allegedly promising you.
660
00:35:52,966 --> 00:35:53,176
Hmm.
661
00:35:53,506 --> 00:35:53,896
Yeah.
662
00:35:54,241 --> 00:35:58,591
And, and in that, I think in that
regard, it becomes, I mean, just becomes
663
00:35:58,591 --> 00:36:00,811
part of the CAD or the pro process.
664
00:36:00,811 --> 00:36:05,521
If, if you went through that with CAD
or RMS is probably going to, there's
665
00:36:05,521 --> 00:36:07,651
probably gonna be a lot of similarities.
666
00:36:07,831 --> 00:36:09,931
And I, I'm a big.
667
00:36:10,336 --> 00:36:17,776
Advocate of not only do you talk about
implementation you talk about how it's
668
00:36:17,776 --> 00:36:24,346
going to fit into the day-to-day operation
for, for the police department, something
669
00:36:24,346 --> 00:36:26,026
like a records management system.
670
00:36:26,026 --> 00:36:27,496
It's, it's usually taken care of.
671
00:36:27,526 --> 00:36:31,156
'cause that's your, the reason you're
using it is to manage all your records.
672
00:36:31,156 --> 00:36:33,226
So you have data entry folks.
673
00:36:33,316 --> 00:36:38,736
But I, mention a couple of times that
when I was at Cincinnati supported getting
674
00:36:38,786 --> 00:36:44,456
a analyst software we got the software,
but I didn't think it ever really was
675
00:36:44,516 --> 00:36:46,526
fully implemented into the day-to-day.
676
00:36:47,256 --> 00:36:51,006
Operations or the day-to-day
activity of, the analysts.
677
00:36:51,216 --> 00:36:55,926
So I always felt that I was a
failure on, on that aspect of it.
678
00:36:55,926 --> 00:36:57,366
So that's down the road too.
679
00:36:57,996 --> 00:37:01,526
And man, Chris, just like any
software though with police
680
00:37:01,526 --> 00:37:06,116
departments, I mean, you really
have to have folks, good advisors,
681
00:37:06,266 --> 00:37:09,746
decision makers, being able to see.
682
00:37:10,106 --> 00:37:15,596
25 steps down the road as you're
trying to evaluate this and how
683
00:37:15,596 --> 00:37:20,816
it's going to play out, and that's,
that can be really difficult.
684
00:37:21,196 --> 00:37:23,926
Especially with such a new technology.
685
00:37:24,046 --> 00:37:25,006
Yeah, a hundred percent.
686
00:37:25,036 --> 00:37:28,156
, In the IT world there's a whole
group of people that, that's their
687
00:37:28,156 --> 00:37:32,916
job essentially is planning and
shepherding projects and initiatives.
688
00:37:32,996 --> 00:37:37,056
That are specific to a division or
a department or a function they're
689
00:37:37,056 --> 00:37:40,246
typically called it business
analysts essentially, right?
690
00:37:40,246 --> 00:37:45,046
Their job is to assess the current and
future state of the technology to do
691
00:37:45,046 --> 00:37:49,186
the job that other people need to do,
like really being that tech advocate.
692
00:37:49,486 --> 00:37:53,146
And that's unfortunately why I
think we see the analyst as the
693
00:37:53,146 --> 00:37:55,876
technologist kind of bringing it
full circle on that one is that.
694
00:37:56,206 --> 00:37:58,846
I think of everyone who's
typically involved in the process.
695
00:37:58,846 --> 00:38:01,576
Hopefully you've got good IT people
in your department and sometimes
696
00:38:01,576 --> 00:38:03,416
we do but sometimes we don't.
697
00:38:03,446 --> 00:38:07,656
And so the analyst the crime analyst, the
intel analyst, ends up being sometimes
698
00:38:07,656 --> 00:38:12,286
that it BA person because they're
the only ones that are like mentally
699
00:38:12,286 --> 00:38:14,776
trained to think 25 steps ahead on.
700
00:38:14,776 --> 00:38:18,226
Like, is this like, is this a
real solution or is this just a
701
00:38:18,226 --> 00:38:19,756
tool we're putting on the floor?
702
00:38:19,786 --> 00:38:21,166
'cause we think that's where it goes.
703
00:38:21,446 --> 00:38:21,926
I think.
704
00:38:22,331 --> 00:38:26,021
Again, that's just a hat we end up wearing
even when we don't want to sometimes.
705
00:38:26,171 --> 00:38:26,501
Yeah.
706
00:38:26,681 --> 00:38:27,011
Alright.
707
00:38:27,061 --> 00:38:33,031
Anything else in terms of the
near future of AI and analysis,
708
00:38:33,151 --> 00:38:33,756
do you wanna mention, do.
709
00:38:34,506 --> 00:38:36,216
, We've seen a lot of change already.
710
00:38:36,276 --> 00:38:40,236
I think we're gonna see more and more,
I think governments and, and public
711
00:38:40,236 --> 00:38:45,126
safety as a whole have been one of the
groups more hesitant to adopt this stuff.
712
00:38:45,126 --> 00:38:49,546
I think we're finally seeing that
hesitancy peak, and I think well,
713
00:38:49,546 --> 00:38:51,346
I'll, I'll, I'll throw the gauntlet.
714
00:38:51,346 --> 00:38:53,521
I think like 2026 will
be a year where we see.
715
00:38:54,226 --> 00:38:57,966
A lot of rapid development and
adoption in law enforcement agencies.
716
00:38:57,966 --> 00:39:00,326
I think it's been around just enough.
717
00:39:00,326 --> 00:39:03,266
I think there's starting to be some
use cases out there that we're gonna
718
00:39:03,266 --> 00:39:05,556
see a lot of agencies leaning into it.
719
00:39:05,866 --> 00:39:08,716
And I think unfortunately though,
one of the things that's gonna drive
720
00:39:08,716 --> 00:39:12,796
that is, a lot of the staffing gaps
that agencies are still dealing with.
721
00:39:12,796 --> 00:39:17,156
I think the, the ultimate question
is gonna be, I use AI to fill gaps
722
00:39:17,156 --> 00:39:18,716
that are left by the vacancies.
723
00:39:18,716 --> 00:39:21,416
I can't fill, whether
that's sworn or civilian.
724
00:39:21,666 --> 00:39:26,236
What kind of optimization can AI get the
agency or the department so we can keep
725
00:39:26,236 --> 00:39:28,456
doing the job or, or do the job better?
726
00:39:28,786 --> 00:39:31,366
I think that's what 2026 is gonna show us.
727
00:39:31,366 --> 00:39:34,646
And my hope is that analysts see this.
728
00:39:34,956 --> 00:39:38,386
This technology taking off and
decide to jump on for the ride
729
00:39:38,386 --> 00:39:39,736
and then see where it takes them.
730
00:39:39,926 --> 00:39:42,836
'Cause I think in a lot of
instances it won't be optional.
731
00:39:43,646 --> 00:39:48,091
I was I, I was trying to see about.
732
00:39:48,451 --> 00:39:53,401
I had a coworker mention this the other
day and I was quickly trying to Google it.
733
00:39:53,611 --> 00:39:58,861
I thought he said that 30% of
calls now are handled by ai.
734
00:39:59,021 --> 00:40:01,571
I can't, I can't, I can't find
it right away, so I'm not sure.
735
00:40:01,631 --> 00:40:04,351
I'm not for sure , on that number, but.
736
00:40:05,546 --> 00:40:11,966
Certainly you're, you can see where if
you're told to do more with less and
737
00:40:11,966 --> 00:40:14,396
it, and this it's always interesting.
738
00:40:14,396 --> 00:40:18,176
If you ever, anybody that's ever
been part of police departments
739
00:40:18,176 --> 00:40:24,176
and budget discussions, you know
that the, employees, the humans.
740
00:40:24,586 --> 00:40:29,946
Or the biggest part of the budget,
it's like 97% of the budget is, the
741
00:40:29,946 --> 00:40:33,516
salaries that go to the humans or
the benefits that go to the humans.
742
00:40:33,696 --> 00:40:34,116
Right.
743
00:40:34,526 --> 00:40:39,386
Maybe 97 is a little high, but I mean
it's still, the vast majority is of the
744
00:40:39,386 --> 00:40:42,746
budget deals with salary and compensation.
745
00:40:42,746 --> 00:40:47,936
So if you're being told to do more
with less and you, you get something
746
00:40:47,936 --> 00:40:52,736
where there's grants or other
opportunities to get a technology,
747
00:40:52,766 --> 00:40:57,986
that also feeds into this idea of
like, oh, well, we're below capacity.
748
00:40:58,546 --> 00:41:03,046
You won't let us hire, but you'll
give us money for technology.
749
00:41:03,046 --> 00:41:05,936
This is going to also feed this fire.
750
00:41:06,086 --> 00:41:06,926
Yeah, for sure.
751
00:41:06,926 --> 00:41:11,086
And I think I don't know the numbers,
but I know in Virginia there are
752
00:41:11,086 --> 00:41:16,576
a few localities that have piloted
or are currently piloting ai to
753
00:41:16,576 --> 00:41:18,586
improve their 9 1 1 processes.
754
00:41:18,586 --> 00:41:19,426
I know Arlington.
755
00:41:19,901 --> 00:41:23,091
County in Virginia has done
some work in that space.
756
00:41:23,181 --> 00:41:26,551
I know Fairfax County was I think
recently in the news as either
757
00:41:26,761 --> 00:41:28,261
doing it or rolling it out.
758
00:41:28,601 --> 00:41:34,401
And I think Virginia Beach as well
has done some ai like AI stuff.
759
00:41:34,401 --> 00:41:36,351
I think it was like non-emergency
calls they were doing.
760
00:41:36,351 --> 00:41:41,061
Were going to AI first to kind of
alleviate the burden on 9 1 1 dispatchers.
761
00:41:41,646 --> 00:41:42,486
All right, Chris.
762
00:41:42,516 --> 00:41:43,686
Very good.
763
00:41:43,716 --> 00:41:44,676
This has been great.
764
00:41:45,151 --> 00:41:46,801
Great to catch up with you.
765
00:41:46,801 --> 00:41:52,321
I was, I was shocked to learn when I went
back and thought, I knew I had you on the
766
00:41:52,321 --> 00:41:54,631
show, but I didn't realize it was 2022.
767
00:41:54,631 --> 00:41:58,351
I was like, oh, it was almost
four years ago that I had you on.
768
00:41:58,351 --> 00:42:03,991
So it's shocked me that time had
gone by , so much between us talking.
769
00:42:03,991 --> 00:42:08,011
So I definitely want to have
you on again, talk more ai.
770
00:42:08,786 --> 00:42:09,866
With you.
771
00:42:09,956 --> 00:42:16,236
So for the listeners, be, sure to look
for another episode with Chris and I
772
00:42:16,236 --> 00:42:22,206
talking ai Chris, for the listener,
if they need or want to contact you.
773
00:42:22,206 --> 00:42:23,706
What's the best way to contact you?
774
00:42:23,971 --> 00:42:27,361
Yeah, so you can find me on LinkedIn.
775
00:42:27,421 --> 00:42:28,501
I'm pretty easy to find there.
776
00:42:28,501 --> 00:42:30,691
Christopher Cruz in Richmond, Virginia.
777
00:42:30,921 --> 00:42:34,521
You can also find me I work for the
Virginia State Police so you can look
778
00:42:34,521 --> 00:42:36,951
me up there and find my contact info.
779
00:42:37,641 --> 00:42:37,701
Yeah.
780
00:42:37,701 --> 00:42:40,581
And do you have any
upcoming presentations or.
781
00:42:41,191 --> 00:42:42,001
Or public events?
782
00:42:42,181 --> 00:42:44,401
I don't think I have any upcoming ones.
783
00:42:44,401 --> 00:42:46,561
I literally just did one yesterday.
784
00:42:46,711 --> 00:42:46,771
Yeah.
785
00:42:46,801 --> 00:42:50,331
I did a, a presentation on
social engineering attacks.
786
00:42:50,331 --> 00:42:51,561
Very cyber, that one.
787
00:42:51,561 --> 00:42:53,726
But yeah, I don't have anything
upcoming at the moment.
788
00:42:54,531 --> 00:42:54,981
For once.
789
00:42:56,091 --> 00:42:59,121
No, it's, and that's interesting, when
I asked you that question, it sounded
790
00:42:59,121 --> 00:43:01,161
like you were like a standup comedian.
791
00:43:01,161 --> 00:43:04,611
Where can people come see you if they
didn't wanna come see you kind of thing.
792
00:43:04,701 --> 00:43:08,271
So, man, that would be interesting
to have that many events where
793
00:43:08,271 --> 00:43:09,411
you're like, rattling off.
794
00:43:09,411 --> 00:43:11,811
Yeah, you can come see me at
these five different cities.
795
00:43:13,701 --> 00:43:16,701
Alright, Chris, I'll give
you the last word again.
796
00:43:16,701 --> 00:43:17,811
Thank you for your time.
797
00:43:17,811 --> 00:43:19,101
It was great catching up with you.
798
00:43:19,101 --> 00:43:23,461
This is a great perspective., What's
your last word for our listeners today?
799
00:43:23,611 --> 00:43:26,561
Yeah, I think AI is not
here to replace analysts.
800
00:43:26,611 --> 00:43:28,411
I think it's about amplifying impact.
801
00:43:28,481 --> 00:43:30,861
But it takes the analysts
wanting that to happen.
802
00:43:30,861 --> 00:43:31,551
For it to happen.
803
00:43:31,551 --> 00:43:33,591
So if you're out there
and you're listening.
804
00:43:33,876 --> 00:43:36,066
Want it to happen and
it'll happen for you.
805
00:43:36,276 --> 00:43:36,996
Very good.
806
00:43:36,996 --> 00:43:39,366
Thank You again, Chris, and you be safe.
807
00:43:39,456 --> 00:43:40,056
Thanks Jason.
808
00:43:40,056 --> 00:43:40,536
Really appreciate it.
809
00:43:40,704 --> 00:43:43,074
Thank you for making it to
the end of another episode of
810
00:43:43,074 --> 00:43:44,484
Analyst Talk with Jason Elder.
811
00:43:44,514 --> 00:43:47,994
You can show your support by sharing
this in other episodes found on
812
00:43:47,994 --> 00:43:51,024
our website at www dot podcasts.
813
00:43:51,499 --> 00:43:52,279
Dot com.
814
00:43:52,489 --> 00:43:55,609
If you have a topic you would like
us to cover or have a suggestion
815
00:43:55,609 --> 00:43:58,059
for our next guest, please send us
an email at LEAPpodcasts@gmail.com
816
00:44:00,619 --> 00:44:01,459
next time analysts.
817
00:44:01,609 --> 00:44:02,299
Keep talking.