WEBVTT

1
00:00:00.160 --> 00:00:06.480
Invent every industry industry. Inside Analysis
is your source of information and insight about

2
00:00:06.480 --> 00:00:10.640
how to make the most of this
exciting new era. Learn more and Inside

3
00:00:10.679 --> 00:00:15.640
analysis dot com, Inside Analysis dot
com and now here's your host, Eric

4
00:00:15.880 --> 00:00:23.480
Kavanaugh. Well, ladies and gentlemen, Hello, and welcome back once again

5
00:00:23.640 --> 00:00:27.320
to the only coast to coast radio
show in the USFA that's all about the

6
00:00:27.359 --> 00:00:32.359
information economy. It's time for Inside
Analysis or truly Eric Kavanaugh here and folks.

7
00:00:32.399 --> 00:00:36.600
I am very excited to have Frankly, a legend in the industry with

8
00:00:36.679 --> 00:00:39.560
us. Today we're going to be
talking with Michael Berthold. He is the

9
00:00:39.640 --> 00:00:45.159
founder and CEO of a company called
NIME that's spelled k n i M and

10
00:00:45.200 --> 00:00:50.280
they are an open source analytics and
data science platform, a visual platform for

11
00:00:50.359 --> 00:00:53.520
doing data science, which is good
stuff because even though there are lots of

12
00:00:53.520 --> 00:00:57.479
people who can code very well,
almost anyone can look at visuals to move

13
00:00:57.520 --> 00:01:00.799
boxes around on a screen. So
that's what they figured out. So at

14
00:01:00.799 --> 00:01:03.719
that, Michael, welcome to Inside
Analysis. Tell us a bit about yourself

15
00:01:03.759 --> 00:01:07.040
and NIME in which you folks are
working on these days. Thanks thanks for

16
00:01:07.079 --> 00:01:11.840
the invite, Eric and having us
run the show. As you already said,

17
00:01:11.920 --> 00:01:15.920
NIME is very much about visual programming
low code for data science. And

18
00:01:15.959 --> 00:01:19.599
we started that many years ago as
a platform really for as a workbench almost

19
00:01:19.599 --> 00:01:23.599
at my group at the University of
University of Constants to be able to kind

20
00:01:23.599 --> 00:01:27.840
of deploy our research results to the
real world to practitioners wanting to use that.

21
00:01:29.400 --> 00:01:34.400
And it's grown from there to become
one of the only open source visual

22
00:01:34.560 --> 00:01:38.760
data flow platforms for doing anything you
want to do with data. And I

23
00:01:38.760 --> 00:01:42.560
mean I kind of I'm always a
little bit careful about calling it data science

24
00:01:42.640 --> 00:01:47.400
because that often scares people because they
say, I just want to do data

25
00:01:47.400 --> 00:01:49.159
wrangling. I don't care about the
science seed beats, and that's kind of

26
00:01:49.200 --> 00:01:52.799
scares them off. And that's what
Nie does as well. So a lot

27
00:01:52.799 --> 00:01:56.760
of applications that we see in real
life is just in large airports, just

28
00:01:57.359 --> 00:02:01.200
getting data in the right shapes from
many different sources. Yeah, and I

29
00:02:01.280 --> 00:02:05.079
will tell you I'm a huge fan
of open source. In fact, we

30
00:02:05.120 --> 00:02:07.879
built a website a number of years
ago and we feed it with a technology

31
00:02:07.879 --> 00:02:13.280
that we built called media Lens and
it's called inside open source. And the

32
00:02:13.319 --> 00:02:15.759
reason we launched it was because I
realized there is so much happening in the

33
00:02:15.800 --> 00:02:20.680
open source world. There's the Apache
Foundation, as the Linux Foundation. There

34
00:02:20.680 --> 00:02:24.479
are lots of other projects outside of
those organizations as well, But in the

35
00:02:24.520 --> 00:02:30.840
origin of open source I found fascinating. So I first started researching this in

36
00:02:30.879 --> 00:02:35.080
two thousand and five when I was
working for the Data Warehousing Institute as their

37
00:02:35.319 --> 00:02:39.280
web evangelist, and Katrina had just
struck New Orleans. I just moved out

38
00:02:39.280 --> 00:02:43.199
of New Orleans, so we watched
it all happen on TV. I was

39
00:02:43.280 --> 00:02:47.360
just, of course terrifying. But
I remember that the senators from Louisiana asked

40
00:02:47.400 --> 00:02:53.560
for a quarter of a trillion dollars
to rebuild southern Louisiana. And I happened

41
00:02:53.599 --> 00:02:57.319
to know through a past life and
past clients in the government space down there,

42
00:02:57.319 --> 00:03:00.719
that the politicians are good at making
money peer and I thought to myself,

43
00:03:00.719 --> 00:03:04.919
this is a very bad situation,
because if all that money just floods

44
00:03:04.919 --> 00:03:07.879
in, it's going to flood right
out, not where anyone intended, and

45
00:03:07.080 --> 00:03:09.719
a lot of it is going to
disappear. So I went on this high

46
00:03:09.759 --> 00:03:14.800
horse, if you will, and
started doing research, and it took me

47
00:03:14.879 --> 00:03:19.599
to open source and I put forth
this theory about open source government, about

48
00:03:19.599 --> 00:03:23.719
publishing all the government data such the
citizens can see where the money goes and

49
00:03:23.800 --> 00:03:28.280
understand. And I basically said,
look, with the sarbainez Ox the Act

50
00:03:28.319 --> 00:03:31.199
in the States, which came out
of the Enron debacle, corporations had to

51
00:03:31.319 --> 00:03:35.560
document their processes for how they come
up with their numbers and had to be

52
00:03:35.639 --> 00:03:38.879
very transparent about that stuff. And
I thought, well, if corporate America

53
00:03:38.960 --> 00:03:40.759
has to do it, why doesn't
the government do it as well? And

54
00:03:40.800 --> 00:03:45.960
people thought I was crazy, But
a couple of things happened that were amazing.

55
00:03:46.319 --> 00:03:49.479
One guy paid attention. Out of
forty thousand people, I emailed,

56
00:03:49.520 --> 00:03:53.039
one guy paid attention, and he
worked for the Heritage Foundation. He went

57
00:03:53.159 --> 00:03:57.599
and talked to He basically testified to
Congress and said, we can have citizen

58
00:03:57.639 --> 00:04:00.960
auditors. This stuff really can happen. And he really leveraged the imprimiture of

59
00:04:00.039 --> 00:04:04.319
TDWY and sure enough they did it. House past the bill, Senate passed

60
00:04:04.319 --> 00:04:08.680
the bill co sponsored by a guy
named Barack Obama who was a Senator for

61
00:04:08.719 --> 00:04:11.599
Illinois, and then President George W. Bush, believe it or not,

62
00:04:11.759 --> 00:04:15.319
signed the Federal Funding Accountability and Transparency
Act in September six, twenty six,

63
00:04:15.399 --> 00:04:18.680
two thousand and six, and I
almost fell off my chair. I was

64
00:04:18.720 --> 00:04:23.199
like, Wow, they actually did
it. But the reason I bring this

65
00:04:23.360 --> 00:04:27.079
up is to talk about the power
of open source. And I realized at

66
00:04:27.079 --> 00:04:32.560
the time the Apache web server had
just surpassed the Microsoft Web Server as the

67
00:04:32.639 --> 00:04:35.319
number one web server. And I
thought to myself, well, that's very

68
00:04:35.319 --> 00:04:40.079
interesting. And I was also studying
service oriented architecture at the time, and

69
00:04:40.120 --> 00:04:43.560
I thought, well, if you
have all this open source code and you

70
00:04:43.560 --> 00:04:46.079
have a service oriented architecture, you
should be able to plug and play and

71
00:04:46.120 --> 00:04:50.319
sort of take stuff out and put
stuff in and have a very composable environment.

72
00:04:50.600 --> 00:04:53.639
And I thought, well, that's
not going to be very good news

73
00:04:53.680 --> 00:04:57.720
for the SAPs and the Oracles of
the world, because they liked the monolith.

74
00:04:57.759 --> 00:05:00.480
They like control, and they have
control of all that stuff. And

75
00:05:00.720 --> 00:05:04.040
it took longer than I thought,
but about ten years later open source just

76
00:05:04.319 --> 00:05:10.279
blew up the market with haddoob and
with the Kafka And of course you have

77
00:05:10.399 --> 00:05:15.000
NIME. So tell me a bit
about the open source foundation of NIME and

78
00:05:15.279 --> 00:05:20.600
what drove your decision there and what
it means for your customers. That's a

79
00:05:20.759 --> 00:05:26.759
very interesting story. I didn't know
about that open data movement coming from Katrina

80
00:05:26.800 --> 00:05:30.319
break in the old is so NIME
is a bit atypical in its open source

81
00:05:30.399 --> 00:05:33.839
models. I mean fundamentally, there
are really three different ways to do that.

82
00:05:33.920 --> 00:05:39.240
You can have distributions like Linox,
and you're essentially making money by packaging

83
00:05:39.279 --> 00:05:43.439
them up nicely and supporting them,
but essentially there's not really new code that

84
00:05:43.439 --> 00:05:46.720
you're aiding around it. Not quite
true. I mean they are little installers

85
00:05:46.720 --> 00:05:49.240
and that type of stuff, but
that's fundamentally the idea behind distributions. You

86
00:05:49.240 --> 00:05:53.519
can also have what people often refer
to as open core, where you have

87
00:05:53.600 --> 00:05:57.040
something that you open source, but
it's kind of it's more of a teaser,

88
00:05:57.199 --> 00:05:59.720
it's more baitware, and if you
really want to use this in production,

89
00:05:59.759 --> 00:06:02.199
you have to buy some commercial bits
and pieces around it. And then

90
00:06:02.240 --> 00:06:06.120
you have, of course also the
Mango debs that are essentially databases that type

91
00:06:06.120 --> 00:06:10.040
of stuff. Also maybe even a
data bax with Spark. They have really

92
00:06:10.040 --> 00:06:13.519
cool open source technology and they make
it accessible to you in the cloud.

93
00:06:14.240 --> 00:06:17.519
For charge Time is different in that
we have one open source piece that's the

94
00:06:17.720 --> 00:06:24.120
analytics platform that allows anybody who wants
to build these workflows and execute these workflows

95
00:06:24.160 --> 00:06:27.600
and pretty much do anything they want
to do with data. And then we

96
00:06:27.680 --> 00:06:31.360
have a commercial complement software to that, which we call the Nine Business Hub,

97
00:06:31.759 --> 00:06:36.160
which allows people to productionize that and
collaborate. So when you have more

98
00:06:36.199 --> 00:06:40.839
than one person using the analytics platform
in your organization and you want to deploy

99
00:06:40.920 --> 00:06:44.439
that as a web interface or a
rest service, or you want to collaborate

100
00:06:44.920 --> 00:06:47.920
and have compliance and governments kind of
features, that's when you buy the hub

101
00:06:47.959 --> 00:06:54.120
from us. And the reason the
open source platform is open source. The

102
00:06:54.160 --> 00:06:58.720
open source analytics platform is open source
is I'm not too religious around it,

103
00:06:58.759 --> 00:07:01.680
but to me, it's in that
they science field in particular, you can't

104
00:07:02.079 --> 00:07:06.959
exist with proprietary software. There's so
much cool new stuff going on on a

105
00:07:08.040 --> 00:07:12.920
daily basis in resourched groups and other
types of environments that you're essentially standing on

106
00:07:12.959 --> 00:07:16.120
the shoulder of many, many giants. A lot of functionality inside nine is

107
00:07:16.160 --> 00:07:21.560
actually based on open source libraries,
so it seems kind of unfair almost to

108
00:07:21.600 --> 00:07:26.720
put that into a proprietary umbrella.
And also it enables us to be fair

109
00:07:26.800 --> 00:07:30.759
to get in in rolls, make
inrolls into academic into teaching environments much easier

110
00:07:30.800 --> 00:07:34.199
because they can just use the open
source platform for teaching. But it also

111
00:07:34.240 --> 00:07:40.839
we have lot of open source contributors
that are contributing additional functionality into the Linemen

112
00:07:41.319 --> 00:07:44.560
platform. So it's really I'd like
to see it as a win win situation

113
00:07:44.680 --> 00:07:48.800
also for our customers because they essentially
get a lot of maintained functionality from us.

114
00:07:49.240 --> 00:07:54.079
In addition, they have access to
all of those community functions out there.

115
00:07:55.560 --> 00:07:57.040
Yeah, it's very cool. I
mean, there are a lot of

116
00:07:57.040 --> 00:07:59.720
good things about open source. One
of the things that I've heard over the

117
00:07:59.800 --> 00:08:03.920
year is that bad code goes away
because all these eyes can see now.

118
00:08:05.240 --> 00:08:09.959
The one shortcoming up come across is
that the open source project gets to MVP

119
00:08:11.160 --> 00:08:15.000
status, if you will, minimum
viable product and then doesn't typically go past

120
00:08:15.079 --> 00:08:18.439
that just because it works now we
just kind of move beyond that. But

121
00:08:18.480 --> 00:08:24.079
what are your thoughts about that,
in particular about how you make sure that

122
00:08:24.160 --> 00:08:31.199
you have truly finished products and that
you're able to deliver robust platform analytics ongoing

123
00:08:31.319 --> 00:08:35.240
for all of your clients. How
much effort does that take internally from your

124
00:08:35.240 --> 00:08:39.840
developers to stay on top of the
platform and make sure everything's working. That's

125
00:08:39.879 --> 00:08:41.759
a very good compoint. I mean, I tend to joke that ninety nine

126
00:08:41.840 --> 00:08:46.440
percent of all PhD projects turn into
open source projects and then they kind of

127
00:08:46.480 --> 00:08:50.480
die away and feel away and never
really turn into something useful in production.

128
00:08:50.879 --> 00:08:56.080
Now I am a probably about half
of our development team, like eighty developers

129
00:08:56.080 --> 00:09:01.200
at nine are focused exclusively on the
Luditic platform and just making sure that core

130
00:09:01.720 --> 00:09:09.879
works and works in a professional environment. We have our own extensions, which

131
00:09:09.879 --> 00:09:13.919
are of course maintained by ourselves,
so we have the same quality assurance there.

132
00:09:13.399 --> 00:09:18.600
We have what we call trusted community
extensions, where we're in close collaboration

133
00:09:18.759 --> 00:09:22.200
with the community contributing those extensions,
so we can also make sure there's quality

134
00:09:22.240 --> 00:09:26.919
assurance there as well. And then
there is i'd call it the long tail

135
00:09:26.000 --> 00:09:31.000
of extensions that are experimental, right. The nice thing is that everybody can

136
00:09:31.240 --> 00:09:35.240
use those and play with those and
explore new technologies, and then when we

137
00:09:35.320 --> 00:09:39.919
see increased usage of some of these
experimental extensions, we can move them into

138
00:09:39.960 --> 00:09:43.840
the trusted extensions as well. Interesting
that makes a lot of sense. So

139
00:09:43.879 --> 00:09:48.120
you're trying out things, you've got
these extensions in trials basically, and then

140
00:09:48.159 --> 00:09:52.960
once you see there's a lot of
activity, then you throw some developer support

141
00:09:52.000 --> 00:09:56.159
behind it to harden it is the
term we typically use right to make it

142
00:09:56.559 --> 00:09:58.919
sure that it's bulletproof, that it
really does what you want it to do.

143
00:10:00.080 --> 00:10:03.080
That makes a lot of sense,
and you do end to end and

144
00:10:03.200 --> 00:10:07.600
NIME does everything from data ingestion,
data pipelines, number crunching, model building,

145
00:10:07.720 --> 00:10:11.360
all that kind of fun stuff is
in the nine analytics platform. Is

146
00:10:11.360 --> 00:10:15.080
that right, Yes, that's true. So we have everything from about the

147
00:10:15.080 --> 00:10:18.120
ETL part loading the data. We
can access about four under different data sources.

148
00:10:18.559 --> 00:10:22.639
We can access databases, strange file
formats. So of course we can

149
00:10:22.720 --> 00:10:28.519
also execute bits and pieces on different
execution environments like doing the ETL directly insider

150
00:10:28.639 --> 00:10:31.480
database or in snowflake, or in
data breaks or in our loop in the

151
00:10:31.480 --> 00:10:37.360
old days. And then we go
all the way to visualization the enderlytics functionality,

152
00:10:37.360 --> 00:10:39.279
and a lot of that, as
I said before, is of course

153
00:10:39.320 --> 00:10:43.799
based on each arts for the visualization, a lot of Pithen libraries see libraries,

154
00:10:45.320 --> 00:10:48.679
our libraries, Java libraries. For
some of the machine learning functionality,

155
00:10:48.159 --> 00:10:50.879
we have integrations with TensorFlow. If
you wanted to do that. We have

156
00:10:50.919 --> 00:10:56.639
integrations with the other deep learning libraries, we have integrations with XG boosts.

157
00:10:56.679 --> 00:11:01.120
Pretty much everything is in there.
But the other piece is so often when

158
00:11:01.159 --> 00:11:05.399
people talk about data science stainly mean
this kind of from the data to the

159
00:11:05.480 --> 00:11:09.799
reporter, to the endpoint or to
the model. But the business hub then

160
00:11:09.840 --> 00:11:13.600
also covers the rest of this journey, right, deploying into others, managing

161
00:11:13.600 --> 00:11:18.879
a three training models when needed,
and monitoring their their performance well. Right,

162
00:11:18.919 --> 00:11:22.559
because at the end of the day, you want these algorithms to connect

163
00:11:22.600 --> 00:11:28.519
into your business, whether it's spirit
of marketing or for manufacturing or supply chain

164
00:11:28.039 --> 00:11:33.440
or whatever it is. You want
it to affect some outcome in the business,

165
00:11:33.440 --> 00:11:37.120
and so that involves connecting to operational
systems, right. It involves connecting

166
00:11:37.120 --> 00:11:41.919
to EERP systems or CRM systems or
things of this nature. That's where the

167
00:11:41.919 --> 00:11:46.440
magic happens. And a lot of
times that's the hard part, right.

168
00:11:46.440 --> 00:11:50.440
I Mean, I've heard many stories
about models that just don't get deployed because

169
00:11:50.840 --> 00:11:54.919
maybe the companies didn't have the wherewithal
or they didn't have the expertise to do

170
00:11:54.960 --> 00:12:00.440
it properly. But being able to
plug the algorithms into operational systems and then

171
00:12:00.639 --> 00:12:03.679
monitor how those models perform and switch
them out right, because you've got your

172
00:12:03.919 --> 00:12:09.240
production model and have your challenger models
that are sitting at by the wayside waiting

173
00:12:09.279 --> 00:12:13.000
to get pulled in and being able
to switch over to a new model.

174
00:12:13.039 --> 00:12:18.480
When a model that's in production starts
faltering, that's a critical piece and that,

175
00:12:18.519 --> 00:12:20.720
I guess is that done in business
hub with you folks. That's something

176
00:12:20.759 --> 00:12:24.080
on the business subside exactly. So
as business hup, you can deploy models

177
00:12:24.120 --> 00:12:30.120
which really are deployed nine workflows.
You can deploy in them as a rest

178
00:12:30.120 --> 00:12:33.000
service or as a wet application,
and then people consume it, and you

179
00:12:33.039 --> 00:12:39.399
can constantly monitor what's happening in production
and then potentially replace retrain or just alert

180
00:12:39.399 --> 00:12:41.720
the data science team and just say, hey, this is so out of

181
00:12:41.720 --> 00:12:43.720
backwth reality. We don't really know
how to fix that, do something about

182
00:12:43.720 --> 00:12:48.120
it. The nice thing is that
you don't have to switch code in between.

183
00:12:48.200 --> 00:12:52.440
Right in the old days, always
somebody coded the model of the strains

184
00:12:52.480 --> 00:12:56.519
in some strange language and then it
was reprogrammed by it into some production language.

185
00:12:56.879 --> 00:13:01.840
On our case, on the hubside, workflow that was trained is also

186
00:13:01.879 --> 00:13:05.360
the one that runs in production.
That's interesting. So one of the other

187
00:13:05.440 --> 00:13:11.080
hurdles that people are running into is
when they use Jupiter notebooks to write their

188
00:13:11.120 --> 00:13:13.720
model or to build their model to
test with data, and then they want

189
00:13:13.759 --> 00:13:16.679
to go put that into production,
and it's just this step by step tedious

190
00:13:16.720 --> 00:13:22.480
process of copying over code and values
and all these things, and that falters

191
00:13:22.600 --> 00:13:24.639
often. That, from what I
understand, is a really serious problem.

192
00:13:24.919 --> 00:13:28.919
But I guess do you not have
that challenge because you're not using the Jupiter

193
00:13:28.960 --> 00:13:33.720
notebooks typically and people are just in
the environment in the analytics platform building out

194
00:13:33.759 --> 00:13:37.519
their models after they pull in their
data, et cetera. So you're already

195
00:13:39.240 --> 00:13:43.559
production ready when the process begins.
Is that about right? That's a very

196
00:13:43.639 --> 00:13:46.600
nice summary. Yes, So what
you use, what you use on the

197
00:13:46.600 --> 00:13:52.039
creation site when you actually train the
model is exactly that piece of the workflow

198
00:13:52.080 --> 00:13:54.840
gets then moved into production and executed
by exactly the same engine, so you

199
00:13:54.840 --> 00:14:00.519
don't also have a translation issue there. The other piece that people often lose

200
00:14:00.679 --> 00:14:05.200
in this going from training to production
is all of the feature engineering that you

201
00:14:05.279 --> 00:14:09.639
did, all the feature transformations that
you tend to lose them, so you

202
00:14:09.799 --> 00:14:11.759
only can take the model and move
that in production, but you can't do

203
00:14:11.840 --> 00:14:16.879
the transformations. And in lime you
can grab automatically the part of the DOPFOW

204
00:14:16.960 --> 00:14:22.080
that has the transformations and the model
applying the model, take that rep that

205
00:14:22.120 --> 00:14:26.120
automatically and deploy it to line business
up. Well, that's pretty cool.

206
00:14:26.559 --> 00:14:31.799
And you cover also two different industries, right, so you do insurance,

207
00:14:31.240 --> 00:14:37.279
healthcare. I would imagine financial services
all sorts of different industries because it's more

208
00:14:37.279 --> 00:14:41.080
of a horizontal solution instead about right, Yes, that's absolutely true. We

209
00:14:41.080 --> 00:14:46.039
have customers and users in pretty much
every industry. Yeah. Well, and

210
00:14:46.080 --> 00:14:48.879
we're going to talk about large language
models here in a minute in our next

211
00:14:48.039 --> 00:14:52.080
segment, But before we get there, I'll just throw out one of my

212
00:14:52.159 --> 00:14:56.120
theories to you and see what you
think about this. To me, this

213
00:14:56.480 --> 00:15:01.600
explosion of AI through foundations models,
including large language models, is really a

214
00:15:01.679 --> 00:15:07.039
major call to action for organizations to
get their data house in order. And

215
00:15:07.080 --> 00:15:11.799
what I mean is that data governance. If you don't have data governance,

216
00:15:11.840 --> 00:15:15.440
if you don't even know what data
governance is from an organizational perspective, you're

217
00:15:15.480 --> 00:15:20.679
going to have a hard time responsibly
leveraging AI. Would you agree that companies

218
00:15:20.759 --> 00:15:24.200
really do need to take a very
hard look at their end to end data

219
00:15:24.240 --> 00:15:28.399
management life cycles, processes, understand
governance, who gets access to what data?

220
00:15:28.840 --> 00:15:33.240
Even understanding a broad inventory of your
data sources, would you agree that

221
00:15:33.360 --> 00:15:39.639
is paramount to do that before pulling
the trigger on some AI. I totally

222
00:15:39.799 --> 00:15:43.039
agree. And the funny in a
way, it's funny that we've been preaching

223
00:15:43.080 --> 00:15:48.120
this for also data science processes for
a long time, this government's topic,

224
00:15:48.360 --> 00:15:52.600
and nobody really cared, and now
people really cared breaking up. Isn't that

225
00:15:52.639 --> 00:15:56.720
fascinating? I mean, I just
I've been in this business a long time.

226
00:15:56.759 --> 00:16:00.000
I've been talking about data governance,
analytics, AI, all this stuff

227
00:16:00.039 --> 00:16:03.240
for twenty years, right, And
we talked about data governance twenty years ago

228
00:16:03.279 --> 00:16:07.080
and fifteen years ago and ten years
ago, and basically nobody was doing it.

229
00:16:07.159 --> 00:16:11.279
I mean, you couldn't even it
wasn't even easy to do because you

230
00:16:11.279 --> 00:16:15.360
could either control access at the database
level, which is hard to access controls,

231
00:16:15.480 --> 00:16:18.600
or at the application level. But
there's nothing in the middle. And

232
00:16:18.679 --> 00:16:21.720
really it's in the middle. And
now with the cloud, that's one of

233
00:16:21.759 --> 00:16:25.519
the nice things about the cloud is
that it is this de facto marshaling area

234
00:16:25.799 --> 00:16:30.600
for functionality and data. And now
we have the capacity to apply very fine

235
00:16:30.639 --> 00:16:34.240
grain controls on things, on data
sets, on types of data. For

236
00:16:34.279 --> 00:16:38.559
example, we can scan and find
PII and then know, okay, flag

237
00:16:38.639 --> 00:16:41.639
this as sensitive. There are lots
of things we can do these days that

238
00:16:41.720 --> 00:16:45.120
we just kind of couldn't do ten
years ago. Real quick, one minute,

239
00:16:45.120 --> 00:16:48.440
what do you think is that about? Writer? We finally can do

240
00:16:48.519 --> 00:16:52.879
this stuff and so we are doing
it. What do you think I'd probably

241
00:16:52.480 --> 00:16:56.720
explained it slightly differently and say we
could have done it probably before as well,

242
00:16:56.759 --> 00:17:00.440
at least some of those aspects,
but people just didn't care enough because

243
00:17:00.440 --> 00:17:03.880
there was not enough arm in it. Now. But now that everybody who

244
00:17:03.880 --> 00:17:08.640
does anything with JENNYI is the danger
of sending data anywhere. People are really

245
00:17:08.680 --> 00:17:12.720
really baking up and seeing the pain
there. That's right. Well, it

246
00:17:12.759 --> 00:17:15.160
is like I say, it's a
call to arms, it's a call to

247
00:17:15.279 --> 00:17:18.799
action, and your organizations have got
to do it because you don't want to

248
00:17:18.799 --> 00:17:22.319
wind up in the crosshairs of an
audit. You don't want to wind up

249
00:17:22.319 --> 00:17:25.960
with a breach. You don't want
to wind up getting sued by someone because

250
00:17:26.000 --> 00:17:30.119
their information has now been leaked to
sensitive the sensitive resources out there. Well,

251
00:17:30.119 --> 00:17:34.079
folks, don't touch that doll.
We're talking all about AI and analytics

252
00:17:34.119 --> 00:17:37.160
platforms, and next up we're going
to dive into these large language models that

253
00:17:37.200 --> 00:17:41.920
are just taking the business world by
absolute storm. It's really quite fascinated to

254
00:17:41.960 --> 00:17:44.720
watch. But don't shut out.
That will be right back. You're listening

255
00:17:44.759 --> 00:17:56.519
to Inside Analysis. Welcome back to
Inside Analysis. Here's your host, Eric

256
00:17:56.640 --> 00:18:03.279
Tabanac. Folks back here on Inside
Analysis talking to Michael Berthol. He is

257
00:18:03.319 --> 00:18:07.160
the CEO and founder of a company
called NIME. That's k n I m

258
00:18:07.599 --> 00:18:11.640
Look these folks up online, an
open source analytics platform. It's wonderful stuff.

259
00:18:11.680 --> 00:18:15.799
It's like a giant candy store for
analysts to go play and have fun.

260
00:18:17.160 --> 00:18:19.359
But I want to talk to you
about these large language models, Michael,

261
00:18:19.359 --> 00:18:22.000
and in particular, first of all, the open source side of the

262
00:18:22.000 --> 00:18:26.640
equation. So Meta comes out with
Lama and lama too open source. Open

263
00:18:26.680 --> 00:18:30.720
AI used to be open now it's
not. Now it's the ironically named open

264
00:18:30.759 --> 00:18:36.640
AI because it's a black box.
And with the technology this powerful, I

265
00:18:36.759 --> 00:18:41.799
believe we need we need open source. I don't know that I would get

266
00:18:41.839 --> 00:18:45.920
behind a mandate that they must be
open source, but there needs to be

267
00:18:45.000 --> 00:18:49.759
some transparency into how these things are
working, just so that we can have

268
00:18:52.440 --> 00:18:56.599
our peaceful sleep at night to know
that there aren't bad actors involved somehow.

269
00:18:56.640 --> 00:19:00.559
I mean, certainly for regulated industries
like mantile services, if you bring it

270
00:19:00.599 --> 00:19:04.839
into some workflow for loan approval or
something like that, then you have to

271
00:19:04.839 --> 00:19:08.519
be able to explain how you came
up to your answer. But what are

272
00:19:08.559 --> 00:19:12.119
your thoughts in general about open source
versus closed source? With these large language

273
00:19:12.119 --> 00:19:19.039
models, I think there's a lot
of value in it. The problem is

274
00:19:19.119 --> 00:19:23.759
that, in my opinion, there's
open sourcing large language models. Isn't just

275
00:19:23.839 --> 00:19:27.559
about open sourcing the code, but
you also need to open source how it

276
00:19:27.640 --> 00:19:30.759
was actually trained. So in a
sense, you also need to at least

277
00:19:30.759 --> 00:19:36.359
give open access to the data that
was used for training. Because even if

278
00:19:36.400 --> 00:19:40.279
I give you a model and it
was trained on half copyrighted material that it's

279
00:19:40.279 --> 00:19:42.400
going to spit out again when you
use it, you wouldn't know if you

280
00:19:42.440 --> 00:19:45.839
didn't have access to the training data, right that That's part of that is

281
00:19:47.039 --> 00:19:52.240
is was it supposed to be used? And then I think the other piece

282
00:19:52.319 --> 00:19:56.519
is that what some companies are open
sourcing is only the code to use the

283
00:19:56.559 --> 00:20:02.240
model for predictions. Later to actually
apply it, I still don't know how

284
00:20:02.279 --> 00:20:06.240
it was trained. So that's the
third element that needs open sourcing. And

285
00:20:06.279 --> 00:20:11.680
then I believe one of the key
proprietary ingredients that a lot of these companies

286
00:20:11.759 --> 00:20:15.599
now have is safeguarding code around it
so that some types of answers don't get

287
00:20:15.640 --> 00:20:21.440
produced, some types of inputs aren't
being accepted, and open sourcing that as

288
00:20:21.480 --> 00:20:25.200
well would really really reveal their secret
sauce. And I think that's why they

289
00:20:25.240 --> 00:20:29.160
are the open eyes of the world
are shying away from that one. Right,

290
00:20:29.279 --> 00:20:32.119
No, it does make sense.
I mean, we have proprietary code.

291
00:20:32.160 --> 00:20:34.720
It's not new, but again,
these are very very powerful engines.

292
00:20:36.079 --> 00:20:38.839
And then there's another whole side of
this equation, which is the RAG model

293
00:20:40.200 --> 00:20:47.559
Retrieval augmented Generation, which upon great
reflection, I believe will be the layer

294
00:20:47.799 --> 00:20:53.599
of functionality for governance, for privacy, to a certain extent, for security,

295
00:20:55.079 --> 00:20:56.440
for management. You know, a
lot of that's going to get baked

296
00:20:56.480 --> 00:21:00.359
into the RAG model, where you
could, for example, before you hit

297
00:21:00.400 --> 00:21:03.119
your prompt, before your prompt goes
up to the large language model, have

298
00:21:03.119 --> 00:21:07.519
a layer in between the checks and
sees. Okay, and this is already

299
00:21:07.519 --> 00:21:11.920
happening. Like I asked Gemini a
couple of weeks ago how many electoral votes

300
00:21:11.960 --> 00:21:17.559
are in Georgia and Arizona and some
other state and thought for a second.

301
00:21:17.599 --> 00:21:21.519
It said elections are complex and fast
moving. We recommend you use Google.

302
00:21:22.599 --> 00:21:26.720
It is a guard rail. That's
a guardrail. They exactly built that in

303
00:21:26.839 --> 00:21:27.799
to say no, no, no, no, we don't want to touch

304
00:21:27.880 --> 00:21:30.559
that. Right, And that's in
the RAG model. Right. That's not

305
00:21:30.720 --> 00:21:34.759
like trained in the model. That's
outside the model, but it's the workflow

306
00:21:34.799 --> 00:21:41.839
you have around the engine that's very
very important. Right. I totally agree

307
00:21:41.839 --> 00:21:42.960
with that one. I mean,
that's what I called the safeguards before,

308
00:21:44.000 --> 00:21:47.680
and I think sometimes it's probably not
even part of the context that's part of

309
00:21:47.680 --> 00:21:51.440
the RAG models, but it's really
part of some safeguarding code even around it.

310
00:21:51.480 --> 00:21:53.680
I mean we use that at MIME
as well. So we have built

311
00:21:53.680 --> 00:21:57.599
in what we call KAY inside the
analytics platform that allows you to have a

312
00:21:57.680 --> 00:22:02.559
QA mode. You ask questions this
and this, and Excel out does that

313
00:22:02.599 --> 00:22:06.000
look in a nine doctim and then
it gives you shows you a couple of

314
00:22:06.039 --> 00:22:11.119
notes in nine, and we're of
course filtering that these notes do actually exist,

315
00:22:11.279 --> 00:22:14.119
because every now and then open AI, which we use underneath the hood,

316
00:22:14.720 --> 00:22:18.759
hallucinates and invents notes that Nie probably
should have, but we don't have

317
00:22:18.799 --> 00:22:21.400
it, and that doesn't help the
pro use. All right, that's a

318
00:22:21.480 --> 00:22:26.880
very very simple way of code around
the KAY that is just making sure that

319
00:22:26.960 --> 00:22:30.480
what it spits out is reasonably useful. Mm hmm. Yeah, that's interesting,

320
00:22:30.680 --> 00:22:33.920
and you're going to see more and
more of these AI agents. That's

321
00:22:33.920 --> 00:22:38.000
what everyone is talking about now,
are AI agents, which are like little

322
00:22:38.039 --> 00:22:42.079
bots, semi autonomous bots that can
do various things, and they can check

323
00:22:42.119 --> 00:22:45.079
on each other and they can do
all kinds of stuff. I mean,

324
00:22:45.599 --> 00:22:48.119
it's very interesting to me when we
talk about data science. We talked about

325
00:22:48.119 --> 00:22:52.119
it before, it all seems to
be getting subsumed now into AI. In

326
00:22:52.200 --> 00:22:56.799
conversations about AI, even though there
are lots of different versions of AI,

327
00:22:56.960 --> 00:23:00.559
right, I mean, there are
traditional models, regression models, all sorts

328
00:23:00.599 --> 00:23:03.039
of old fashioned aif you well,
it's still very powerful and still works,

329
00:23:03.799 --> 00:23:07.000
but the new stuff is sucking all
the action out of the room, isn't

330
00:23:07.000 --> 00:23:11.720
that about? Right? Yeah?
We see that as well. And sometimes

331
00:23:11.759 --> 00:23:12.759
I mean, I'm an old guy
by now, I've seen this in the

332
00:23:12.799 --> 00:23:17.440
past. Right when back propagation came
along, everybody was suddenly using grade into

333
00:23:17.440 --> 00:23:19.119
cent for every problem. We just
thought, hey, you can solve this

334
00:23:19.240 --> 00:23:22.119
directly. You don't need to do
grade into set. Then there was support

335
00:23:22.160 --> 00:23:26.240
vector machines, and there was somebody
else, and now then with deep learning,

336
00:23:26.279 --> 00:23:30.240
and now it's AI. So sometimes
we see building workflows for even very

337
00:23:30.240 --> 00:23:33.119
simple things, they're reaching out to
some AI and we just say, hey,

338
00:23:33.119 --> 00:23:37.160
there's a no denignment that does that
computationally a lot less expensive. I

339
00:23:37.240 --> 00:23:41.720
don't use that, So I think, to me, it's it's a bit

340
00:23:41.720 --> 00:23:42.920
of a hype right now, it's
just a new kid in the block.

341
00:23:42.960 --> 00:23:48.480
Everybody wants to play with it and
use it. But the augmented really mixing

342
00:23:48.480 --> 00:23:51.519
it and matching it right with traditional
techniques. I think that's where the true

343
00:23:51.559 --> 00:23:55.960
value lies. Yeah. Well,
and so I'm just guessing here that one

344
00:23:55.960 --> 00:23:59.480
of the nice things about your platform
is that it is an end to end

345
00:23:59.519 --> 00:24:03.720
platform for building models, designing models, training models, pulling the data in

346
00:24:03.759 --> 00:24:08.839
all these things, and it's adjacent
to this business hub. So you have

347
00:24:08.920 --> 00:24:15.079
a marshaling area for ideas and for
testing algorithms and for testing models. Then

348
00:24:15.119 --> 00:24:18.599
you connect it through the business hub
and see what happens and see how it

349
00:24:18.640 --> 00:24:23.160
operates. And it's important to have
this one environment where that takes place,

350
00:24:23.200 --> 00:24:29.119
because when you have multiple tools,
it just takes longer and it's disjointed and

351
00:24:29.160 --> 00:24:33.160
there are connections between the tools and
things change. So it's important to have

352
00:24:33.240 --> 00:24:38.960
that main marshaling area to It's like
a giant analytics sandbox. Is that about

353
00:24:40.000 --> 00:24:42.720
right? That's a very nice description. Absolutely. I tend to say that

354
00:24:44.519 --> 00:24:49.759
data scientists doesn't necessarily need to know
how the method does something, but it

355
00:24:49.839 --> 00:24:52.720
needs to know what the method does. So if it's reaching out to a

356
00:24:52.759 --> 00:24:56.920
Python library or in our library or
Sea library underneath it, it's not that

357
00:24:56.960 --> 00:25:02.200
important, but you still need to
understand what the method actually does underneath to

358
00:25:02.279 --> 00:25:04.640
be able to interpret the results.
It's a simple example. If you don't

359
00:25:04.640 --> 00:25:07.519
know what a regression coefficient is,
you won't be able to interpret it,

360
00:25:07.680 --> 00:25:11.000
but you don't necessarily need to understand
how it was derived from the data.

361
00:25:11.359 --> 00:25:15.240
Yeah, no, that that's pretty
interesting. Let me throw this concept at

362
00:25:15.279 --> 00:25:17.920
you and see what you think about
it. I wrote up an article just

363
00:25:18.200 --> 00:25:21.839
last week. I guess about this. I was flying to a conference in

364
00:25:21.880 --> 00:25:25.599
Denver just thinking about these large language
models and analytics and AI and all this

365
00:25:25.680 --> 00:25:27.839
stuff that have been covering for a
long long time, and I thought to

366
00:25:27.880 --> 00:25:33.359
myself about this concept I call the
executive cockpit, and the idea is that

367
00:25:33.480 --> 00:25:38.440
I think very forward looking organizations are
going to deploy a small language model that

368
00:25:38.519 --> 00:25:41.960
is aligned with their business, like
if it's manufacturing or healthcare or whatever,

369
00:25:44.279 --> 00:25:48.359
in their data center, so on
prem, possibly in the cloud as well,

370
00:25:48.400 --> 00:25:52.440
but I have my thoughts wrapped around
this on prem small language model.

371
00:25:52.640 --> 00:25:56.559
Then you're going to train it on
your ERP, on your salesforce, on

372
00:25:56.599 --> 00:26:00.559
your CRM, on your customer support
for example, your ticket it's like any

373
00:26:00.599 --> 00:26:03.759
of your core enterprise systems. You're
going to train this model on your data,

374
00:26:03.799 --> 00:26:08.640
on your business business data. Then
what you'll do is set up Coughka

375
00:26:08.720 --> 00:26:15.559
topics coming from those systems into a
vector database adjacent to this interface for the

376
00:26:15.640 --> 00:26:19.359
small language model, and that is
where the executives will spend their day running

377
00:26:19.359 --> 00:26:22.359
their business. Because then you could
ask any question at all, how is

378
00:26:22.359 --> 00:26:26.920
our marketing working in APAC? Who
can we let go if we have to

379
00:26:27.000 --> 00:26:32.839
save some money. Where are we
weak in our organization right now? Just

380
00:26:32.880 --> 00:26:36.599
all kinds of different questions and you'll
get all these answers. And I actually

381
00:26:36.640 --> 00:26:38.839
mentioned to a CEO of this one
company because I was trying to get him

382
00:26:40.000 --> 00:26:42.359
to help me do sales enablement for
them, because I have this big audience

383
00:26:42.400 --> 00:26:47.200
I've been marketing to for years.
And one person turns out to be the

384
00:26:47.240 --> 00:26:52.519
next deputy chief data officer for the
IRS. And I sent this email saying,

385
00:26:52.559 --> 00:26:55.240
Hey, this is the lady I've
known for a long long time.

386
00:26:55.319 --> 00:26:57.480
This is what I mean by sales
enablement. Do you guys have the IRS

387
00:26:57.519 --> 00:27:00.759
account? And I fire back you
so I don't know. I don't know

388
00:27:00.759 --> 00:27:03.200
if the other's accounts. I thought
to myself, well, you would know

389
00:27:03.240 --> 00:27:06.920
if you had the executive cockpit,
because you would just ask it, do

390
00:27:07.000 --> 00:27:10.319
we have the IRS account? Who
is the account rep? What's the latest

391
00:27:10.319 --> 00:27:14.119
of this account? Because you're getting
information from all these systems in your private

392
00:27:14.200 --> 00:27:17.480
environment. But what do you think
about this concept? Is that is that

393
00:27:17.599 --> 00:27:19.200
doable? Is that pie in the
sky or what do you think about all

394
00:27:19.200 --> 00:27:25.039
that? It's an interesting idea.
I thought about it similar I mean,

395
00:27:25.119 --> 00:27:29.440
at the end of the day,
you're personalizing a large language model around your

396
00:27:29.480 --> 00:27:33.839
own infrastructure in house data. I
think the challenge there is that in order

397
00:27:33.880 --> 00:27:37.240
to get a really really good model
like that one that's really useful, you

398
00:27:37.319 --> 00:27:41.920
need to train it on a lot
more data than just your own. So

399
00:27:41.920 --> 00:27:45.519
in a sense, you need to
benefit from your competitor's data without actually seeing

400
00:27:45.519 --> 00:27:48.599
that, but kind of learning the
general structure and the general insights, and

401
00:27:48.640 --> 00:27:52.160
then you customize it on your own, which in return kind of means that

402
00:27:52.240 --> 00:27:59.599
you should also be providing your data
to other organizations. It's almost like that's

403
00:27:59.680 --> 00:28:03.680
kind of pree competitive training of these
models so that they're useful for everybody.

404
00:28:03.799 --> 00:28:10.599
I think just training it on your
setup, you need some bigger context than

405
00:28:10.599 --> 00:28:14.079
that. Or maybe you're a company
and you have enough context anyway, But

406
00:28:14.160 --> 00:28:15.920
for every small company, I don't
think you have enough data to really get

407
00:28:15.960 --> 00:28:22.200
meaningful insights. That's very interesting.
That's a good that's a good point because

408
00:28:22.279 --> 00:28:25.119
I'm just I'm wondering to myself and
I'm gonna throw this one at you too.

409
00:28:25.240 --> 00:28:27.480
So one of my AHA moments with
these large language models is when I

410
00:28:27.559 --> 00:28:33.200
realized that when you train them on
a corpus of data. They're not actually

411
00:28:33.240 --> 00:28:37.319
persisting the data verbatim. It's not
like they're taking strings of text and storing

412
00:28:37.359 --> 00:28:42.319
it in a record somewhere, but
rather, in the training process that data

413
00:28:42.440 --> 00:28:48.200
you use will adjust the weights and
biases and the parameters of the model.

414
00:28:48.759 --> 00:28:52.839
So in other words, it's like, huh, well, that's that's very

415
00:28:52.839 --> 00:28:56.839
interesting that it can train in that
fashion and then reflect back to you such

416
00:28:56.920 --> 00:29:02.319
remarkably granular detail about things. And
you know, what I've seen is that

417
00:29:02.359 --> 00:29:06.480
if there is a subject therea that
has been published about widely, like how

418
00:29:06.559 --> 00:29:11.880
computer processors work, or how an
irrigation system works, anything that has a

419
00:29:11.920 --> 00:29:15.960
lot of content on the web that
these engines were trained on, it does

420
00:29:17.119 --> 00:29:18.759
very well. It knows all that
stuff. It's when you get to the

421
00:29:18.799 --> 00:29:22.799
fringe where it's not that much published. And I guess that's kind of your

422
00:29:22.799 --> 00:29:26.680
point about having enough data to train
the model. So if you don't have

423
00:29:26.880 --> 00:29:30.160
enough, you're not going to get
the contours right, and it's going to

424
00:29:30.160 --> 00:29:33.640
be skewed in one direction or other. Is that about right? I think

425
00:29:33.640 --> 00:29:36.079
that's a very good summary. The
contrast is right. I mean, a

426
00:29:36.079 --> 00:29:38.920
colleague of mind wants summarized, since
it's essentially it's a consensus engine. It's

427
00:29:38.960 --> 00:29:45.119
getting the consensus around what a computer
programming is, learns that from the data

428
00:29:45.160 --> 00:29:48.039
and can repeat that. But if
it's just one isolated outcome, it's not

429
00:29:48.039 --> 00:29:52.920
going to be able to recall that
one. Interesting. Yes, So Craig

430
00:29:52.960 --> 00:29:56.079
schmid Huber I think his name is. He's the guy who wrote the papers

431
00:29:56.079 --> 00:30:00.000
on the transformers, and he's based
I guess he's actually in Saudi Arabia these

432
00:30:00.039 --> 00:30:03.960
days, but I want to say
he's German of German origin. And I

433
00:30:04.000 --> 00:30:07.440
was amazed when I realized he wrote
those papers in like the nineteen nineties or

434
00:30:07.480 --> 00:30:11.119
something. And it's just just now
we have the compute to be able to

435
00:30:11.480 --> 00:30:15.279
can you explain that? Is that
what happened is that just the timing was

436
00:30:15.400 --> 00:30:18.759
right now to be able to understand
this and put it into play, because

437
00:30:18.799 --> 00:30:22.720
that was one of the big changes. And now it's able to see like,

438
00:30:22.839 --> 00:30:25.880
you know, ten twelve tokens left
or right as opposed to just like

439
00:30:26.160 --> 00:30:29.720
two or three. And you also
have this, like you say, like

440
00:30:29.759 --> 00:30:33.519
a consensus right where so they are
like I call it almost like an ai

441
00:30:33.640 --> 00:30:36.720
Greek chorus where one is saying I
think it should be an A. I

442
00:30:36.720 --> 00:30:37.240
think it should be a B.
I think it should be a C.

443
00:30:37.440 --> 00:30:41.319
And then the Okay, I'm going
to pick this one. That's very interesting.

444
00:30:41.359 --> 00:30:45.119
It's a very interesting development. But
why do you think it took so

445
00:30:45.240 --> 00:30:48.640
long? Is it just because we
now have the compute to do that?

446
00:30:48.680 --> 00:30:52.119
I think it was a compute party
issue as well. And then some science

447
00:30:52.200 --> 00:30:55.759
tends to have a little bit more
of fun, needs a little bit of

448
00:30:55.759 --> 00:30:59.119
time before it truly has an impact, but mostly waiting for complete power.

449
00:30:59.480 --> 00:31:03.359
I don't know. One way of
looking at what this consensus really does is

450
00:31:03.400 --> 00:31:07.440
I don't know if you watch these
YouTube videas about JGBT playing chess now and

451
00:31:07.480 --> 00:31:11.400
the interesting part is that at the
beginning it does extremely well and does very

452
00:31:11.480 --> 00:31:17.039
sensible things, and part of that
is these opening libraries are all over the

453
00:31:17.039 --> 00:31:22.440
place, so that's extremely well established
consensus. And then somewhere in the middle

454
00:31:22.519 --> 00:31:26.680
it starts inventing bizarre moves and suddenly
new figures pop up on the on the

455
00:31:26.680 --> 00:31:30.559
board out of nowhere, right,
and it has always meaningful explanations for that.

456
00:31:30.000 --> 00:31:36.079
And the problem there is that data
is so sparse that there's no consensus

457
00:31:36.119 --> 00:31:38.680
to learn. So at the beginning
it sounds it almost looks like it understands

458
00:31:38.839 --> 00:31:44.000
chess rules. But the only reason
it does follow the chess rules is that

459
00:31:44.039 --> 00:31:48.279
they're so deeply ingrained in all of
the common material that you see that the

460
00:31:48.680 --> 00:31:52.279
kind of the likelihood of going outside
of the world book is too small.

461
00:31:52.640 --> 00:31:55.400
But somewhere in the middle of the
game it goes completely off the books.

462
00:31:55.799 --> 00:31:59.920
That's interesting, that's wild. So
one of my good friends in the business,

463
00:32:00.359 --> 00:32:02.839
as a gentleman named Usama Fayad,
you may have come across from at

464
00:32:02.839 --> 00:32:07.240
some point. He was the first
chief data officer for Yahoo way back in

465
00:32:07.240 --> 00:32:12.200
the day, and now he runs
the Institute for Experiential AI over at Northeastern

466
00:32:12.359 --> 00:32:15.440
University here in the States. And
I had him in the show, and

467
00:32:15.480 --> 00:32:19.119
he's very funny, he's very candid. He said, these large models,

468
00:32:19.480 --> 00:32:22.720
they're too big, they're not supposed
to work. We don't know why they

469
00:32:22.759 --> 00:32:25.359
work. What are you talking about, this guy who runs this whole operation.

470
00:32:25.400 --> 00:32:28.480
He's joking, we don't even know
how they work. I mean,

471
00:32:28.920 --> 00:32:31.839
how's that for transparency, right,
I mean there's some truth, right,

472
00:32:31.880 --> 00:32:35.960
we don't really know how they come
up with these answers. Right, it's

473
00:32:36.000 --> 00:32:38.039
a wild mix of it. It's
a highly distributed model. We don't know

474
00:32:38.279 --> 00:32:44.640
why a particular answer comes. We
can come up with kind of proxies for

475
00:32:44.680 --> 00:32:46.839
an explanation by wiggling with the inputs
and trying to figure out what happens,

476
00:32:46.880 --> 00:32:50.240
and we can say, ah,
this probably had a lot of influence on

477
00:32:50.279 --> 00:32:53.000
the decision, but we don't know
for sure. That's so wild. I

478
00:32:53.000 --> 00:32:55.640
mean, that's just such a big
deal that you know, but we do.

479
00:32:55.839 --> 00:33:00.359
So now we have all this observability
in the data space, right,

480
00:33:00.400 --> 00:33:04.359
You've got Data Relic and Data Dog
or new relative to say data. All

481
00:33:04.359 --> 00:33:07.640
these different companies are doing observability which
I think spun out of Kubernetes primarily.

482
00:33:08.039 --> 00:33:12.960
But it's very interesting and we need
that kind of observability in these large language

483
00:33:12.960 --> 00:33:14.920
models. I think, I think
that's going to be one of the keys

484
00:33:14.960 --> 00:33:17.160
to success. But folks, don't
touch that dot. Will be right back.

485
00:33:17.319 --> 00:33:29.960
We're talking to Michael berthold from NIME
on Inside Analysis. Standby, Welcome

486
00:33:30.000 --> 00:33:37.400
back to Inside Analysis. Here's your
host, Eric Tabanac. All right,

487
00:33:37.400 --> 00:33:42.240
folks back here on Inside Analysis with
Michael bertholdt founder and CEO of NIME k

488
00:33:42.480 --> 00:33:45.240
n I m E looked them up
online and Michael I was mentioning to you

489
00:33:45.240 --> 00:33:51.880
in the break that I'm wondering to
myself this whole business intelligence industry and there

490
00:33:51.920 --> 00:33:55.880
are hundreds of players these days,
hundreds of companies doing some form of analytics.

491
00:33:55.880 --> 00:34:00.680
Of course, NIME is an whole
analytics platform, in open source analytics

492
00:34:00.720 --> 00:34:04.000
platform, end to end, but
there are lots of point tools, whether

493
00:34:04.039 --> 00:34:07.920
it's visualization or number crunching, olapp
roll, app all this kind of stuff,

494
00:34:07.360 --> 00:34:12.360
and I wonder is all of that
in the crosshairs of these foundational models?

495
00:34:12.400 --> 00:34:15.599
What do you think? That's a
very very interesting question, and we

496
00:34:15.679 --> 00:34:20.280
of course asked ourselves that as well. And I think for some of the

497
00:34:20.559 --> 00:34:23.199
some of the tools that you mentioned, like generating visualizations, that type of

498
00:34:23.199 --> 00:34:29.519
stuff, I do think they are
pretty replaceable by AI type models, because

499
00:34:29.559 --> 00:34:34.360
at the end of the day,
you're doing something, you're generating a code

500
00:34:34.360 --> 00:34:37.119
that generates the visualization based on data, and you judge the output of that

501
00:34:37.239 --> 00:34:40.400
code by just looking at it and
saying this is quite right. So I

502
00:34:40.440 --> 00:34:44.679
think that type of stuff will go
away. And we have in NIME actually

503
00:34:44.840 --> 00:34:47.960
built in what used to be an
each chart scripting editor that has now an

504
00:34:49.000 --> 00:34:52.159
AI element and you don't need to
touch the code anymore. So those types

505
00:34:52.159 --> 00:34:55.920
of wills I believe will go away. The eye tools trying to really find

506
00:34:57.400 --> 00:35:01.239
surprising, interesting new insights in data, I think that type of stuff is

507
00:35:01.239 --> 00:35:07.159
a lot harder to replace because fundamentally
you're trying to find something new, And

508
00:35:07.320 --> 00:35:12.199
like we discussed before the break,
these GENEI models are consensus engines, right,

509
00:35:12.239 --> 00:35:16.159
so they kind of try to gravitate
towards something they've seen more and more

510
00:35:16.199 --> 00:35:21.599
often before. Interesting. That's right, That's an excellent point. Really,

511
00:35:21.639 --> 00:35:28.440
that's that's exactly right. So it's
good for understanding the well trodden path basically,

512
00:35:28.519 --> 00:35:30.039
Like that's what it's very good at
doing is saying, Okay, there's

513
00:35:30.039 --> 00:35:35.039
a highway, it goes that direction, but I want to go wandering around

514
00:35:35.079 --> 00:35:37.800
the forest, and it's not as
good on the fringe basically, so you

515
00:35:38.199 --> 00:35:40.679
will use it. But I mean, so I read an article of some

516
00:35:40.760 --> 00:35:45.599
guy on LinkedIn talking about how he
connected I don't know by ODBC or JDBC

517
00:35:45.800 --> 00:35:50.760
or something in his model with data
sources, and he asked it to queer

518
00:35:50.840 --> 00:35:53.719
the data source and it did it
reached into the database, pulled the information

519
00:35:53.760 --> 00:35:58.880
out and delivered it and you're like, Okay, that's pretty interesting. And

520
00:35:58.920 --> 00:36:01.599
then when I think to my about
what's what could be happening here? Is

521
00:36:02.559 --> 00:36:07.559
in the data warehousing space, for
example, we move so much data around.

522
00:36:07.960 --> 00:36:10.280
It's all the data that's from your
core systems that you've decided to put

523
00:36:10.280 --> 00:36:14.480
in, which is a tremendous amount. Very little of that data ever gets

524
00:36:14.639 --> 00:36:17.920
used a lot of times it's the
it's the summaries or the aggregates or the

525
00:36:19.000 --> 00:36:22.079
roll ups that are used for various
purposes, but a lot of it just

526
00:36:22.119 --> 00:36:25.719
doesn't even get used at all.
And I think that what these large language

527
00:36:25.719 --> 00:36:30.360
models are going to do is kind
of turn the entire model inside out of

528
00:36:30.360 --> 00:36:35.519
how we viewed moving data and analyzing
data and doing things with data. Because

529
00:36:35.559 --> 00:36:38.800
they don't they don't really care.
They're just going to once they're trained on

530
00:36:38.920 --> 00:36:42.679
a certain space. And again,
if you train it on your data,

531
00:36:42.760 --> 00:36:45.280
or if you're in your vector database, you have a lot of embeddings of

532
00:36:45.320 --> 00:36:49.199
your corporate data and you point your
RAG model there, well, you can

533
00:36:49.239 --> 00:36:53.760
get answers to things very quickly that
before would have required running reports and doing

534
00:36:53.800 --> 00:36:57.880
etl. And doing all this stuff, and I think that in many use

535
00:36:57.920 --> 00:37:01.000
cases, these models are going to
short circuit all that stuff and you're just

536
00:37:01.000 --> 00:37:04.760
not going to have to do as
much that stuff anymore at all. But

537
00:37:04.800 --> 00:37:08.400
what do you think about that?
I think there is some truth to it,

538
00:37:08.440 --> 00:37:14.440
because fundamentally, what these models won't
necessarily do is actually look at all

539
00:37:14.480 --> 00:37:17.639
of the data, but they're going
to apply a lot of common standard practices

540
00:37:17.679 --> 00:37:21.880
to that. And standard sounds a
little bit too limiting. I think there's

541
00:37:21.920 --> 00:37:24.880
a huge wealth of standard practices that
people do apply to the data, and

542
00:37:24.920 --> 00:37:30.719
that's part of this consensus engine,
and so that the AI models will try

543
00:37:30.760 --> 00:37:34.960
out a lot of those things a
lot faster than you ever would. So

544
00:37:35.079 --> 00:37:37.719
absolutely, and there's a good chance
that some of these insights that will be

545
00:37:37.800 --> 00:37:44.679
generated are interesting to you. But
then continuing the exploration and saying, I

546
00:37:44.679 --> 00:37:50.320
mean, how do they always say
the Eureka moment is usually preceded by Oops,

547
00:37:50.360 --> 00:37:54.920
that's strange. I think you'll have
these moments right, and AI doesn't

548
00:37:54.920 --> 00:37:58.639
do that. AI doesn't say this
is weird. I should diggle in a

549
00:37:58.679 --> 00:38:02.000
little bit deeper because that's outside of
the consensus. So it will continue doing

550
00:38:02.119 --> 00:38:07.079
kind of like you said, will
continue the normal path. And that's what

551
00:38:07.320 --> 00:38:15.679
I believe. The human intuition,
curiosity oops detection capability is going to be

552
00:38:15.440 --> 00:38:20.679
relevant for a long time. I
like this oops detection. That's good stuff.

553
00:38:20.880 --> 00:38:22.199
Well, there was a gentleman I
had on the show years ago who

554
00:38:22.239 --> 00:38:25.559
did something. He said something a
lot like that. He basically said,

555
00:38:25.599 --> 00:38:30.199
AI doesn't have to be the ability
to be like, hmm, that's kind

556
00:38:30.239 --> 00:38:34.480
of weird. What's going on with
that? Right? Because it's just processing

557
00:38:34.519 --> 00:38:37.679
information and doing what it's been told
to do, which is just reflect backwards

558
00:38:38.000 --> 00:38:40.760
based upon a prompt, and it's
training. It's a very simple thing.

559
00:38:42.280 --> 00:38:45.480
I mean, it's very complex in
terms of how it got there, but

560
00:38:45.559 --> 00:38:47.559
nonetheless it you know, one thing
that did annoy me, I will say

561
00:38:47.639 --> 00:38:52.119
is in the early days when that
New York Times reporter was getting deep with

562
00:38:52.199 --> 00:38:57.920
the with chat GPT and trying to
like tease out of it whether it's sentient

563
00:38:58.039 --> 00:39:00.719
or something. I'm like, dude, that is a miss use of the

564
00:39:00.760 --> 00:39:05.039
technology, Like that is not whether
you should be using this thing for to

565
00:39:05.079 --> 00:39:07.519
try to like what trick it into
revealing that it's really alive. And you

566
00:39:07.559 --> 00:39:10.599
know, what are you been talking
about. And I think that's part of

567
00:39:10.639 --> 00:39:15.280
the downside these days is that.
And I'm a media person myself, but

568
00:39:15.360 --> 00:39:20.639
a lot of times the media will
just sort of glom onto some narrative about

569
00:39:20.679 --> 00:39:23.880
something and it's very hard for them
to decouple from that and get down to

570
00:39:23.960 --> 00:39:25.960
brass tacks. And that's what we
do in the show. In fact,

571
00:39:27.000 --> 00:39:29.639
I used to say at the beginning
of every show, the show, it's

572
00:39:29.679 --> 00:39:32.480
all about getting down to the brass
tax of what actually happens in the data

573
00:39:32.519 --> 00:39:36.639
world and what you do with this
stuff. And I think it is important

574
00:39:36.639 --> 00:39:39.760
that people keep in their minds the
purpose of this technology, why are you

575
00:39:39.880 --> 00:39:43.679
using it, Where is it appropriate
to use it, and where is it

576
00:39:43.719 --> 00:39:45.639
not appropriate to use it? And
that's just basic common sense, right,

577
00:39:46.519 --> 00:39:50.360
Yes, I totally agree. I
mean I go pretty much in line with

578
00:39:50.400 --> 00:39:53.159
also the European aii that they just
passpect. I mean, if it's not

579
00:39:53.440 --> 00:39:58.039
mission critical, if it's not safety
critical, you can trust a system that

580
00:39:58.199 --> 00:40:01.159
is wrong in I don't know,
point one percent of all cases. If

581
00:40:01.159 --> 00:40:06.480
it's controlling nuclear power plants, better
not be wrong in point one percent of

582
00:40:06.519 --> 00:40:08.880
all cases, right, That's right. You got to watch out where it's

583
00:40:09.119 --> 00:40:13.039
so where do you see a lot
of use case of your clients. I

584
00:40:13.039 --> 00:40:15.760
mean, obviously some of your clients
are using large language models. Where are

585
00:40:15.760 --> 00:40:22.360
you seeing success stories in that space
right now? So there's a lot of

586
00:40:22.400 --> 00:40:25.039
success stories in other areas of the
business, as you probably probably know.

587
00:40:25.159 --> 00:40:30.599
Undoubtedly no checking legal contracts, doing
marketing material, that type of stuff.

588
00:40:30.639 --> 00:40:35.760
There's a lot of value in applying
GENII on the data analytic space. Honestly,

589
00:40:35.880 --> 00:40:37.679
we don't. We see a lot
of interest. There's a lot of

590
00:40:37.719 --> 00:40:43.039
people that say, oh cool,
I can build a customized chatboard using name

591
00:40:43.159 --> 00:40:46.559
that's not really our core business.
And then the real applications tend to be

592
00:40:46.599 --> 00:40:52.480
around text processing, just where Jennai
is really strong. And then instead of

593
00:40:52.840 --> 00:41:00.239
using outdated antique libraries for sentiment analysis
or text segmentation, you're just handing over

594
00:41:00.320 --> 00:41:02.239
to an AI model and say,
hey, segment this, or extract the

595
00:41:02.320 --> 00:41:06.519
key components or create a summary for
that type of stuff. It's amazing.

596
00:41:06.920 --> 00:41:12.440
So I see I'm also as image
mining extensions. I think that's the next

597
00:41:13.119 --> 00:41:17.239
setup where we can use image processing
capabilities of jen Ai for a lot of

598
00:41:17.280 --> 00:41:21.519
the number crunching. I mean,
we've all seen these cases where you can't

599
00:41:21.559 --> 00:41:23.960
add two numbers, doesn't really know
what the prime numbers. This is the

600
00:41:24.079 --> 00:41:29.920
understanding of the concept of a number, right. So there. I think

601
00:41:29.920 --> 00:41:36.360
it's more as as a tool to
help you build workflows, build datazations,

602
00:41:36.480 --> 00:41:38.840
but only as a helper, right. So that's actually an excellent point I

603
00:41:38.880 --> 00:41:44.800
wanted to get into. I believe
that we're just scratching the surface of using

604
00:41:44.840 --> 00:41:49.320
these models as a component in a
workflow. So you mentioned, for example

605
00:41:49.360 --> 00:41:53.079
summarization. That is hugely powerful.
I mean, you know you can enter

606
00:41:53.800 --> 00:41:59.719
especially for policies, for complex policies
for law, for example, for legal

607
00:42:00.719 --> 00:42:04.960
protocols and when to file motions,
what motions you can file, what you

608
00:42:05.000 --> 00:42:07.960
have to do according to I mean, you used to have to pay lawyer's

609
00:42:07.960 --> 00:42:09.719
a lot of money to tell you
that stuff. Now if you just get

610
00:42:09.719 --> 00:42:13.559
access to the rules, will load
them into a large language model and just

611
00:42:13.559 --> 00:42:17.880
start asking questions. That is an
incredibly powerful use case because it used to

612
00:42:17.920 --> 00:42:22.039
take a lot of time to sort
through the process of how to do something.

613
00:42:22.079 --> 00:42:24.440
Now you just ask it. How
do I fire someone? First step,

614
00:42:24.519 --> 00:42:28.440
send them a letter saying they're not
performing properly. Second step, you

615
00:42:28.480 --> 00:42:30.880
know, monitor their behavior. Third
step with all this stuff, it's like

616
00:42:30.920 --> 00:42:34.840
there, it is like, wow, talk about saving time. I mean,

617
00:42:34.840 --> 00:42:37.840
it saves time. And here's another
big soapbox issue. It improves morale

618
00:42:38.519 --> 00:42:44.280
because nobody wants to spend their time
scratching their head reading through just dreadful documentation.

619
00:42:44.400 --> 00:42:46.719
Nobody likes doing that, nobody,
So all that stuff is going to

620
00:42:46.760 --> 00:42:52.159
go away, right, what do
you think? I totally agree your examples

621
00:42:52.199 --> 00:42:57.760
center a lot around firing people,
by the way, But I think I

622
00:42:57.800 --> 00:43:00.239
tend to say, and people ask
me if Jenny going to make data science

623
00:43:00.280 --> 00:43:04.039
lives easier? I say no,
I don't think so. But it's going

624
00:43:04.079 --> 00:43:07.920
to make it nicer because it's going
to remove all of that boring stuff and

625
00:43:07.000 --> 00:43:10.840
we can now focus on the really
interesting but more complex stuff. So it's

626
00:43:10.880 --> 00:43:15.400
going to make it more interesting,
more complex. That's interesting. Yeah,

627
00:43:15.440 --> 00:43:19.880
wow? And I think you do
want to document things, and you can

628
00:43:19.880 --> 00:43:22.199
have it document things for you too, right, you can just throw a

629
00:43:22.199 --> 00:43:24.920
whole bunch of stuff and document this, okay exactly. I mean we now

630
00:43:24.960 --> 00:43:29.360
have a component on the hubs that
takes a nine workflow and explains what the

631
00:43:29.480 --> 00:43:31.440
nine workflow does. And we do
that by just shipping it off to Jennai.

632
00:43:31.480 --> 00:43:35.599
It's perfect for that. Wow,
that's amazing. All right, folks,

633
00:43:35.679 --> 00:43:37.719
Well, Podcast pull a segment coming
up next. We're listening to Inside

634
00:43:37.719 --> 00:43:45.320
Analysis. All right, folks back
here on Inside Analysis talking to Michael Bertholdt

635
00:43:45.480 --> 00:43:51.480
is the founder and CEO of nine
K. I am and Michael. I

636
00:43:51.559 --> 00:43:53.199
know what it's like in a software
company. There's always a roadmap. You're

637
00:43:53.239 --> 00:43:57.519
always working on something, and we
talked about a couple of key things.

638
00:43:57.760 --> 00:44:01.760
Governance. There's model governance, there's
data governance, there's it governance. What

639
00:44:01.760 --> 00:44:07.400
are you working on in the governance
space? Thanks for asking that, Like,

640
00:44:07.400 --> 00:44:09.880
like you know, all the road
maps are changing all the time.

641
00:44:10.440 --> 00:44:15.159
But what we're currently working on.
We had actually this model government's topic on

642
00:44:15.239 --> 00:44:17.000
the work right, brend I've been
working on that for a couple of years

643
00:44:17.039 --> 00:44:22.639
now. So the idea of being
able to monitor what models are doing automatically

644
00:44:22.920 --> 00:44:24.960
we train them. We talked about
that and I think the first episode,

645
00:44:25.960 --> 00:44:31.119
but what we added now is the
ability to also govern the AI usage of

646
00:44:31.159 --> 00:44:35.199
people that are creating nine work clows. So, first of all, when

647
00:44:35.239 --> 00:44:38.559
somebody is creating nine work clos using
the NIME analytics platform and uses this built

648
00:44:38.559 --> 00:44:43.639
in AI. We call it KAI
for NIME AI. We need to make

649
00:44:43.639 --> 00:44:46.679
sure that gets channel to an IT
approved AI. Right. Maybe that's just

650
00:44:46.760 --> 00:44:50.519
for expense purposes, you know,
I'm going to have too much consumption in

651
00:44:50.519 --> 00:44:53.320
the cloud money you make that in
house or it's really a data privacy issue.

652
00:44:53.559 --> 00:44:58.639
But the more worrisome part for people
is that you I mean one of

653
00:44:58.679 --> 00:45:02.000
the springs of the Namealytics platform,
the workflow concepts, is that everybody can

654
00:45:02.119 --> 00:45:06.880
use any technology they want. Right, they can reach out to experimental libraries,

655
00:45:06.880 --> 00:45:08.599
they can reach out to our stuff, to Python stuff, to whatever.

656
00:45:09.280 --> 00:45:15.239
But by now they can of course
also connect to various different AI providers,

657
00:45:15.920 --> 00:45:20.199
and we need a way for them
for Central Light Tea Governance to be

658
00:45:20.239 --> 00:45:25.559
able to make sure that the nine
workflow users inside the organization can only use

659
00:45:25.639 --> 00:45:30.599
approved AIS. So maybe maybe marketing
can use an AI in the cloud,

660
00:45:30.679 --> 00:45:35.599
but maybe legal shouldn't or HR shouldn't, right, And that's something we have

661
00:45:35.639 --> 00:45:38.719
built in into the name hub.
Now that we can limit the types of

662
00:45:38.880 --> 00:45:44.679
AIS, you can their users can
reach out to from the nine ndlydrics platform

663
00:45:45.159 --> 00:45:49.719
and they get to choose from one
of the approved AIS that it central light

664
00:45:49.920 --> 00:45:52.960
set up and said, okay,
here's an AI that's consumption light. That's

665
00:45:52.960 --> 00:45:55.679
for the easy tasks, here's the
one for whatever the tech team. Here's

666
00:45:55.719 --> 00:46:00.960
the one that's for compliant data.
And we also allow they're set up on

667
00:46:01.000 --> 00:46:06.119
the hubside of safeguarding workflows so that
you can before the data gets sent out

668
00:46:06.199 --> 00:46:10.000
to say a cloud AI provider,
it gets screened for private information or maybe

669
00:46:10.000 --> 00:46:15.320
the data automatically it gets anonymized before
it gets sent out. Yeah, that's

670
00:46:15.440 --> 00:46:21.239
very important stuff and the US are
you also able to do some fin apps

671
00:46:21.400 --> 00:46:24.159
on that, In other words,
see how much it's costing to leverage this

672
00:46:24.239 --> 00:46:28.840
AI engine versus that AI engine and
do some cost optimization. Is that something

673
00:46:28.840 --> 00:46:32.039
you can do. We can do
that as part of a nine workflow and

674
00:46:32.079 --> 00:46:35.639
you could build that. You could
build that in there as well. But

675
00:46:35.639 --> 00:46:39.159
we are currently offering to our customers
the abilities to monitor our consumption so they

676
00:46:39.159 --> 00:46:43.480
have a bit of an eye on
that. But it's not automatically re routing

677
00:46:43.519 --> 00:46:47.000
to different AIS. But that's just
an added functionality under the wood m hm.

678
00:46:47.320 --> 00:46:51.280
Yeah, it's all in the workflows
basically, and that's where we're going

679
00:46:51.320 --> 00:46:54.679
in the last segment where I think
that we're just the beginning of leveraging these

680
00:46:54.719 --> 00:47:00.639
technologies because what they're really very good
at is pattern recognition, right, even

681
00:47:00.679 --> 00:47:04.920
just the vectorization the embedding is basically
how it stores it as a point in

682
00:47:05.039 --> 00:47:09.480
array basically, and just understanding how
it can map those two things. It's

683
00:47:09.519 --> 00:47:13.679
not just word not just text generation. I think we're going to get some

684
00:47:13.719 --> 00:47:17.800
really interesting things in terms of pattern
recognition and then recommendations. I mean,

685
00:47:17.800 --> 00:47:22.079
I think that these little AI agents, these assistants are going to be extremely

686
00:47:22.119 --> 00:47:27.440
helpful in all facets of business.
You know, to be able to very

687
00:47:27.519 --> 00:47:30.679
quickly give you a customer profile when
you're on the phone with someone, or

688
00:47:30.760 --> 00:47:34.840
to be able to give you summarization
of text on demand. I mean,

689
00:47:34.880 --> 00:47:37.960
really, I think the hardest challenge
is going to be changing mindsets and changing

690
00:47:38.039 --> 00:47:43.480
day to day behaviors and workflows.
What do you think final thoughts? I

691
00:47:43.559 --> 00:47:45.159
totally agree with that, But you
see that inside NIM as well. I

692
00:47:45.199 --> 00:47:47.880
mean, the developers were the first
ones that really said we want to use

693
00:47:47.880 --> 00:47:51.440
this, you want to use this, but then getting the rest of the

694
00:47:51.559 --> 00:47:53.400
organization to also seriously think about it. It's like, what can HR do,

695
00:47:53.480 --> 00:47:57.559
but can legal really do with it. It's a huge time saver and

696
00:47:57.599 --> 00:48:02.400
it's largely untipped. I totally give
with you check and they're interesting times ahead,

697
00:48:02.760 --> 00:48:06.360
yes, to say the least.
Well, what a fantastic conversation,

698
00:48:06.400 --> 00:48:08.880
folks, look them up online.
Michael Berthold b E R T H O

699
00:48:09.079 --> 00:48:13.239
L D from nine k N I
M E. We'll talk to you next

700
00:48:13.239 --> 00:48:19.639
time. Folks, you've been listening
to Inside Analysis. For more local radio

701
00:48:19.840 --> 00:48:27.199
every day, listen to KCAA.
Redlands Ranch Market is a unique, full

702
00:48:27.239 --> 00:48:31.760
service international or grocery store that specializes
in authentic food items from Mexico, India,

703
00:48:31.800 --> 00:48:37.559
and from many Mediterranean and Asian countries, including popular items from the US.

704
00:48:37.679 --> 00:48:40.719
They offer fresh baked items from their
in house bakery, housemade tortillas from

705
00:48:40.760 --> 00:48:45.719
their tortillarea, a delicious array of
prepared Mexican foods, a terrific fresh food

706
00:48:45.719 --> 00:48:50.840
and juice bar, and a large
selection of meats, seafoods and deli sandwiches,

707
00:48:50.960 --> 00:48:53.599
salads, and hellal meats. Their
produce department is stocked full with fresh,

708
00:48:53.719 --> 00:48:59.480
local and hard to find international fruits
and vegetables that you cannot find anywhere

709
00:48:59.480 --> 00:49:04.280
else. Don't forget to step into
the massive beer cave and experience the largest

710
00:49:04.320 --> 00:49:07.920
selection of domestic, artisan and imported
beers in the IE. They can also

711
00:49:08.039 --> 00:49:13.000
cater your next event with one of
the delicious takeout catering trays of food.

712
00:49:13.159 --> 00:49:17.119
Visit them at Redlands ranch Market dot
com. That's Redlands Ranch Market dot com.

713
00:49:17.159 --> 00:49:24.039
Redlands Ranch Market a unique and fun
shopping destination. To Hebotea Club's original

714
00:49:24.119 --> 00:49:29.039
pure power to rcosuper tea comes from
the only tree in the world that fungus

715
00:49:29.119 --> 00:49:31.559
does not grow on. As a
result, it naturally has anti fungal,

716
00:49:31.760 --> 00:49:38.159
anti infection, anti viral, antibacterial, anti inflammation, and anti parasite properties.

717
00:49:38.239 --> 00:49:42.480
So the tea is great for healthy
people because it helps build the immune

718
00:49:42.480 --> 00:49:46.119
system, and it can be truly
miraculous for someone fighting a potentially life threatening

719
00:49:46.159 --> 00:49:51.079
disease due to an infection, diabetes, or cancer. The tea is also

720
00:49:51.199 --> 00:49:54.280
organic and naturally caffeine free. A
one pound package of tea is forty nine

721
00:49:54.400 --> 00:49:59.440
ninety five which includes shipping. To
you order, please visit to Heboteaclub dot

722
00:49:59.480 --> 00:50:02.840
com. Ta Hebo is spelled T
like tom, a h ee b like

723
00:50:02.960 --> 00:50:07.840
boyo. They continue with the word
T and then the word club. The

724
00:50:07.920 --> 00:50:12.320
complete website is to Hebot club dot
com or call us at eight one eight

725
00:50:12.559 --> 00:50:16.000
six one zero eight zero eight eight
Monday through Saturday, nine am to five

726
00:50:16.039 --> 00:50:21.800
pm California time. That's eight one
eight six one zero eight zero eight eight

727
00:50:21.960 --> 00:50:28.159
to Hebot club dot com. With
sixty years of fascinating facts. This is

728
00:50:28.199 --> 00:50:32.320
the man from yesterday and back in
time. We go to this time in

729
00:50:32.400 --> 00:50:37.800
nineteen seventy, Stevie Wonder discloses he's
gonna marry Miss Sorita Wright in just a

730
00:50:37.800 --> 00:50:43.199
few months. Here in nineteen seventy, so Rita Wright is a secretary who

731
00:50:43.239 --> 00:50:47.000
works for a motown. Stevie Wonder
says they have been composing songs together,

732
00:50:58.360 --> 00:51:00.880
and from this time in twenty thirty, goodbye to Click and Clack, as

733
00:51:00.960 --> 00:51:07.440
car talk hosts Tom and Ray Magliosi
are retiring after doing their car talk radio

734
00:51:07.480 --> 00:51:12.239
program for some thirty five years.
Look, she said, I want something

735
00:51:12.239 --> 00:51:16.840
that goes from zero to two hundred
in four seconds or less, and my

736
00:51:17.000 --> 00:51:22.360
birthday's coming up. You could surprise
me for her birthday. He bought her

737
00:51:22.719 --> 00:51:31.199
a brand new bathroom scale, and
from this time in nineteen fifty five,

738
00:51:31.360 --> 00:51:37.199
NBC launches Monitor. It's a radio
program that lasts all weekend, and it's

739
00:51:37.199 --> 00:51:40.840
supposed to be a cousin of TV's
Today Show and Tonight Show in that the

740
00:51:42.159 --> 00:51:45.719
listener doesn't know what's coming up next. Listen again on the hour for NBC

741
00:51:45.880 --> 00:51:58.880
Monitor News with more at Man from
Yesterday dot Com. PAULM. Spreeze Dispensary

742
00:51:58.920 --> 00:52:02.639
reminds everyone during these challenging times they
understand the importance of mental health. Offering

743
00:52:02.719 --> 00:52:07.519
daily deals on everything from top quality
flowers to edibles and discounts on CBD products.

744
00:52:07.519 --> 00:52:10.840
Palm Springs Dispensary's goal is to help
you win the battle within. As

745
00:52:10.880 --> 00:52:15.760
one of the area's most beautiful wellness
shops. Located on Garnet Avenue, visit

746
00:52:15.920 --> 00:52:21.440
the Palm Springs Dispensary dot com.
That's the Palm Springs Dispensary dot Com.

747
00:52:21.719 --> 00:52:27.400
Hi, I'm Lannie Swardblow and I'm
back on KCAA ten fifty am and Express

748
00:52:27.480 --> 00:52:31.000
one oh six point five FM every
Tuesday at eight pm. My show is

749
00:52:31.039 --> 00:52:38.320
Beyond Common Sense. It's Lanny Sense
featuring me Lanni Swardlowe, kcaa's resident gay

750
00:52:38.719 --> 00:52:45.440
Jewish liberal potsmoking race mixing, left
handed atheist, an evangelical, fundamentalist,

751
00:52:45.639 --> 00:52:52.719
Christian nationalist, Worst Nightmare with subjects
that no one else will touch in quite

752
00:52:52.800 --> 00:52:57.440
the same way, Every Tuesday at
eight pm on Express one oh six point

753
00:52:57.480 --> 00:53:04.960
five FM, The Legacy ten fifty
AA and live screaming on Kcaradio dot Com.

754
00:53:05.119 --> 00:53:08.039
Bob Byla here with my home improvement
tip of the day. Gas powered

755
00:53:08.079 --> 00:53:12.559
fireplaces have been showing up in more
and more homes in the past few years.

756
00:53:12.840 --> 00:53:15.920
They're clean, easy to use,
and add a nice ambiance to the

757
00:53:15.960 --> 00:53:20.239
home. However, one drawback is
the noises they sometimes make. If you

758
00:53:20.280 --> 00:53:23.280
hear a popping noise when the burner's
on, it may signal small leaks around

759
00:53:23.360 --> 00:53:28.440
joints in the burner assembly. To
test for leaks, first turn off the

760
00:53:28.440 --> 00:53:31.639
burner and once the ceramic logs have
cooled off, remove them from the firebox.

761
00:53:32.119 --> 00:53:36.159
Next, mix a bit of liquid
detergent with water and pour it into

762
00:53:36.199 --> 00:53:39.599
a spray bottle. Turn the now
exposed burner assembly back on and look for

763
00:53:39.639 --> 00:53:44.159
any small bursts of flame that may
be popping around the joints. If you

764
00:53:44.159 --> 00:53:46.960
don't see any, try spraying a
little detergent makes on the various joints and

765
00:53:47.000 --> 00:53:52.079
fittings in the burner assembly. If
you see bubbles, you found the leak.

766
00:53:52.559 --> 00:53:54.280
If you can't find a leak,
or the leak you find appears to

767
00:53:54.280 --> 00:53:58.280
be a hole in the assembly itself, you'll want to call in a pro.

768
00:53:58.679 --> 00:54:01.400
Get more info at bobyla dot com
and right here at home to meet

769
00:54:01.480 --> 00:54:08.840
Bob Deever, you like to safely
leverage bank money to earn double digit returns

770
00:54:09.039 --> 00:54:14.679
income tax free, with guarantees and
no downside market risk. How can you

771
00:54:14.719 --> 00:54:17.400
do this? This is Farence host
of the Your Personal Bank Show. One

772
00:54:17.679 --> 00:54:22.280
You fund a high cash value policy
one time to earn dividends and interest.

773
00:54:22.400 --> 00:54:28.079
Two establish a bank line of credit
using the cash in your policy as collateral.

774
00:54:28.199 --> 00:54:30.599
When you earn more in dividends from
your policy than the interest the bank

775
00:54:30.719 --> 00:54:35.440
charges, you keep the difference,
and the difference is average two to five

776
00:54:35.480 --> 00:54:38.760
percent annually in your favor for the
past forty plus years. Three, the

777
00:54:38.840 --> 00:54:44.719
bank funds contributions years two to twenty
plus. Each year the bank adds funds,

778
00:54:44.719 --> 00:54:46.960
your rate of return increases. Your
average rate of return can grow too

779
00:54:47.000 --> 00:54:52.440
strong, double digits annually within a
few years. Contact me you your Personal

780
00:54:52.480 --> 00:54:55.559
bank dot com. Your Personal Bank
dot com or eight sixty six two six

781
00:54:55.679 --> 00:55:00.039
eight four four two two eight sixty
six two six eight four two two for

782
00:55:00.159 --> 00:55:06.079
more info, or tune in to
your Personal Bank show. Your Personal Bank

783
00:55:06.119 --> 00:55:09.920
Show airs Tuesdays at four pm right
here on CASEAA ten fifty am and one

784
00:55:09.920 --> 00:55:15.400
oh six point five am. This
station that leaves no listeners behind. KCAA

785
00:55:15.599 --> 00:55:19.119
Radio has openings for one hour talk
shows. If you want to host a

786
00:55:19.239 --> 00:55:22.639
radio show, now is the time. Make kca your flangship station. Our

787
00:55:22.719 --> 00:55:27.679
rates are affordable and our services are
second to none. We broadcast to a

788
00:55:27.719 --> 00:55:31.559
population of five million people plus.
We stream and podcast on all major online

789
00:55:31.559 --> 00:55:37.480
audio and video systems. If you've
been thinking about broadcasting a weekly radio program

790
00:55:37.559 --> 00:55:42.519
on real radio plus the internet,
contact our CEO at two eight one five

791
00:55:42.760 --> 00:55:46.360
nine nine ninety eight hundred two eight
one five nine nine ninety eight hundred.

792
00:55:46.440 --> 00:55:50.840
You could skype your show from your
home to our Redlands, California studio,

793
00:55:50.880 --> 00:55:54.159
where our live producers and engineers are
ready to work with you personally. A

794
00:55:54.239 --> 00:56:00.400
radio program on KCAA is the perfect
work from home advocation in these ut full

795
00:56:00.400 --> 00:56:05.119
times. Just type kca radio dot
com into your browser to learn more about

796
00:56:05.119 --> 00:56:07.880
hosting a show on the best station
in the nation, or call our CEO

797
00:56:08.000 --> 00:56:15.440
for details too. Eight one five
eight hundred. An auto dealer says he's

798
00:56:15.440 --> 00:56:19.719
saving all the good cars for himself. It's the Onion Radio News. This

799
00:56:19.880 --> 00:56:24.920
is Doyle Redland reporting. Owner Jim
Gannetti of Gannetti Chevrolet admitted to reporters today

800
00:56:25.039 --> 00:56:30.159
that his slogan the best cars for
the best prices is a lie. Instead

801
00:56:30.159 --> 00:56:35.679
of selling the best cars, Gonnetti
has secretly been keeping all the best cars

802
00:56:35.679 --> 00:56:39.280
for himself. He now says the
guilt and storage costs have finally forced him

803
00:56:39.320 --> 00:56:44.320
to come clean. It was too
much. Every time I heard my own

804
00:56:44.440 --> 00:56:47.000
jingle, I I just felt like
a fraud. Gnnetti added that he was

805
00:56:47.039 --> 00:56:52.920
afraid of becoming like his father,
legendary auto dealer Carl Gannetti, who died

806
00:56:52.960 --> 00:56:58.320
alone in a house filled with two
hundred eighty three Sedans. Doyle Redland for

807
00:56:58.360 --> 00:57:02.119
The Onion Radio News online at the
Onion dot com. Del Walmsley here,

808
00:57:02.239 --> 00:57:06.039
the first thing you're going to have
to learn is that until you stop expecting

809
00:57:06.119 --> 00:57:08.800
our politicians or anyone else to change
your life, your life isn't going to

810
00:57:08.920 --> 00:57:13.079
change. The only person who can
change your life is you. But you

811
00:57:13.159 --> 00:57:15.480
need to know how. Listen to
my show, The Del Walmsley Radio Show,

812
00:57:15.519 --> 00:57:21.000
where the Hype ends and the help
begins right here on CACAA now broadcasting

813
00:57:21.039 --> 00:57:24.159
on ten fifty AM and one oh
six point five FM, the stations that

814
00:57:24.239 --> 00:57:31.800
leave no listener Behind NBC News Radio, I'm Chris Gragio. At least fifteen

815
00:57:31.880 --> 00:57:37.000
storm related deaths are being reported after
severe weather slammed the nation's midsection. Authorities

816
00:57:37.000 --> 00:57:40.280
in Texas say seven people, including
two children, were killed when a tornado

817
00:57:40.480 --> 00:57:45.079
roared through Cook County, just north
of Dallas. Other deaths were reported in

818
00:57:45.079 --> 00:57:50.039
Oklahoma, Arkansas, and Kentucky as
violent storms caused widespread damage and knocked out

819
00:57:50.079 --> 00:57:54.559
power to tens of thousands. The
International Criminal Court is facing sharp criticism from

820
00:57:54.599 --> 00:57:59.760
both sides of the Aisle for pursuing
arrest warrants for top Israeli officials over the

821
00:57:59.800 --> 00:58:05.159
world in Gaza, Florida. Democrat
Jared Moskowitz on Fox News Sunday Today said

822
00:58:05.199 --> 00:58:08.679
the move by the ICC was politically
motivated. The ICCs are relevant, they

823
00:58:08.719 --> 00:58:13.320
have no jurisdiction. We might as
well call them the Harry Potter Ministry.

824
00:58:13.360 --> 00:58:16.800
Of Magic. Congressman Moscowitz said there's
no equivalence between Israel and Hamas. The

825
00:58:16.880 --> 00:58:22.719
chief prosecutor at the ICC is seeking
arrest warrants for Israeli Prime Minister Benjamin net

826
00:58:22.840 --> 00:58:27.519
Yahoo and Hamas officials, as the
conflict in Gaza has killed tens of thousands

827
00:58:27.519 --> 00:58:31.719
of civilians. The Hushmuny trial of
former President Donald Trump heads into its final

828
00:58:31.719 --> 00:58:37.679
phase this week. Larry Kovsky Now
with a preview. Closing arguments are expected

829
00:58:37.719 --> 00:58:42.679
Tuesday in a Manhattan courtroom, and
Senator Tim Scott predicts the former president will

830
00:58:42.719 --> 00:58:45.960
be found not guilty. Appearing on
CNN State of the Union, the South

831
00:58:46.000 --> 00:58:51.960
Carolina Republican said Americans no, when
they see a two tiered justice system,

832
00:58:52.039 --> 00:58:55.800
we don't want to see the weaponizing
of our justice system against our political opponents.

833
00:58:57.000 --> 00:59:00.119
We want to see fairness, no
thumb on the scale. If that's

834
00:59:00.159 --> 00:59:05.760
the case, he will be found
innocent. Trump faces thirty four Fella accounts

835
00:59:05.760 --> 00:59:09.320
in the case in which he is
accused of falsifying business records related to a

836
00:59:09.360 --> 00:59:15.320
payment to adult film actor Stormy Daniels. Larry Kovsky reporting and the NBA's Western

837
00:59:15.320 --> 00:59:21.119
Conference finals continue tonight in Dallas.
With the Mavericks currently up two games to

838
00:59:21.199 --> 00:59:28.320
none on the Minnesota Templewolves. I'm
Chris Karragio, NBC News Radio, NBC

839
00:59:28.400 --> 00:59:32.800
News on KCAA, Lowlanda sponsored by
Teamsters Local nineteen thirty two, Protecting the

840
00:59:32.840 --> 00:59:43.239
Future of Working Families Teamsters nineteen thirty
two dot org. You're listening to an

841
00:59:43.360 --> 01:00:00.199
encore presentation of this program KCAA,
The Inland Talk Express and Return to Zulu

842
01:00:00.239 --> 01:00:05.800
out the Justice Watch Crew, Rose
of nu Yez, Michael Bloud, Parsh,

843
01:00:05.960 --> 01:00:08.400
Doctor Kilbasher. Today, like each
week, we'll be discussed