WEBVTT

1
00:00:02.879 --> 00:00:07.799
The information economy as a rod.
The world is teeming with innovation as new

2
00:00:07.839 --> 00:00:14.080
business models reinvent every Intensis is your
source of information and insights about how to

3
00:00:14.119 --> 00:00:18.920
make the most of this exciting new
era. Learn more and Inside Analysis dot

4
00:00:18.920 --> 00:00:28.359
Comside Analysis dot com. And now
here's your host through Eric Kavanaugh. All

5
00:00:28.440 --> 00:00:31.800
Right, ladies and gentlemen, Hello
and welcome back once again to the only

6
00:00:31.879 --> 00:00:35.679
coast to coast radio show that's all
about the information economy. It's called Inside

7
00:00:35.679 --> 00:00:39.359
Analysis. Your host here, Eric
Kavanaugh. On a very special episode,

8
00:00:39.399 --> 00:00:44.320
folks were going to dive into the
roaring hottest topic in the world of analytics

9
00:00:44.320 --> 00:00:49.840
and AI and business intelligence as well. These days the large language models,

10
00:00:49.960 --> 00:00:54.359
or foundational models as they're called AI. Chat bots, for example, have

11
00:00:54.520 --> 00:00:58.000
been around for quite a few years
now. Chatbots they were a bit rusty,

12
00:00:58.240 --> 00:01:00.240
I think when they first came on
the market. Maybe sometimes they weren't

13
00:01:00.679 --> 00:01:06.640
quite ready for prime time. But
my how things have changed, and certainly

14
00:01:06.680 --> 00:01:11.120
open Ai threw down the gauntlet with
chat GPT, but then Google has barred.

15
00:01:11.400 --> 00:01:15.519
Of course, GitHub has co pilots. You have data breaks with Dolly

16
00:01:15.640 --> 00:01:19.799
two point zero already, So these
things are suddenly everywhere and everyone's talking about

17
00:01:19.799 --> 00:01:25.760
in generative AI. That's the sort
of broader subject area involved here, and

18
00:01:25.799 --> 00:01:29.239
we're very pleased to have with us
Mike Finley. He's the co founder and

19
00:01:29.319 --> 00:01:33.359
chief technology officer for a company called
Answer Rockets. They've been in the BI

20
00:01:33.439 --> 00:01:36.920
space for quite some time now.
We've been tracking them doing some pretty interesting

21
00:01:36.959 --> 00:01:41.040
things, and we're going to talk
about Max. Introducing Max, and I

22
00:01:41.079 --> 00:01:45.040
have to say, remember the old
movie It Wasn't Gunball Rally was one of

23
00:01:45.120 --> 00:01:47.719
those kind of movies. Though in
the end the guy goes push the button

24
00:01:47.799 --> 00:01:52.680
Max, he goes boom. All
these crazy things happen. But Max sounds

25
00:01:52.719 --> 00:01:57.079
pretty interesting to me and folks,
the extent to which these large language models

26
00:01:57.079 --> 00:02:02.799
and foundation role models can change your
workflow is really quite remarkable. One of

27
00:02:02.799 --> 00:02:06.959
the smartest people I've met in a
while. We interviewed him last week down

28
00:02:06.959 --> 00:02:09.840
in Austin, Texas, and he
was telling me he'll get technical briefs that

29
00:02:09.879 --> 00:02:14.159
are one hundred pages long that he
has to consume over the weekend. He'll

30
00:02:14.199 --> 00:02:16.439
just dump it right in the chat
cheepts. Give me a three page summary,

31
00:02:16.879 --> 00:02:22.680
Boom done. I mean, that's
like a dynamic cliffs notes engine basically

32
00:02:22.800 --> 00:02:25.439
is what you can use it for. So these things are very very powerful,

33
00:02:25.919 --> 00:02:30.319
and they're very complex under the hood. They could do some amazing things.

34
00:02:30.439 --> 00:02:32.240
Now, there are these hallucinations we've
heard about. Maybe we'll talk about

35
00:02:32.280 --> 00:02:36.439
that, but let's get down to
brass tacks and bring in Mike Finley from

36
00:02:36.439 --> 00:02:39.080
answer Rocket. Thanks for your time
today. Tell us a bit about MAX

37
00:02:39.159 --> 00:02:44.639
and how MAX is helping with your
bi initiatives. Sure, Eric, thank

38
00:02:44.680 --> 00:02:47.719
you, and thanks for that introduction
and the good launch for the conversation.

39
00:02:49.000 --> 00:02:53.560
Yeah. So answer rocket helps businesses
have I like to say they have conversations

40
00:02:53.560 --> 00:02:55.879
with their data. Right. If
you think of all the data that a

41
00:02:55.960 --> 00:03:00.120
business has, and it's expensive to
collect it, to collate it, to

42
00:03:00.199 --> 00:03:04.560
save it away somewhere, being able
to have a conversation with that data,

43
00:03:04.680 --> 00:03:07.439
right, to treat it like it's
the brain of the business, like it's

44
00:03:07.479 --> 00:03:10.240
a simulation of the operations, so
you can go in and say what happened

45
00:03:10.240 --> 00:03:14.000
and when and why? Right,
That's that's what answer rocket does. Right,

46
00:03:14.039 --> 00:03:16.360
That's that's our expertise. Um,
not so much. I want to

47
00:03:16.400 --> 00:03:21.159
set right that's been done many many
times. But answer rocket really is about

48
00:03:22.240 --> 00:03:25.639
allowing that conversational interaction with with your
database. And that's what it's been,

49
00:03:27.039 --> 00:03:29.120
uh, you know, for ten
years now, and of course with the

50
00:03:29.159 --> 00:03:35.520
advent of language models, specifically most
recently with gpt UM, that's suddenly taken

51
00:03:35.560 --> 00:03:39.599
a huge leap forward in terms of
that conversational capability, not just to be

52
00:03:39.639 --> 00:03:44.039
able to identify what the user wants
if they are clever enough to sort of

53
00:03:44.080 --> 00:03:46.599
walk their way through the right keywords
and the right commands, but just to

54
00:03:46.639 --> 00:03:51.879
be able to vary candidly in any
language of their choice, have that conversation

55
00:03:52.039 --> 00:03:54.960
with the data, uh, and
have have that that agent which is max

56
00:03:55.240 --> 00:03:59.800
feel like a coworker who really understands, who really grocks what's going on in

57
00:03:59.800 --> 00:04:02.879
the business. That's what we built
well. And what's cool is these are

58
00:04:03.000 --> 00:04:11.479
text generators, but they generate syntactically
correct tax they generate meaning on having absorbed

59
00:04:11.960 --> 00:04:16.120
goodness knows how many volumes of text
and the training of these models, and

60
00:04:16.160 --> 00:04:23.560
so they are very powerful and they
can very quickly ascertain certain aspects of your

61
00:04:23.600 --> 00:04:27.319
business, assuming you've connected the data
to them or fed the data into them.

62
00:04:27.360 --> 00:04:30.519
Basically to feed this data into a
chat gpt right, and every one's

63
00:04:30.519 --> 00:04:33.199
out there too, Dolly, whatever, whichever one you're talking about. Of

64
00:04:33.240 --> 00:04:36.360
course, cold Pilot is great for
coding, right and geth hub it's it's

65
00:04:36.360 --> 00:04:42.079
doing the coding side. And Gmail
was on this I think first, at

66
00:04:42.160 --> 00:04:46.319
least in terms of general availability.
You would note in Gmail if you're in

67
00:04:46.360 --> 00:04:47.040
the condician your sentence, it's I
was like, you know, a year

68
00:04:47.040 --> 00:04:49.920
and a half, two years ago. Yeah, I was like, oh

69
00:04:49.920 --> 00:04:54.959
wow, this is getting pretty interesting, and it was pretty good zoo.

70
00:04:55.079 --> 00:05:00.199
Right, So these foundational models,
and it's basically a language AI bot that

71
00:05:00.279 --> 00:05:03.480
can interpret your data and then give
you some insights about it, right,

72
00:05:04.199 --> 00:05:09.319
Yeah, so absolutely. And there's
the thing that the language model is really

73
00:05:09.319 --> 00:05:13.000
good at is that interaction with the
human. It's good at understanding what the

74
00:05:13.079 --> 00:05:16.920
human means, like the subtleties and
the intense, and it's good at explaining

75
00:05:16.920 --> 00:05:19.519
the answer back to the human.
Right, that's the part that it really

76
00:05:19.600 --> 00:05:24.680
was trained to do. That's what
it's really good good at is faithfully looking

77
00:05:24.759 --> 00:05:28.160
up and using all the correct facts
that are in your database. Right.

78
00:05:28.360 --> 00:05:30.639
And so that's where answer rocket comes
in. That's why this is such a

79
00:05:30.720 --> 00:05:34.240
strong partnership between answer rocket and a
language model. It says, the language

80
00:05:34.240 --> 00:05:39.120
model can hold that that interaction with
the human with all the subtleties, with

81
00:05:39.160 --> 00:05:44.639
all the conversational aspects of evolution,
of solving a problem, but allowing answer

82
00:05:44.680 --> 00:05:47.639
Rocket on the back end to both
retrieve the correct data, make the correct

83
00:05:47.800 --> 00:05:53.199
decisions and calculations from that data,
and then fact check the results that before

84
00:05:53.199 --> 00:05:56.199
the language model ever presents it back. Right, So you really have to

85
00:05:56.399 --> 00:05:59.839
You have to treat the language model
like like you would treat a human co

86
00:06:00.079 --> 00:06:02.040
worker. What would you do?
You would train them on your business.

87
00:06:02.519 --> 00:06:06.439
You would make sure that they are
an expert in the area that you're talking

88
00:06:06.480 --> 00:06:10.240
about, and you would fact check
their results. That's what you would do

89
00:06:10.240 --> 00:06:12.959
if it was were a human that
you're talking to. So you have to

90
00:06:13.000 --> 00:06:15.839
do the same thing with the language
model. An answer Rocket does that by

91
00:06:15.000 --> 00:06:19.319
essentially wrapping that interaction. So I
had a great fund one the other day.

92
00:06:19.360 --> 00:06:21.560
You know, a customer said,
hey, what were the top five

93
00:06:21.639 --> 00:06:29.480
products in you know, this category
last quarter? And Max sent that to

94
00:06:29.519 --> 00:06:31.079
the language model and said, hey, language model, what do they want?

95
00:06:31.319 --> 00:06:33.600
A Language model came back and said, oh, they want you to

96
00:06:33.839 --> 00:06:41.279
examine the product table and rank the
top five members for the whatever, right

97
00:06:41.560 --> 00:06:45.319
and so answer Rocket went away and
got that data, produced the answer back

98
00:06:45.360 --> 00:06:46.680
and said, by the way of
these five, this one's an outlier.

99
00:06:46.800 --> 00:06:49.759
Right. That's something that GPT wouldn't
have known by itself, but we were

100
00:06:49.800 --> 00:06:54.480
able to inject that in. Now
GPT gives a great answer back to the

101
00:06:54.560 --> 00:06:57.319
user saying, oh, there these
are the top five. Good news,

102
00:06:57.319 --> 00:07:00.160
there was an outlier, and it
looks like that was due to growth in

103
00:07:00.199 --> 00:07:03.959
California. Right. And then the
user followed up with um, um okay,

104
00:07:04.040 --> 00:07:10.160
which ones were the dogs? And
so to those kinds of questions.

105
00:07:10.160 --> 00:07:14.399
This is really where the language model
and the bi tool working together is beautiful,

106
00:07:14.639 --> 00:07:17.680
because if I tried to write code
that said, if then else to

107
00:07:17.720 --> 00:07:20.879
figure out what were the dogs,
right or which ones were the dogs,

108
00:07:23.240 --> 00:07:26.279
I'd be writing code until I'm retired, and um, you know, I'd

109
00:07:26.360 --> 00:07:29.519
never have a chance to finish.
The language model knew exactly what they meant.

110
00:07:29.519 --> 00:07:31.879
It said, switch to bottom that's
it, from top from top five,

111
00:07:31.920 --> 00:07:35.120
switch to the bottom five and answer
the same question it was. It's

112
00:07:35.199 --> 00:07:39.959
it's those moments, are I mean, it's surely goosebump before a guy like

113
00:07:40.079 --> 00:07:45.279
me who's been intech long time never
really faced a problem that I didn't know

114
00:07:45.319 --> 00:07:47.680
how to code. Uh you know, it's it's um. And now these

115
00:07:47.759 --> 00:07:51.319
language models are just doing things that
are fundamentally different. I mean, imagine

116
00:07:51.319 --> 00:07:55.000
if you brought a contractor into your
house and you said, hey, I

117
00:07:55.360 --> 00:07:57.839
want to add a wall here,
and they just scratch their head and said,

118
00:07:58.319 --> 00:08:00.399
I don't know how to add a
wall there. That can be done,

119
00:08:00.519 --> 00:08:03.480
right. It's almost like like somebody
said, make me a floating bathtub.

120
00:08:03.959 --> 00:08:05.839
I don't know how to make a
floating bathtub. That can't be done.

121
00:08:05.959 --> 00:08:09.439
These language models are doing things that
are so fundamentally new. And again

122
00:08:09.920 --> 00:08:13.160
you hear a lot about this in
the press. Not even their creators know

123
00:08:13.240 --> 00:08:15.800
exactly how they work. And so
I think we're just at the we're at

124
00:08:15.800 --> 00:08:20.199
the beginning of all the possibilities where
we've seen that. It's kind of like

125
00:08:20.199 --> 00:08:24.040
that old Star Trek transparent aluminum moment, you know, when Scotty needed to

126
00:08:24.040 --> 00:08:28.279
save a whale from the twentieth century
or something and he made transparent aluminum.

127
00:08:28.399 --> 00:08:31.639
Right, we have no idea.
We're at that moment. We know we

128
00:08:31.679 --> 00:08:35.200
have transparent alluminum, what do we
do with it? Right? This is

129
00:08:35.240 --> 00:08:39.000
where all the engineering wrapping the language
model is coming about now. And this

130
00:08:39.080 --> 00:08:41.720
is the whole ecosystem, the fountain
of companies that are starting up around this

131
00:08:41.799 --> 00:08:45.679
tech, you know, is so
this is very interesting. So I think

132
00:08:45.679 --> 00:08:50.279
about data warehousing, business intelligence,
even business analytics to kind of stretch it

133
00:08:50.320 --> 00:08:56.600
a bit further, all of these
disciplines revolve around transactional data and it's really

134
00:08:56.879 --> 00:09:01.440
numbers. I mean, there are
tags that assigned to objects like products,

135
00:09:01.480 --> 00:09:07.360
for example, services, but really
you're talking about the numbers game that's moving

136
00:09:07.399 --> 00:09:11.639
all around. Well, what's missing
in that is the language to interpret and

137
00:09:11.720 --> 00:09:16.759
explain what the ugers mean. And
that's the magic that these large language models

138
00:09:16.759 --> 00:09:22.639
are bringing into the equation is they
can find ways to represent the facts in

139
00:09:22.639 --> 00:09:28.559
a fashion that is meaningful and understandable
and inspiring to the user. Right,

140
00:09:28.919 --> 00:09:31.759
that's right, and that's really working
at the user's level. Right. So,

141
00:09:31.759 --> 00:09:33.720
if you go, if you go
really back to the beginning of computing,

142
00:09:33.840 --> 00:09:37.279
right in the eighteen hundreds, you
know, you think of a problem,

143
00:09:37.440 --> 00:09:41.960
you express it to the machine as
a program. So right away you

144
00:09:41.000 --> 00:09:46.159
are having to write your question in
the computer's language. The answer comes back

145
00:09:46.200 --> 00:09:48.200
to you as a spreadsheet. Right, so you are getting the answer back

146
00:09:48.240 --> 00:09:52.960
in the computer's language. That paradigm
is broken. One hundred and fifty years

147
00:09:52.000 --> 00:09:56.039
in we finally broke out right.
Now now we can ask the question on

148
00:09:56.120 --> 00:10:01.480
our language and we get the answer
back in our uage as well. Uh.

149
00:10:01.519 --> 00:10:03.720
The analogy that I like to make
is, um, you know data,

150
00:10:03.919 --> 00:10:07.519
That data is digitizing your business.
Right, it's a it's you know,

151
00:10:07.559 --> 00:10:11.720
it's uh uh, it's the matrix. Right. The data should show

152
00:10:11.720 --> 00:10:15.279
you everything about your business. If
you just know how to look at the

153
00:10:15.360 --> 00:10:18.600
data, you can see the business
right. And this is this is exemplified

154
00:10:18.639 --> 00:10:22.919
by seasoned executives who can look at
a wall of numbers and spreadsheets and say,

155
00:10:22.919 --> 00:10:26.159
oh my god, what are we
going to do? Right? Um,

156
00:10:26.200 --> 00:10:28.240
and nobody else in the room knows
what it is, but they know

157
00:10:28.320 --> 00:10:31.360
the numbers to look for. Well, um, if you try to if

158
00:10:31.360 --> 00:10:35.080
you try to um uh digitize the
business like that, you lose a lot

159
00:10:35.120 --> 00:10:39.240
of the subtleties unless you have all
of that depth. Right. So it's

160
00:10:39.320 --> 00:10:41.440
kind of like you know, if
I if I digitize Romeo and Juliet,

161
00:10:41.519 --> 00:10:45.840
well, it's boy meets girl,
boy falls in love with girl. Boy

162
00:10:45.879 --> 00:10:50.039
loses girl. Right, it becomes
a very boring story. Right. I

163
00:10:50.080 --> 00:10:54.120
can't I can't digit I can't make
transactional records out of what happened to my

164
00:10:54.159 --> 00:10:58.440
business. The story is more subtle
than that. The story is unstructured,

165
00:10:58.840 --> 00:11:01.120
right, And and that's where that's
where these language models are helping us.

166
00:11:01.360 --> 00:11:05.159
Take that data, which by definition
has to be structured because I have to

167
00:11:05.200 --> 00:11:09.840
be able to record a series of
things that happen in the business and turn

168
00:11:09.919 --> 00:11:13.759
them into unstructured information, which is
really where all the subtleties are, right,

169
00:11:13.759 --> 00:11:18.879
It's where all the competitive edges,
all the opportunities to improve, you

170
00:11:18.919 --> 00:11:22.480
know, the whole idea of strength
and weaknesses. Right, these are soft

171
00:11:22.519 --> 00:11:26.120
things. These aren't digitized. And
this is really where the language models is

172
00:11:26.159 --> 00:11:31.279
helping is helping business users shine.
Yeah. Well, the data is like

173
00:11:31.320 --> 00:11:35.840
the wire frame essentially, and then
the lms are filling in all the colors

174
00:11:35.879 --> 00:11:41.720
and the shapes and the dimensions and
the fluidity for example, to view this

175
00:11:41.919 --> 00:11:46.279
rich experience to help understand what's going
on. And you know, in terms

176
00:11:46.320 --> 00:11:50.159
of structured data versus unstructured data,
well, you can kind of get into

177
00:11:50.159 --> 00:11:54.840
a semantics debate about the fact that
languages do have structure, it's center acts,

178
00:11:56.399 --> 00:12:01.840
and these llms understand that syntax is
probably better than people do and can

179
00:12:01.960 --> 00:12:05.440
explain things to you. And that's
really the other interesting side of the equation,

180
00:12:05.519 --> 00:12:09.039
right is you are having that conversation
with your data. We've talked about

181
00:12:09.080 --> 00:12:13.080
that for thirty years and it's business, but it's never really been that.

182
00:12:13.240 --> 00:12:18.080
It's always been playing with the numbers, looking at visualizations, kind of identifying

183
00:12:18.120 --> 00:12:22.120
patterns. But to have an engine
then explain that to you in English or

184
00:12:22.240 --> 00:12:28.720
French or German or code, that
gets very, very interesting because now you

185
00:12:28.759 --> 00:12:33.159
can really flesh out your understanding and
it is a process you go back and

186
00:12:33.159 --> 00:12:35.759
forth and say, hey, can
you explain this? Can you explain that

187
00:12:35.320 --> 00:12:39.240
I saw data bricks was talking about
with Dolly, how you can use it

188
00:12:39.279 --> 00:12:43.919
to explain your code to you?
So, you know, let's think about

189
00:12:43.279 --> 00:12:48.919
digital transformation, which of course is
hugely important these days. And how many

190
00:12:48.919 --> 00:12:52.159
of these old programs are written in
coball. Well, there aren't that many

191
00:12:52.200 --> 00:12:54.480
developers that write coball anymore. A
lot of them are retired, and so

192
00:12:54.519 --> 00:12:58.200
this has been a big sticking point. But it doesn't have to be because

193
00:12:58.200 --> 00:13:03.799
you can just take corpus of code, PLoP it into your single instance chat,

194
00:13:03.799 --> 00:13:07.480
GBT or whatever and say, explain
to me what this is doing,

195
00:13:07.679 --> 00:13:11.840
and it'll do that, right,
sure, Yeah, extracted or translated or

196
00:13:11.840 --> 00:13:13.720
do any number of things. Yeah. And you know, keep in mind

197
00:13:13.720 --> 00:13:16.360
it's you're still in that, You're
still in that. Treat it like you

198
00:13:16.360 --> 00:13:20.679
would treat a human in other words, Uh, make sure it's trained on

199
00:13:20.720 --> 00:13:22.279
the thing you're asking it to do, and then verify the results. Right.

200
00:13:22.360 --> 00:13:28.360
So there's no free lunch even with
the language models, but there is

201
00:13:28.399 --> 00:13:31.639
a lot of acceleration that comes from
from the capabilities that they're bringing that and

202
00:13:31.679 --> 00:13:35.840
I think that's the term. I
think we should pick on that term and

203
00:13:35.879 --> 00:13:41.159
really drill into it. It does
accelerate the process of learning, of designing,

204
00:13:41.559 --> 00:13:46.480
of understanding, of sharing, of
articulating. All these things are accelerated

205
00:13:46.480 --> 00:13:50.480
by this kind of an engine.
And you know, as a writer myself,

206
00:13:50.559 --> 00:13:54.279
I'm a little bit leary of just
using it to do writing, but

207
00:13:54.320 --> 00:13:58.519
I will say it's very good at
giving you the contours of a conversation,

208
00:13:58.559 --> 00:14:01.679
are very good at giving you the
pill is of a certain topic. If

209
00:14:01.679 --> 00:14:05.480
we want to talk about data science, what are the top five issues in

210
00:14:05.600 --> 00:14:09.600
data science. It will give you
five really good issues. That's right for

211
00:14:09.720 --> 00:14:11.919
data science. Absolutely absolutely. It's
like a teacher that's helping you all along

212
00:14:11.919 --> 00:14:16.360
the way, right, yes,
so um, kind of like improv yes

213
00:14:16.440 --> 00:14:20.679
and right um. And the end
part is that by the same by the

214
00:14:20.720 --> 00:14:24.080
same token, that it knows everything
there is to know about things that are

215
00:14:24.159 --> 00:14:30.039
general to the marketplace. It knows
very little about your business. It knows

216
00:14:30.120 --> 00:14:33.360
very little about the seasonal behavior of
the sales of one of your products.

217
00:14:33.360 --> 00:14:39.039
It knows very little about the supply
chain issues that you have in monsoon season.

218
00:14:39.159 --> 00:14:43.360
Right. It knows very little about
your your competitor and your price wars.

219
00:14:43.440 --> 00:14:48.039
Right. So that's where that's where
again this one two combination of that

220
00:14:48.159 --> 00:14:52.840
deically and how max works is really
to say, hey, the language model

221
00:14:52.919 --> 00:14:56.399
is talking to the user about the
data, but answer Rocket is feeding in

222
00:14:56.519 --> 00:15:01.440
those observations, those facts, those
insights, not just in the form of

223
00:15:01.440 --> 00:15:05.399
saying January was fifteen million February it
was twenty million. No. Instead,

224
00:15:05.440 --> 00:15:09.960
where January was an outlier even though
seasonally we expected it to be high,

225
00:15:11.000 --> 00:15:13.879
it was actually way outside of its
even seasonal high parameters. And the reason

226
00:15:15.039 --> 00:15:16.879
is these three drivers, and those
drivers, one of them, by the

227
00:15:16.879 --> 00:15:20.360
ways, that a record high for
all time. We provide that feedback back

228
00:15:20.399 --> 00:15:24.039
to the language model. It gets
a chance to say, oh, I've

229
00:15:24.080 --> 00:15:31.080
seen a bunch of recent highs because
everybody's moving to South Carolina, or maybe

230
00:15:31.120 --> 00:15:33.000
it says, oh, that is
an unusual thing. I get a chance

231
00:15:33.039 --> 00:15:37.200
to digest that and provide it back
to the user. So it's it's the

232
00:15:37.279 --> 00:15:41.440
partnering on on one side, the
answer rocket being able to really rock your

233
00:15:41.440 --> 00:15:46.519
business and dive in and understand your
data, but the language model being able

234
00:15:46.559 --> 00:15:48.799
to make sense and have a conversation
about those facts back to the user.

235
00:15:50.320 --> 00:15:56.399
Yeah, it provides the contact,
It provides the language to explain what's happening,

236
00:15:56.440 --> 00:15:58.600
and you do have to kind of
iterate, and I think in our

237
00:15:58.639 --> 00:16:03.039
next segment we'll get into a bit
more detail on how you actually deploy these

238
00:16:03.080 --> 00:16:07.000
things, because you want to deploy
a chat, GPT or one of these

239
00:16:07.000 --> 00:16:11.679
other various AI engines to be able
to have some controls around that, right,

240
00:16:11.720 --> 00:16:15.159
because this is one issue that you
have to watch out for, and

241
00:16:15.240 --> 00:16:18.759
it's just a question of knowing what
the tool can do and what the tool

242
00:16:18.840 --> 00:16:22.799
can't do what it's designstering to do. And just very quickly here, I

243
00:16:22.840 --> 00:16:26.840
remember when the New York Times editor
was asking it all these esoteric questions and

244
00:16:26.960 --> 00:16:30.080
existential questions and all these things,
and it got all this weird stuff back.

245
00:16:30.799 --> 00:16:33.720
That's not how the tool is meant
to be used, right, So

246
00:16:33.759 --> 00:16:40.639
this is it's sort of an interesting
journey to see what happens. But to

247
00:16:40.679 --> 00:16:44.480
preclude or I'm sorry to conclude that
it must have been sentient because they're saying

248
00:16:44.480 --> 00:16:48.399
these things as a pretty significant stretch, right, absolutely, And you could

249
00:16:48.440 --> 00:16:53.159
even take it a step further and
say, essentially that journalist was basically writing

250
00:16:53.200 --> 00:16:57.279
a story and the language model was
simply helping him finish that story in an

251
00:16:57.279 --> 00:17:00.799
interesting way. Right. It didn't
want to marry. I mean, it

252
00:17:00.879 --> 00:17:04.079
had no concept of being married,
right, right, But there are a

253
00:17:04.079 --> 00:17:08.000
lot of novels out there where where
you know, there's conversations that happen like

254
00:17:08.039 --> 00:17:14.359
that. So absolutely, it's it's
um, it's incorrect to personify the language

255
00:17:14.359 --> 00:17:19.039
model. It's more like realizing that
actually it's talking about a cultural issue that

256
00:17:19.359 --> 00:17:22.599
the journalists tapped into. Yeah,
no, that's exactly right. So you

257
00:17:22.640 --> 00:17:26.839
do need to understand what these tools
are good for, what they do,

258
00:17:26.960 --> 00:17:30.880
how they work roughly to to really
get the most value from them, and

259
00:17:30.920 --> 00:17:33.559
again trying to trick it into revealing
that it's really sent in is probably not

260
00:17:33.599 --> 00:17:37.079
a very good use case, at
least not in the business world. If

261
00:17:37.160 --> 00:17:47.200
maybe in the end, back you
were listening to inside analysis, what if

262
00:17:47.240 --> 00:17:49.759
you could own a piece of the
future. What if you could build your

263
00:17:49.799 --> 00:17:55.519
next castle not on sand, but
on the bedrock of a modern blockchain ecosystem.

264
00:17:55.960 --> 00:18:00.920
The first Internet gold rush made millionaires, The second wave is mine billionaires,

265
00:18:00.359 --> 00:18:04.440
but the third wave is just gathering
now and anyone can get in on

266
00:18:04.480 --> 00:18:10.279
the action. Hop online to crowdpointtech
dot com to learn how you can secure

267
00:18:10.279 --> 00:18:15.240
a foothold in the blockchain revolution.
Whatever your passion, wherever you want to

268
00:18:15.279 --> 00:18:19.359
go in life, there's an opportunity
awaiting you right now. Go to crowdpointtech

269
00:18:19.440 --> 00:18:25.400
dot com to learn how the blockchain
will fuel the next generation of innovation in

270
00:18:25.440 --> 00:18:30.079
this globally connected world. That's crowd
pointech dot com, your trusted agent in

271
00:18:30.160 --> 00:18:36.039
an untrusted world. What's the longest
running radio show in the world focused on

272
00:18:36.119 --> 00:18:41.000
data DM Radio? Want to be
a guest sometime Send an email to info

273
00:18:41.039 --> 00:18:45.359
at dmradio dot biz. That's Info
at dmradio dot biz. What if you

274
00:18:45.400 --> 00:18:49.079
could own a piece of the future. What if you could build your next

275
00:18:49.119 --> 00:18:53.920
castle not on sand, but on
the bedrock and modern blockchain ecosystem. The

276
00:18:55.039 --> 00:19:00.319
first Internet gold rush made millionaires.
The second wave is minting billionaires, but

277
00:19:00.440 --> 00:19:03.839
the third wave is just gathering now
and anyone can get in on the action.

278
00:19:04.440 --> 00:19:08.720
Hop online to crowdpointtech dot com to
learn how you can secure a foothold

279
00:19:08.759 --> 00:19:12.720
in the Whatever your passion, wherever
you want to go in life, there's

280
00:19:12.720 --> 00:19:18.960
an opportunity awaiting you right now.
Go to crowdpointtech dot com to learn how

281
00:19:18.000 --> 00:19:23.119
the blockchain will fuel the next generation
of innovation in this globally connected world.

282
00:19:23.640 --> 00:19:30.839
That's crowdpointtech dot com. Your trusted
agent in an untrusted world. What's the

283
00:19:30.920 --> 00:19:34.200
longest running radio show in the world
focused on data? DM Radio. Want

284
00:19:34.200 --> 00:19:38.960
to be a guest sometime? Send
an email to Info at dmradio dot biz.

285
00:19:40.079 --> 00:19:45.160
That's Info at DM radio dot biz. Do you need to get your

286
00:19:45.200 --> 00:19:49.240
hands on some extra money right now? Maybe twenty five thousand or more.

287
00:19:49.400 --> 00:19:52.599
If you're a homeowner, now is
a perfect time to get cash out.

288
00:19:52.680 --> 00:19:56.640
While homes in many neighborhoods like yours
have gone up in money for anything,

289
00:19:56.759 --> 00:20:00.480
it's yours. You can buy an
investment property, payoff higher interest debt,

290
00:20:00.680 --> 00:20:06.240
or make home improvements if you need
twenty five thousand, fifty thousand or more.

291
00:20:06.559 --> 00:20:10.240
Now is the time home values are
up and so is your equity.

292
00:20:10.319 --> 00:20:12.319
We offer you a way to use
it. No need to use your savings

293
00:20:12.440 --> 00:20:15.960
called New American Funding. Now and
see how much cash out you can get.

294
00:20:17.119 --> 00:20:21.720
Call eight hundred seven one h three
seven three nine, eight hundred seven

295
00:20:21.839 --> 00:20:26.000
one h three seven three nine,
eight hundred seven one h three seven three

296
00:20:26.119 --> 00:20:30.880
nine. That's eight hundred seven one
h thirty seven thirty nine. NMLS sixty

297
00:20:30.920 --> 00:20:36.079
six oh six Www dot MLS,
Consumer XS dot org. This is not

298
00:20:36.119 --> 00:20:38.720
an offer or commitment to end subject
to borrow or improperty qualifications. Not all

299
00:20:38.759 --> 00:20:45.799
borrowers will qualify. Terms and conditions
apply equal housing opportunity. When a player's

300
00:20:45.799 --> 00:20:48.920
sudden cardiac event brought a national football
game to a halt, it's shown a

301
00:20:49.000 --> 00:20:53.720
spotlight on the importance of CPR readiness. Now, with youth sports in full

302
00:20:53.759 --> 00:20:59.319
swing, the American Heart Association is
rallying parents and coaches to be ready in

303
00:20:59.480 --> 00:21:03.440
emergence. To be ready learn hands
only CPR. It's a skill anyone can

304
00:21:03.519 --> 00:21:08.680
learn in minutes. Just visit haart
dot org slash hands only CPR and nationally

305
00:21:08.720 --> 00:21:15.039
supported by Elevant's Health Foundation. Now
you can fly anywhere in the world and

306
00:21:15.160 --> 00:21:18.759
pay discount prizes on your airline tickets. Book a flight today to London,

307
00:21:18.880 --> 00:21:23.480
Paris, Madrid or anywhere else you
want to go and pay a lot less

308
00:21:23.599 --> 00:21:30.279
guarantee. Call the International Travel Department
right now at Local eight hundred two nine

309
00:21:30.319 --> 00:21:34.160
eight five seven eight three eight hundred
two nine eight five seven eight three.

310
00:21:34.200 --> 00:21:44.000
That's eight hundred two nine eight fifty
seven eighty three. Welcome back to Inside

311
00:21:44.039 --> 00:21:52.920
Analysis. Here's your host, Eric
Tavanaugh. All right, folks, back

312
00:21:52.920 --> 00:21:56.279
here on Inside Analysis. At fascinating
conversation here with Mike Finley, co founder

313
00:21:56.279 --> 00:22:02.279
and CTO of Answer Rocket. We're
getting some great answers about these large language

314
00:22:02.319 --> 00:22:07.759
models and how to fuse them with
your traditional business intelligence environment. You know,

315
00:22:07.079 --> 00:22:11.240
always we talk about these things.
Don't throw the baby out the bathwater

316
00:22:11.319 --> 00:22:12.759
or no, this is not going
to solve all your problems. It gives

317
00:22:12.759 --> 00:22:21.279
you a very powerful context onto and
infuse into your traditional business intelligence or analytics

318
00:22:21.680 --> 00:22:23.480
environment. So, Mike, I'd
like to ask you just to kind of

319
00:22:23.480 --> 00:22:29.640
walk us through, especially this instance
conversation, how you go into maybe as

320
00:22:29.680 --> 00:22:32.799
your some cloud platforms, saying this
is what I want, kind of walk

321
00:22:32.839 --> 00:22:36.759
us through how that all happens and
what it means. Absolutely. Yeah,

322
00:22:36.799 --> 00:22:42.599
So look exactly what does a GPT
or a palm or a barred look like

323
00:22:42.759 --> 00:22:45.599
if you if you walked up to
it with the physical thing, what would

324
00:22:45.599 --> 00:22:48.759
it be. It's a bunch of
computers. It's a bunch of servers,

325
00:22:48.839 --> 00:22:52.279
right, that's it. Uh,
And they have storage and they have whatever.

326
00:22:52.319 --> 00:22:55.480
But it's basically about a bunch of
computers. So if you want to

327
00:22:56.279 --> 00:22:59.839
use the same instance of GPT that
all the teenagers are using to write their

328
00:23:00.200 --> 00:23:02.920
papers, right, then you go
to the open AYE website and go to

329
00:23:03.000 --> 00:23:07.279
chat GBT and you're hitting the same
computer that all those other folks are hitting.

330
00:23:07.559 --> 00:23:10.119
Um. And when you do that, it comes with some warnings.

331
00:23:10.200 --> 00:23:11.839
Right, anything you send to it
is going to be used by it to

332
00:23:11.960 --> 00:23:15.759
learn how to do things better.
That's how it got as good as it

333
00:23:15.839 --> 00:23:18.880
is, right. But if you
if you don't want that to happen,

334
00:23:18.960 --> 00:23:22.160
if you want the benefit of all
that learning that that thing did, but

335
00:23:22.319 --> 00:23:26.519
you want your data to be kept
secret on the way, to be used

336
00:23:26.559 --> 00:23:30.160
only for you, um, for
your benefit in the future, and never

337
00:23:30.200 --> 00:23:33.839
revealed to anybody else. Well,
then all you need is another copy of

338
00:23:33.839 --> 00:23:38.839
that big server UM. And that's
exactly what what Microsoft offers through Azure.

339
00:23:40.000 --> 00:23:42.240
Same same as if you had a
database or a compute server. Now they're

340
00:23:42.279 --> 00:23:47.400
offering they're offering instances, private instances
of the language models, and uh,

341
00:23:47.480 --> 00:23:52.000
it's as easy as adding one of
those resources through an open an Azure account,

342
00:23:52.400 --> 00:23:56.559
create that model added and begin using
it. And at that point it's

343
00:23:56.960 --> 00:24:00.079
it's going to it's going to be
a transfer of all the knowledge in the

344
00:24:00.160 --> 00:24:04.640
publicly available model that's out there.
Latest one, I think is June thirteen

345
00:24:04.759 --> 00:24:08.519
is the last kind of finished to
published a GPT three five turbo, and

346
00:24:08.559 --> 00:24:12.240
then it'll be used for your benefit
and again it won't be shared with anybody

347
00:24:12.240 --> 00:24:18.039
else, with all of the same
credentials and the same assurances that come with

348
00:24:18.400 --> 00:24:22.839
saving your data into an Azure database
or using an Azure compute resource to process

349
00:24:22.960 --> 00:24:26.880
your your web server. So it's
it's as secure as all the other cloud

350
00:24:26.920 --> 00:24:33.079
services that you're already using. Yeah, and then so once you get that

351
00:24:33.160 --> 00:24:37.519
instantiated, you can start throwing your
information at it. Right. I joked

352
00:24:37.519 --> 00:24:42.759
about my friend who just dumped dumps
these two hundred page technical briefings into instance

353
00:24:42.799 --> 00:24:47.599
and he gets back a summary.
So from a client perspective, would your

354
00:24:47.599 --> 00:24:51.799
clients then start loading their reports into
this thing? Would you connect the data

355
00:24:51.839 --> 00:24:55.799
warehouse to it? How does that
work? Yeah? Yeah, So once

356
00:24:55.799 --> 00:24:59.640
it's up and running, there's a
there's not a way to dump your data

357
00:24:59.680 --> 00:25:03.400
into it. It's still very much
like the thing that you see that everybody

358
00:25:03.480 --> 00:25:06.839
uses online, the chat GPT.
You you give it a prompt and it

359
00:25:06.839 --> 00:25:10.359
gives you a response. Right.
So the key then is to have a

360
00:25:10.359 --> 00:25:14.079
piece of software um that essentially is
acting as the front end. That's that's

361
00:25:14.119 --> 00:25:18.759
bridging the semantic gap between the business
use and that back end. So let

362
00:25:18.799 --> 00:25:22.359
me give you some examples. Um, if you let's say that your your

363
00:25:22.680 --> 00:25:25.720
your your users are gonna be talking
about some obscure part numbers. You know

364
00:25:25.720 --> 00:25:29.200
they're they're gonna they're gonna look like
numbers. So if if you just go

365
00:25:29.240 --> 00:25:32.640
to the public GPT instance in reference
number six two five nine, five,

366
00:25:32.680 --> 00:25:36.079
eight or whatever. Um, that
public instance is going to think that's a

367
00:25:36.160 --> 00:25:38.079
number. It's not going to know
anything about it. If if there's a

368
00:25:38.119 --> 00:25:42.759
semantic layer between your business user and
the language model, then that layer can

369
00:25:42.799 --> 00:25:47.640
augment that number with something indicating to
the language model. By the way,

370
00:25:48.200 --> 00:25:51.720
this is a part number, so
that when the language model sees it,

371
00:25:51.559 --> 00:25:53.079
it knows, oh, that's a
part number. Now now I know how

372
00:25:53.119 --> 00:25:56.960
to talk about that thing as a
part number, not as not as a

373
00:25:56.079 --> 00:26:00.319
number. Right. Um, this
happens with zip codes, right, happens

374
00:26:00.359 --> 00:26:06.519
with um you know anything like excuse
me, like quantities? Is it a

375
00:26:06.519 --> 00:26:08.359
percent or is it a dollar?
Value? In business? That's really important

376
00:26:08.920 --> 00:26:15.480
in the in the largely numbers with
suggested formats right in the in the business

377
00:26:15.480 --> 00:26:18.839
world, we need for the language
model to treat it very precisely like what

378
00:26:18.880 --> 00:26:22.599
it's intended to be and not sort
of run away and imagine a different future

379
00:26:22.640 --> 00:26:26.200
for that number. Makes sense?
Yeah, And so what you're doing then

380
00:26:26.240 --> 00:26:30.680
when you talk about this bridge to
the semantic layer basically is answer rocketing.

381
00:26:30.799 --> 00:26:36.799
Your instance is going to be the
front end, and it's reaching into chat

382
00:26:36.920 --> 00:26:42.519
GPT to generate context around specific discussions
and topics and things, supply chain issues,

383
00:26:42.599 --> 00:26:47.039
or whatever the case may be.
So basically it's it's now augmenting.

384
00:26:47.079 --> 00:26:52.160
It's part of your information management program, and it is a text generator,

385
00:26:52.200 --> 00:26:56.759
context generator really designed to facilver already
doing. Is that right? That's right,

386
00:26:56.759 --> 00:27:00.440
that's right. You can you can
very much think of it as it's

387
00:27:00.480 --> 00:27:03.680
like the concierge. Right. So
instead of instead of going in and saying,

388
00:27:04.000 --> 00:27:07.880
hey, business user, here is
Excel. Please get your answers right,

389
00:27:08.200 --> 00:27:11.160
imagine you had you had a concierge
who would say to the business user,

390
00:27:11.240 --> 00:27:14.880
what do you need to know about
this spreadsheet? Um? And then

391
00:27:14.920 --> 00:27:18.000
the user would say, I need
to know where the you know where the

392
00:27:18.039 --> 00:27:22.400
problem areas are. I'm maybe I'm
managing a bunch of franchises and I need

393
00:27:22.440 --> 00:27:26.559
to know today where should I go? Right? So you could you could

394
00:27:26.599 --> 00:27:29.000
do that in a spreadsheet by saying, give me the most recent day,

395
00:27:29.079 --> 00:27:33.240
rank those by the labor ratio or
whatever, and that would tell you where

396
00:27:33.240 --> 00:27:34.559
you need to go today. Or
you could say to the concierge, please

397
00:27:34.559 --> 00:27:38.000
look this up for me. GPT
plays that that that role in answer rocket,

398
00:27:38.079 --> 00:27:41.440
Right, So you've got all your
data in the system. The user

399
00:27:41.519 --> 00:27:45.240
is very much in a conversation with
GPT, but GPT is getting all the

400
00:27:45.319 --> 00:27:49.480
facts fed into it from from answer
Rocket. So if the user says,

401
00:27:49.519 --> 00:27:53.240
hey, who's in trouble, GPT
has no idea who's in trouble, but

402
00:27:53.240 --> 00:28:00.039
it knows that that means find me
the bottom stores by labor ratio and shrinkage,

403
00:28:00.119 --> 00:28:03.039
right, or whatever those those key
parameters are. Because it's AI is

404
00:28:03.119 --> 00:28:07.519
goal driven, so you give it
those goals and then answer rocket will find

405
00:28:07.000 --> 00:28:10.559
the way or the products that are
ranked a certain way, or whatever the

406
00:28:11.000 --> 00:28:15.000
thing is the users looking for,
and then the GPT will communicate that back

407
00:28:15.039 --> 00:28:18.160
to the user. Hey, um, you said five, but really two

408
00:28:18.240 --> 00:28:21.440
will really focus on these two the
other three or these three, but they're

409
00:28:21.519 --> 00:28:25.039
much better off, so don't worry
about them today. Right, And this

410
00:28:23.880 --> 00:28:27.640
is how the music is made,
right. It's it's really a magical experience

411
00:28:27.680 --> 00:28:30.240
because in the past, the user
would have just gotten the dump of all

412
00:28:30.279 --> 00:28:33.480
this information. Um, you know, they'd be looking at a bar chart

413
00:28:33.519 --> 00:28:37.200
that they're going to have to interpret. Instead, they can just get some

414
00:28:37.400 --> 00:28:42.400
very straightforward kinds of instructions and conversation
and they and they can they can continue

415
00:28:42.640 --> 00:28:47.519
that line of thinking and questioning and
drilling down and getting more well, I

416
00:28:47.559 --> 00:28:52.160
guess one question I have is in
the typical setup, what is the actual

417
00:28:52.359 --> 00:28:55.200
front end, because like what chet
GPT, you're in this environment, you've

418
00:28:55.200 --> 00:28:57.200
got to prompt you throw things at
it, etc. Is that still how

419
00:28:57.240 --> 00:29:03.200
you would do it ideally with primarily
in the answer rocket environment. And then

420
00:29:03.640 --> 00:29:06.839
the the chat GPT is kind of
the background. It is. The chat

421
00:29:06.880 --> 00:29:08.119
GPT is definitely in the background,
and the reason for that is because it

422
00:29:08.160 --> 00:29:11.920
needs to be under security um and
needs to it needs to have you know,

423
00:29:12.359 --> 00:29:17.599
this user is authenticated for what data
they have credentials to access? Right,

424
00:29:17.720 --> 00:29:21.799
So so it's really you know,
it's a it is a business intelligence

425
00:29:21.880 --> 00:29:26.000
enterprise tool, right. It comes
with all the sophistication of role management security,

426
00:29:26.400 --> 00:29:30.880
you know, row level isolation of
that data, um. All those

427
00:29:30.920 --> 00:29:36.240
things are very important even through this
chat conversation. So the experience though,

428
00:29:36.279 --> 00:29:38.720
the experience, the actual visual experience
the user has is very much like a

429
00:29:38.799 --> 00:29:44.200
chat GPT, augmented with some charts
and some other drop downs and click clicky

430
00:29:44.319 --> 00:29:47.480
type things that you would see you
know in a in a dashboard or something

431
00:29:47.519 --> 00:29:51.440
like that. But but it feels
very much like a chat conversation UM and

432
00:29:51.440 --> 00:29:55.039
that that helps obviously if you're if
you're on a you know, if you're

433
00:29:55.039 --> 00:29:57.640
on a desktop. It helps the
user to feel like that's a live,

434
00:29:57.680 --> 00:30:02.680
interactive chat. But it also supports
all sorts of other paradigms that you might

435
00:30:02.759 --> 00:30:06.599
use, like a like a cell
phone based or even embedding this capability in

436
00:30:06.599 --> 00:30:10.880
a third party API so that it
can be surfaced inside you know, Microsoft

437
00:30:10.880 --> 00:30:12.799
Teams or some other some other chat
a slack, you know, some other

438
00:30:14.119 --> 00:30:18.799
chat type interface. That's interesting,
So would it makes sense to feed in

439
00:30:19.119 --> 00:30:23.079
reports like your annual reports, let's
say, or your corarely reports. Would

440
00:30:23.079 --> 00:30:27.720
it makes sense to feed that into
your instance of chat NPT and then over

441
00:30:27.839 --> 00:30:33.720
time? Because what one thing I've
noticed that's really quite interesting is these large

442
00:30:33.759 --> 00:30:40.480
language models, these foundational models will
sort of seemingly intuitively understand certain constructs,

443
00:30:40.599 --> 00:30:44.920
even market dynamics and things of this
nature. So you know, you could

444
00:30:44.960 --> 00:30:48.680
look at it and say, hey, I have to lay off ten percent

445
00:30:48.799 --> 00:30:52.160
of the people in this company because
things do not look a good. Where

446
00:30:52.200 --> 00:30:56.440
should I look to first do that? And it can come back and give

447
00:30:56.440 --> 00:31:00.359
you some rationale for cutting back here
cutting back the area can be very specific

448
00:31:00.400 --> 00:31:04.240
in those recommendations. Right, yeah, absolutely so. UM, so the

449
00:31:04.960 --> 00:31:10.240
UM you are correct. But let's
talk about the way that you said it.

450
00:31:10.279 --> 00:31:12.240
So you said, can I feed
in my annual report to chat GPT.

451
00:31:12.680 --> 00:31:17.000
The short version of that is,
UM, you can't feed it in

452
00:31:17.000 --> 00:31:22.599
in the sense of saying, can
I make the GPT instance change because I've

453
00:31:22.640 --> 00:31:26.200
given it this report, so it
permanently knows that information that that's not a

454
00:31:26.240 --> 00:31:29.559
thing yet. All right, So
that's that's something that may be coming in

455
00:31:29.559 --> 00:31:32.599
the future. UM. We hear
a lot of things. But what you

456
00:31:32.599 --> 00:31:34.160
can do, you can do.
What you can do is say, okay,

457
00:31:34.400 --> 00:31:37.960
now the user has just asked me
a question about layoffs. What documents

458
00:31:37.960 --> 00:31:42.400
do I have that um, that
talk about layoffs? Um, from the

459
00:31:42.400 --> 00:31:45.599
past. Oh, it looks like
I've got this annual report from from five

460
00:31:45.680 --> 00:31:49.519
years ago when we went through a
down sort downturn that talks about layoffs.

461
00:31:49.799 --> 00:31:55.480
Now, let me augment what GPT
has right now for this question. Let

462
00:31:55.519 --> 00:32:00.519
me augment what it knows by giving
it that document, and then it will

463
00:32:00.519 --> 00:32:04.839
answer that question in the context of
that annual report that it has from the

464
00:32:04.839 --> 00:32:07.440
past. Right, So, it's
this is the discipline of prompt engineering.

465
00:32:07.440 --> 00:32:10.839
You may have heard that term.
That's very much about It's very much about

466
00:32:10.839 --> 00:32:15.920
saying, what all do I need
to put into that prompt so that the

467
00:32:15.000 --> 00:32:20.440
language model has what it needs to
answer, but always keeping in mind I

468
00:32:20.599 --> 00:32:23.319
have to be able to fact check
it. I can't I can't leave that.

469
00:32:23.640 --> 00:32:29.319
I can't have the question be completely
open ended because it will hallucinate things

470
00:32:29.519 --> 00:32:32.519
above all else, it wants to
answer, so it will make up an

471
00:32:32.519 --> 00:32:35.960
answer. If you say, you
know how many McDonald's are on the dark

472
00:32:35.960 --> 00:32:38.079
side of the moon, it's going
to give you an answer. My favorite,

473
00:32:38.720 --> 00:32:43.200
my favorite test for language models when
we're playing around with them is I'll

474
00:32:43.200 --> 00:32:46.359
say Hakuna and the answer always comes
back matadha, and it wants to have

475
00:32:46.359 --> 00:32:51.160
a celebration about that. It really
wants to fill in a blank. It

476
00:32:51.240 --> 00:32:53.279
will not leave Yeah, it will
not leave a blank. So so you

477
00:32:53.279 --> 00:32:55.599
have to be careful when you ask
it a question. You're going to get

478
00:32:55.640 --> 00:32:59.160
an answer. If you want that
to be based on fact, you have

479
00:32:59.240 --> 00:33:02.039
to number one and ensure that you
fed it those facts, and number two

480
00:33:02.160 --> 00:33:07.079
verify that whatever it extracted is still
true or holds true. Or is accurate.

481
00:33:08.279 --> 00:33:12.119
You're reminding me of a lesson I
learned when we went to Europe as

482
00:33:12.559 --> 00:33:16.200
teenagers, and in certain countries they
would tell you don't ask for directions from

483
00:33:16.240 --> 00:33:20.119
the locals because they'll be embarrassed that
they can't answer, and so they'll just

484
00:33:20.119 --> 00:33:22.440
give you some answer. And that's
kind of like what check If you do

485
00:33:22.759 --> 00:33:25.440
right, they're gonna be hereing what
ask they ask it for? Right,

486
00:33:25.559 --> 00:33:31.480
bild to verified and that that gets
back to traditional business intelligence data warehousing and

487
00:33:32.200 --> 00:33:37.440
similar practices, because that's where you
verify, Okay, is it really true

488
00:33:37.519 --> 00:33:40.039
that we had this many sales and
you go check the warehouse and check and

489
00:33:40.079 --> 00:33:43.359
balance. You have to watch out
before, right, Yeah, validation and

490
00:33:43.440 --> 00:33:45.720
audits absolutely yeah. And so we
we build that into the kind of the

491
00:33:45.759 --> 00:33:52.880
real time aspect of what the language
model is processing and is returning back.

492
00:33:52.960 --> 00:33:57.720
And we also put in the guardrails
that say, look, don't answer if

493
00:33:57.720 --> 00:34:01.240
you don't have this information and so, and that's in the answer rocket layer.

494
00:34:01.359 --> 00:34:05.640
Is that right? So basically it's
it's a layer of instruction. You've

495
00:34:05.680 --> 00:34:09.239
built in some guardrails that say,
Okay, no matter what the question is,

496
00:34:09.280 --> 00:34:13.159
if you don't know the answer,
don't come up with something. Yeah,

497
00:34:13.719 --> 00:34:15.480
I mean, the language model thinks
it's taken the sat you know,

498
00:34:15.679 --> 00:34:21.119
it literally thinks I must answer every
question and because it's it's better to take

499
00:34:21.119 --> 00:34:22.480
a guess than not to right.
So so you have to be you have

500
00:34:22.519 --> 00:34:24.599
to be careful. It doesn't know. You have to tell it. Hey,

501
00:34:24.599 --> 00:34:28.119
the right answer if you don't know, the right answer is I don't

502
00:34:28.119 --> 00:34:30.239
know. And so these are these
are the very subtle. You know,

503
00:34:30.280 --> 00:34:32.920
one of the one of the best
engineers I know the other day said,

504
00:34:32.960 --> 00:34:37.000
look, let's stop calling it prompt
engineering and let's call it what it is,

505
00:34:37.119 --> 00:34:40.079
rhetoric, right, Like a lot
of the instructions, Yeah, a

506
00:34:40.079 --> 00:34:45.119
lot of the instructions to the prompt
are really just getting the rhetoric correct,

507
00:34:45.199 --> 00:34:47.440
right, getting the rhetoric that because
at the end of the day, it

508
00:34:47.559 --> 00:34:51.840
is a language model. That's what
it understands language. And so it'll do

509
00:34:51.840 --> 00:34:54.320
what you tell it. Um,
it will generally not do what you don't

510
00:34:54.360 --> 00:34:59.440
tell it. But you have to
verify. You have to verify that before

511
00:34:59.480 --> 00:35:01.960
you you a business user to trust
it well. And you know, so

512
00:35:02.119 --> 00:35:07.679
let's just kind of talk about rubber
meets road with analytics and business right in

513
00:35:07.800 --> 00:35:09.519
the day. I mean, I
get all excited about technologies because I'm in

514
00:35:09.559 --> 00:35:13.159
the space and I'm a geek.
But at the end of the day,

515
00:35:13.199 --> 00:35:16.719
the whole point of all these exercises
is to run your business better, and

516
00:35:16.760 --> 00:35:21.000
a big part of that is the
storytelling. Is to get the data and

517
00:35:21.039 --> 00:35:25.039
then explain to the board of directors, your boss, a partner, whoever,

518
00:35:25.159 --> 00:35:30.880
explain to some other party what's happening
and why you want to take X,

519
00:35:30.039 --> 00:35:34.480
Y or Z decision. And that's
also where it can be very helpful

520
00:35:34.519 --> 00:35:37.119
because you can say, I have
to present to the board of directors tonight,

521
00:35:37.400 --> 00:35:40.840
what are the five things I need
to be sure to say, and

522
00:35:40.960 --> 00:35:45.760
please be sure to focus on the
treasurer who always asks the hardest questions,

523
00:35:45.760 --> 00:35:47.840
and that it will give you a
whole bunch of stuff to work with.

524
00:35:49.440 --> 00:35:52.440
You go through that and choose kind
of cherry pick what makes sense for you

525
00:35:52.559 --> 00:35:57.800
right absolutely. You know it's funny. I my undergraduate was in physics and

526
00:35:57.840 --> 00:36:00.239
I went into business. And one
of the reasons why is because I think

527
00:36:00.280 --> 00:36:04.599
business is the hardest problem there is. And the reason the reason is the

528
00:36:04.599 --> 00:36:07.599
hardest problem there is. It's very
straightforward as soon as you know the right

529
00:36:07.639 --> 00:36:09.519
answer, somebody else copies you and
runs away with it. Right, That's

530
00:36:09.599 --> 00:36:14.079
that's how competition works. So so
you have to get to get it right

531
00:36:14.119 --> 00:36:19.159
again and again and again. Well
that's why these language models, basically intelligence

532
00:36:19.239 --> 00:36:23.599
on tap is the ultimate power tool
for business. Right And and they're they're

533
00:36:23.719 --> 00:36:27.920
very new now, like literally,
you know, six months old. But

534
00:36:28.000 --> 00:36:31.719
imagine if after six months of its
existence, it was unreliable, it was

535
00:36:31.760 --> 00:36:35.719
expensive, it was I still remember
you know, I was in the cash

536
00:36:35.760 --> 00:36:39.480
register business back then, and a
chain of pizzeria's own name, the CTO

537
00:36:39.639 --> 00:36:44.119
said I will put the cash register
in the cloud when you put the pizza

538
00:36:44.159 --> 00:36:46.880
oven in the cloud, right,
uh, and and so you know that

539
00:36:46.880 --> 00:36:50.519
that kind of fear exists. But
now you know, look at Square and

540
00:36:50.559 --> 00:36:52.599
all these other ones that they're very
much cloud based and and so that's where

541
00:36:52.639 --> 00:36:55.760
these language models are going. There, they are they're going towards that that

542
00:36:55.880 --> 00:37:00.480
point. Yeah, so yeah,
this is this is fast that extel folks

543
00:37:00.519 --> 00:37:14.000
don't touch up. That will be
right back. You're listening to Inside Analysis

544
00:37:19.679 --> 00:37:23.000
Ray stand up to the next financial
crisis. That our top economists are saying

545
00:37:23.079 --> 00:37:29.159
is that our doorsteps By allocating a
percentage of your IRA into physical gold and

546
00:37:29.239 --> 00:37:34.119
silver with a tax free rollover,
you can diverse the holdings from turbulent markets

547
00:37:34.119 --> 00:37:38.719
and economic downturns by putting your IRA. Find out how to safeguard your assets

548
00:37:38.719 --> 00:37:44.760
with a tax free rollover with a
Genesis Gold Ira, the only IRA that

549
00:37:44.840 --> 00:37:49.559
can hold physical precious metals. Call
now for your free gold and silver report.

550
00:37:49.760 --> 00:37:53.679
Protect your IRA today with one simple
phone call and learn how to quantify

551
00:37:53.880 --> 00:37:59.880
for up to ten thousand dollars in
free silver called Genesis Gold Group. Empowering

552
00:38:00.199 --> 00:38:05.960
faith driven Stewardship. Eight hundred six
four four eight six one one eight hundred

553
00:38:06.119 --> 00:38:10.920
six four four eight six one one
eight hundred six four four eight six one

554
00:38:12.039 --> 00:38:16.159
one. That's eight hundred six four
four eighty six eleven. Can your IRA

555
00:38:16.440 --> 00:38:21.719
stand up to the next financial crisis
that our top economists are saying is at

556
00:38:21.719 --> 00:38:27.000
our doorsteps. By allocating a percentage
of your IRA into physical gold and silver

557
00:38:27.199 --> 00:38:31.360
with a tax free rollover, you
can diversify in safeguard your holdings from turbulent

558
00:38:31.400 --> 00:38:37.000
markets and economic downturns by putting your
IRA back on the gold standard. Find

559
00:38:37.039 --> 00:38:42.840
out how to safeguard your sts with
a tax free rollover with a Genesis Gold

560
00:38:42.960 --> 00:38:46.760
IRA, the only IRA that can
hold physical precious metals. Call now for

561
00:38:46.840 --> 00:38:52.360
your free gold and silver report.
Protect your IRA today with one simple phone

562
00:38:52.360 --> 00:38:57.800
call and learn how to qualify for
up to ten thousand dollars in free silver

563
00:38:58.079 --> 00:39:02.360
called Genesis Gold Group empower Ring Faith
Driven Stewardship. Eight hundred and six four

564
00:39:02.519 --> 00:39:08.360
four eight six one one eight hundred
six four four eight six one one eight

565
00:39:08.440 --> 00:39:14.199
hundred and six four four eight six
one one. That's eight hundred six four

566
00:39:14.360 --> 00:39:19.440
four eighty. Do you own a
timeshare? We'll face the facts. You

567
00:39:19.519 --> 00:39:22.639
made a mistake, You made a
bad purchase. A timeshare is not an

568
00:39:22.639 --> 00:39:28.119
investment. It's a money pit that
continues forever. If you use your time

569
00:39:28.159 --> 00:39:31.679
share, that's great. But if
you don't and you want illegally get out

570
00:39:31.719 --> 00:39:37.360
of your contract, call my friends
right now at the Timeshare Exit Hotline.

571
00:39:37.519 --> 00:39:40.800
They're an experienced team of lawyers who
help good people like you get out of

572
00:39:40.800 --> 00:39:45.360
a timeshare contract that they just don't
want. Don't throw away your money on

573
00:39:45.519 --> 00:39:50.159
maintenance fees, use it for things
you really want. We can help you

574
00:39:50.360 --> 00:39:53.920
end your time share contract and stop
the money drain. Immediately. If you

575
00:39:53.960 --> 00:39:59.119
are ready to move on with your
time share, call our team right now.

576
00:39:59.360 --> 00:40:02.639
Yes your timeshare now with a free
call eight hundred two eight nine O

577
00:40:02.960 --> 00:40:08.280
four one three O four one three, eight hundred two or one three.

578
00:40:08.320 --> 00:40:15.400
That's eight hundred two eight nine zero
four thirteen. When a player's sudden cardiac

579
00:40:15.400 --> 00:40:20.360
event brought a national football game to
a halt, it's shown a spotlight on

580
00:40:20.400 --> 00:40:25.199
the importance of CPR readiness. Now
with youth American Heart Association is rallying parents

581
00:40:25.199 --> 00:40:30.480
and coaches to be ready in an
emergency. To be ready, learn hands

582
00:40:30.480 --> 00:40:34.360
only CPR. It's a skill anyone
can learn in minutes dot org, slash

583
00:40:34.400 --> 00:40:38.880
hands only CPR. Hands only CPR
is nationally supported by an Elevant's Health Foundation.

584
00:40:39.960 --> 00:40:45.440
Now you can fly anywhere in the
world and pay discount prices on your

585
00:40:45.480 --> 00:40:49.440
airline tickets. Book a flight today
to London, Paris or else you want

586
00:40:49.440 --> 00:40:54.000
to go and pay a lot less
guarantee. Call the International Travel Department right

587
00:40:54.039 --> 00:41:00.920
now at low cost airlines eight hundred
two nine eight five seven eight eight hundred

588
00:41:00.000 --> 00:41:05.360
drew nine eight five seven eight three. That's eight hundred two nine eight fifty

589
00:41:05.400 --> 00:41:15.039
seven eighty three Express Welcome back to
Inside Analysis. Here's your host, Eric

590
00:41:15.199 --> 00:42:39.599
Kavanaugh. Hard folks, back here
on Inside Analysis. They fascinating conference from

591
00:42:39.639 --> 00:42:45.039
Answer Rocket. We're talking all about
foundational models, large language models. What

592
00:42:45.079 --> 00:42:47.119
do they mean? How do you
use them? Well, there are lots

593
00:42:47.119 --> 00:42:50.400
of different ways, folks, I
mean you can you can't even count the

594
00:42:50.440 --> 00:42:52.280
number of ways that you can use
these things. But they're very good for,

595
00:42:52.360 --> 00:42:55.239
of course, generating That's what they're
supposed to do, is generate text.

596
00:42:55.519 --> 00:43:00.400
There are other models that are used
for generating graphics. Mid journey.

597
00:43:00.440 --> 00:43:05.719
I think it's one that's through the
discord channel is just shockingly amazing. But

598
00:43:05.800 --> 00:43:07.800
we'll focus on the text. And
we just took a quick break there and

599
00:43:07.840 --> 00:43:12.679
the break might get a really good
point talking about dark data. So what

600
00:43:12.920 --> 00:43:16.039
is dark data? Well, large
enterprises have had massive amounts of data.

601
00:43:16.119 --> 00:43:20.000
Most of it is unstructured. What
do they say, ten percent, maybe

602
00:43:20.079 --> 00:43:22.920
fifteen percent is structured, by which
we mean sitting in a relational databases,

603
00:43:23.400 --> 00:43:28.440
structured in a way to facilitate reporting
and access and so forth. But all

604
00:43:28.480 --> 00:43:32.440
the emails, all the PDFs,
all the PowerPoint presentations, there is a

605
00:43:32.519 --> 00:43:37.960
tremendous amount of knowledge based into these
things, and people who are old like

606
00:43:38.039 --> 00:43:43.239
me remember something called knowledge management and
that was going to solve all our problems

607
00:43:43.280 --> 00:43:46.079
like twenty five years ago, and
it really didn't get too far in part

608
00:43:46.119 --> 00:43:50.760
because we didn't have the compute,
we didn't have the semantic models. Yet

609
00:43:51.000 --> 00:43:54.000
now we have all that stuff and
more. We have these engines and that's

610
00:43:54.000 --> 00:44:00.519
what they are. They are engines
that can process information. So what do

611
00:44:00.519 --> 00:44:02.760
you think about the future of dark
data in the enterprise and why does that

612
00:44:02.800 --> 00:44:07.559
matter money? Oh? Absolutely,
yeah, So look, dark data is

613
00:44:07.599 --> 00:44:09.840
the future, is what I would
say. Dark data is becoming the becoming

614
00:44:09.840 --> 00:44:15.119
the light data. The fact is
we've thought of is the terminal state for

615
00:44:15.159 --> 00:44:19.920
information, Right Like once a human
reads a spreadsheet, draws some conclusions,

616
00:44:19.920 --> 00:44:22.119
writes it out, it goes in
a folder somewhere areage. That's all we

617
00:44:22.400 --> 00:44:25.679
ever get to do with it.
But now these language models are able to

618
00:44:27.320 --> 00:44:30.079
They're able to read that and make
use of it and answer questions from it

619
00:44:30.199 --> 00:44:35.360
if we can find the right information
to answer the question. And this is

620
00:44:35.440 --> 00:44:38.880
where the technology area. It's not
it's not well understood, but it is

621
00:44:38.920 --> 00:44:44.159
really important. Embeddings and vector databases, right, So, so these these

622
00:44:44.400 --> 00:44:50.639
embeddings and embeddings is just a a
sophisticated way of saying what the computer or

623
00:44:50.639 --> 00:44:53.920
what the model thinks about. So
it might be it's two o'clock in Tokyo.

624
00:44:54.039 --> 00:44:57.880
That's a statement. The model has
an embedding for that. It might

625
00:44:57.920 --> 00:45:01.239
be you know, last quarter sales
were down three percent and that's because,

626
00:45:01.639 --> 00:45:05.800
uh, you know, we had
a supply chain shortage in Asia. Whatever,

627
00:45:06.079 --> 00:45:07.800
whatever that statement, anything that you
can say that you can write a

628
00:45:07.840 --> 00:45:13.039
two hundred page document has an embedding. And that embedding is just a bunch

629
00:45:13.119 --> 00:45:15.760
of numbers, a bunch of numbers
that means something to the language model.

630
00:45:15.800 --> 00:45:22.239
So the way that the way that
dark data becomes light is you index it

631
00:45:22.280 --> 00:45:25.280
with embeddings, You feed it to
the language models UM, and you get

632
00:45:25.320 --> 00:45:29.400
from the language model, you get
back kind of an index. Right.

633
00:45:29.840 --> 00:45:32.159
And now anytime a question comes in, a new question comes in, Hey,

634
00:45:32.440 --> 00:45:37.119
um, when if we had supply
chain problems in um, you know

635
00:45:37.800 --> 00:45:39.960
India, Right, Well, it's
going to go match up against all the

636
00:45:39.960 --> 00:45:45.039
embeddings that it has, and it's
going to India in Asia ut supply chain

637
00:45:45.079 --> 00:45:49.039
shortages, transportation problems, it's going
to find, you know, anything related

638
00:45:49.079 --> 00:45:53.760
to causing shortages of inventory. Why
Because the language model has such a deep

639
00:45:53.880 --> 00:46:00.679
understanding that that it will associate any
problem with with UM, apply with supply

640
00:46:00.760 --> 00:46:06.000
chain management, any problem with transportation, with supply chain management, any kind

641
00:46:06.000 --> 00:46:09.199
of outage, it'll it'll group all
of that and and it really starts to

642
00:46:09.239 --> 00:46:14.800
show truly the power of what these
language models are doing when you look at

643
00:46:14.840 --> 00:46:17.440
how how similar the embeddings for two
things are, even though they might be

644
00:46:17.480 --> 00:46:22.320
in different languages, they might be
very different ways of saying the same thing,

645
00:46:22.360 --> 00:46:25.079
they might be only tangentially related,
but the embeddings will light up.

646
00:46:25.119 --> 00:46:29.719
The embeddings will be similar enough that
the language model will will tell you,

647
00:46:29.920 --> 00:46:34.440
hey, this is this is meaningful
content to answer that question. That's really

648
00:46:34.440 --> 00:46:37.920
where where the superpowers are. It's
almost like the like the Dewey decimal system

649
00:46:37.920 --> 00:46:42.440
where we get you numbers that the
library uses. Imagine there's fifteen hundred numbers,

650
00:46:42.679 --> 00:46:45.360
right, that's the index for that
content. And if you can take

651
00:46:45.480 --> 00:46:49.320
all of your enterprise documents and index
it like that, then you can access

652
00:46:49.360 --> 00:46:59.000
anything you need to access really quickly, get into this embeddings things. That's

653
00:46:59.079 --> 00:47:01.159
very interesting. So what are basically
saying is and I have read up on

654
00:47:01.199 --> 00:47:05.679
this that one of the things they
do in these large language models is that

655
00:47:05.840 --> 00:47:09.679
they provide a numeric value to every
statement that we sent it to paragraph all

656
00:47:09.719 --> 00:47:12.559
sorts of different things. Is that
kind of where you're going with this?

657
00:47:12.599 --> 00:47:15.599
And that works? That's right?
That that is that is the the the

658
00:47:15.679 --> 00:47:20.280
idea of embeddings. Look until until
language models, So like this time this

659
00:47:20.320 --> 00:47:22.519
time last year, right, language
models were kind of a cool thing.

660
00:47:23.000 --> 00:47:28.280
Peripherally, we were all focused on
machine learning, right and classifiers, right,

661
00:47:28.280 --> 00:47:31.920
and those classifiers they were really good
at saying which customers are like other

662
00:47:32.000 --> 00:47:36.239
customers, you know which, But
which products should I recommend to Eric based

663
00:47:36.280 --> 00:47:38.519
on the things that Eric already knows? Right? Uh, which are my

664
00:47:38.599 --> 00:47:43.519
stars and which are my you know, lagging solutions? Right? Um,

665
00:47:44.159 --> 00:47:45.800
that's as good as the language models
could do. They really not sorry,

666
00:47:45.880 --> 00:47:50.519
not language models. The machine learning
models could do. They understood the idea

667
00:47:50.559 --> 00:47:53.760
of what's like and what's not like. The revolution, the giant leap forward

668
00:47:53.800 --> 00:47:59.960
are now able to generalize on that
and say, which of these paragraphs are

669
00:48:00.119 --> 00:48:01.719
like each other? Which of these
statements are like each other? Um,

670
00:48:01.880 --> 00:48:06.199
you know there's that like I like
to when I'm training my engineers about this.

671
00:48:06.360 --> 00:48:08.400
I'll say, look, there's a
million different ways that I can tell

672
00:48:08.440 --> 00:48:12.599
you what time it is. I
can say it's a quarter to three.

673
00:48:12.760 --> 00:48:15.679
I can say it is two four
five right right, right, um,

674
00:48:15.840 --> 00:48:21.440
and a million different ways of saying
that, And yet it's really just the

675
00:48:21.480 --> 00:48:25.320
same numbers at the end. That's
how the numeration system in the language models

676
00:48:25.320 --> 00:48:30.639
works. It produces these these numeric
vectors that match the meaning of the of

677
00:48:30.679 --> 00:48:37.199
the words not not that's itself,
that's interesting. So and this is how

678
00:48:37.239 --> 00:48:38.880
they align things, right, finn
or a dog? Right? I mean

679
00:48:38.880 --> 00:48:44.599
that classification is good at YEP,
but that's a fairly limited thing. And

680
00:48:44.639 --> 00:48:47.079
it's first segmentation, you know,
and I think about you know, primarily

681
00:48:47.079 --> 00:48:52.840
done a couple different things. One
classify, segments organized basically recommend you with

682
00:48:53.320 --> 00:48:58.119
optimize and optimize on a decision point, which is based on lots of different

683
00:48:58.119 --> 00:49:01.360
factors. But this is a much
different environment. I mean, it's it's

684
00:49:01.400 --> 00:49:06.880
predicting text, but it's also giving
you all this context right to help you

685
00:49:06.960 --> 00:49:08.519
understand and so you know, I
mean, the bottom line is you can

686
00:49:08.519 --> 00:49:12.400
just get about stuff. It just
prompt that thing all day long to get

687
00:49:12.800 --> 00:49:15.800
information about how to ramp up.
If I just started as a product manager

688
00:49:15.920 --> 00:49:19.599
for the software company, what should
I get focused on? You know,

689
00:49:19.679 --> 00:49:22.800
that's right, that's right, yep, yeah. And if you rewind back

690
00:49:22.239 --> 00:49:27.199
to two thousand and nine, is
the very first paper from from Jeff Hinton

691
00:49:27.440 --> 00:49:31.760
that described essentially the precursor to all
the language models we're using now. In

692
00:49:31.800 --> 00:49:35.360
that paper, you know it was
it was a little eight line completed.

693
00:49:35.440 --> 00:49:39.559
It was absolutely fascinating. Um and
and all that All that a little eight

694
00:49:39.559 --> 00:49:45.800
line program did was teach itself.
That's it. It just at leased on

695
00:49:45.880 --> 00:49:49.719
everything you gave it. It taught
itself that specific one taught itself to recognize

696
00:49:49.760 --> 00:49:52.159
handwritten digits like you know, I
draw a seven with a dash, and

697
00:49:52.280 --> 00:49:55.559
you don't where I do my ones, you know, kind of crooked,

698
00:49:55.639 --> 00:49:59.599
and you do your ones like a
slash. Right, That's all it did.

699
00:49:59.719 --> 00:50:01.480
But but the thing ones and which
ones were eights and which ones were

700
00:50:01.480 --> 00:50:06.400
threes. It separated them by itself. And that was the chilling moment that

701
00:50:07.440 --> 00:50:09.239
if it could do that, then
all we have to do is get it

702
00:50:09.280 --> 00:50:14.920
to the point where it could separate
annual reports from baseball scores. Um and

703
00:50:14.920 --> 00:50:17.920
and you know from you know,
the Declaration of Independence from uh, you

704
00:50:17.920 --> 00:50:22.440
know, Google reviews. If it
can separate those because it reads them and

705
00:50:22.519 --> 00:50:24.840
knows the difference and starts spacing them
out, then all of a sudden,

706
00:50:24.880 --> 00:50:29.239
in order to do that, in
order to solve that problem, it turns

707
00:50:29.239 --> 00:50:32.159
out, it has to understand in
order to know whether you know, whether

708
00:50:32.239 --> 00:50:37.559
these this list of abs and season
d's is a report card or a you

709
00:50:37.599 --> 00:50:44.440
know um uh and and and nfl
um you know, seasonal training report uh.

710
00:50:44.599 --> 00:50:47.239
For it to know that difference,
it has to understand the principles behind

711
00:50:47.280 --> 00:50:51.760
it. And it was able to
teach itself all that, so we didn't

712
00:50:51.760 --> 00:50:55.199
have to train it on those individual
parts. We win. That's crazy.

713
00:50:55.440 --> 00:50:59.760
I mean, that really is amazing. So now you're teaching it to you

714
00:51:00.039 --> 00:51:04.760
you've instructed it to learn to teach
itself things, and then you're building upon

715
00:51:04.840 --> 00:51:07.760
these. That's why they call it
a foundational model, Right, there's foundation

716
00:51:07.880 --> 00:51:10.840
upon which it's learning. What this
guy say to me the other day about

717
00:51:10.960 --> 00:51:15.800
how doctors can benefit from these things, because a large language model trained on

718
00:51:16.320 --> 00:51:21.639
medical data can be a very very
powerful thing. And today is the dumbest

719
00:51:21.679 --> 00:51:28.440
it will ever be. Gathering podcast
segment is coming up next year. Listenings

720
00:51:28.719 --> 00:51:35.639
Inside Analysis, case AA where every
Day is a Great Day, casey AA,

721
00:51:35.880 --> 00:51:42.320
Lomalinda. This segment sponsored by the
generous support of the Dream team.

722
00:51:42.360 --> 00:51:45.639
Looking for the keys to something bigger
and better, downsizing or relocating to the

723
00:51:45.679 --> 00:51:51.840
perfect spot. Oscar Ramirez from Century
twenty one Lowest Lower real Estate and Matt

724
00:51:51.840 --> 00:51:55.519
Flores from Secure Choice Lending are here
to help you sell or buy with their

725
00:51:55.599 --> 00:52:00.519
trusted and experienced knowledge and advice.
People are called Oscar and Matt at nine

726
00:52:00.599 --> 00:52:06.079
five one seventy five one three two
four nine. That's nine five one seven

727
00:52:06.159 --> 00:52:10.320
five one thirty two forty nine real
estate and loan advisors. Oscar and Matt

728
00:52:10.320 --> 00:52:15.239
can give you a no cost consultation. You don't have to buy anything.

729
00:52:15.559 --> 00:52:19.719
Matt and Oscar can help you figure
your way through the complicated real estate market.

730
00:52:19.760 --> 00:52:24.639
Email Oscar at loislower dot com or
on Instagram if Oscar romeiraz Garcia and

731
00:52:24.880 --> 00:52:30.840
Matt Flores at Secure Choice Lending dot
Com. Don't let today's real estate pitfalls

732
00:52:30.880 --> 00:52:36.960
stop you from dreaming. Make your
new home dreams come true. Dri number

733
00:52:37.039 --> 00:52:45.800
zero two zero seven zero three four
four zahabla espanol Khebot Club's original pure pouty

734
00:52:45.920 --> 00:52:50.639
Arco SUPERTA helps build redcorepuscles in the
blood, which carry oxygen to our organs

735
00:52:50.679 --> 00:52:54.360
and cells. Our organs and cells
need oxygen to regenerate themselves. The immune

736
00:52:54.400 --> 00:52:59.639
system needs oxygen to develop, and
cancer dies in oxygen. So the T

737
00:52:59.840 --> 00:53:02.360
is great for healthy people because it
helps build the immune system, and it

738
00:53:02.440 --> 00:53:07.920
can truly be miraculous for someone fighting
a potentially life threatening disease due to an

739
00:53:07.960 --> 00:53:12.800
infection, diabetes, or cancer.
The T is also organic and naturally caffeine

740
00:53:12.840 --> 00:53:15.079
free. A one pound package of
T is forty nine ninety five, which

741
00:53:15.119 --> 00:53:20.639
includes shipping. To order, please
visit to eebotclub dot com. Tohebo is

742
00:53:20.639 --> 00:53:24.000
spelled T like tom, A H, E, B like boy Oh.

743
00:53:24.239 --> 00:53:29.239
Then continue with the word T and
then the word club. The complete website

744
00:53:29.360 --> 00:53:32.199
is to hubot club dot com or
call us at eight one eight six one

745
00:53:32.320 --> 00:53:37.360
zero eight zero eight eight Monday through
Saturday nine am to five pm California time.

746
00:53:37.559 --> 00:53:42.880
That's eight one eight six one zero
eight zero eight eight toebot club dot

747
00:53:42.960 --> 00:53:51.400
com with sixty years of fascinating facts. This is the man from yesterday and

748
00:53:51.559 --> 00:53:54.199
back in time, we go to
this time in nineteen sixty nine, Senator

749
00:53:54.360 --> 00:54:00.440
Edward Kennedy pleads guilty here believing the
scene of a fatal auto accident which killed

750
00:54:00.559 --> 00:54:05.400
Secretary Mary Joke Packney. Ted Kennedy
is given a two month suspended sentence and

751
00:54:05.639 --> 00:54:08.280
is placed on probation. This morning, I entered a plea of guilty to

752
00:54:08.360 --> 00:54:12.559
the charge of leaving the scene of
an accident. Prior to my appearance in

753
00:54:12.719 --> 00:54:16.119
Chord, it would have been proper
for me to comment on these matters and

754
00:54:16.239 --> 00:54:22.119
from this time. In nineteen seventy
six, the FCC expands citizens band radio

755
00:54:22.400 --> 00:54:27.519
to forty channels from its present twenty
three. Currently here in nineteen seventy six,

756
00:54:27.960 --> 00:54:32.199
there are twenty million CB radios in
use. Radio Shack has the hottest

757
00:54:32.199 --> 00:54:37.960
thing on wheels today, realistic two
ACB radio. We introduced our realistic CB

758
00:54:38.079 --> 00:54:43.599
line way back in nineteen sixty and
today we have sixteen low priced mobile and

759
00:54:43.679 --> 00:54:47.119
waukee talkie models to choose from.
And from this time. In nineteen ninety

760
00:54:47.239 --> 00:54:52.400
seven, Bulworth's announces it's going to
close its four hundred store chain, bringing

761
00:54:52.480 --> 00:54:57.960
an end to an era. Over
nine thousand will lose their jobs. The

762
00:54:58.159 --> 00:55:04.440
store has been around since eighteen seventy
nine. It's world worth Stereo Spectacular seventy

763
00:55:04.480 --> 00:55:08.119
one, a fantastic once a year's
Salem albums, LPs tapes, everything in

764
00:55:08.239 --> 00:55:17.559
stereo, unbelievably low priced with more
Atman from Yesterday dot com. Do you

765
00:55:17.760 --> 00:55:22.559
like to safely leverage bank money to
earn double digit returns income tax free,

766
00:55:22.800 --> 00:55:27.760
with guarantees and no downside market risk? How can you do this? This

767
00:55:27.960 --> 00:55:30.719
is Fern's host of the Your Personal
Bank Show. One. You fund a

768
00:55:30.840 --> 00:55:37.320
high cash value policy one time to
earn dividends and interest to establish a bank

769
00:55:37.400 --> 00:55:40.280
line of credit using the cash in
your policy as collateral. When you earn

770
00:55:40.360 --> 00:55:45.360
more in dividends from your policy than
the interests the bank charges, you keep

771
00:55:45.440 --> 00:55:49.320
the difference, and the difference is
average two to five percent annually in your

772
00:55:49.400 --> 00:55:53.320
favor for the past forty plus years. Three the bank funds contributions years two

773
00:55:53.559 --> 00:55:57.960
to twenty plus. Each year the
bank adds funds your rate of return,

774
00:55:58.039 --> 00:56:01.679
increases your average rate of return and
grow to strong double digits annually within a

775
00:56:01.760 --> 00:56:06.719
few years. Contact me at your
Personal bank dot com, Your Personal bank

776
00:56:06.800 --> 00:56:10.079
dot com or eight six six two
six eight four four two two eight six

777
00:56:10.159 --> 00:56:15.400
six two six eight four four two
two for more info or tune in to

778
00:56:15.480 --> 00:56:20.239
the Your Personal Bank Show. Your
Personal Bank Show airs Tuesday's at four pm

779
00:56:20.519 --> 00:56:23.559
right here on casey AA ten fifty
am and one to six point five FFM.

780
00:56:23.639 --> 00:56:30.119
The station that leaves no listeners behind. Ivy League Auto Transport and Flavors

781
00:56:30.199 --> 00:56:35.400
Mobile Detailing would like to remind listeners. Drunk drivers remains one of the nation's

782
00:56:35.480 --> 00:56:39.079
most serious issues. Forty five percent
of traffic fatalities recording the US caused by

783
00:56:39.159 --> 00:56:43.960
an intoxicated driver. When you get
on the road, do yourself, family,

784
00:56:44.079 --> 00:56:46.880
and your community a favor. Remember
the three CS of safety, Caution,

785
00:56:47.079 --> 00:56:51.039
courtesy and common sense. If your
hand has been on the bottle,

786
00:56:51.320 --> 00:56:54.360
keep your foot off the throttle.
This is a reminder courtesy of Ivy League

787
00:56:54.360 --> 00:56:59.880
Auto Transports and Flavors Mobile Detailing,
serving our area with pride and integrity.

788
00:57:00.440 --> 00:57:05.760
Edge Manufacturing would like to remind listeners. Drunk driving remains one of the nation's

789
00:57:05.840 --> 00:57:10.039
most serious issues. Forty five percent
of traffic fatalities recorded in the US are

790
00:57:10.119 --> 00:57:14.840
caused by an intoxicated driver. When
you get on the road, do yourself,

791
00:57:15.079 --> 00:57:17.800
family, and your community of favor. Remember the three cs of safety,

792
00:57:19.079 --> 00:57:22.280
caution, courtesy, and common sense. If your hand has been on

793
00:57:22.360 --> 00:57:25.400
a bottle, keep your foot off
the throttle. This reminder is courtesy of

794
00:57:25.599 --> 00:57:30.800
Edge Manufacturing, serving our area with
pride and integrity, specializing in truck rolloff,

795
00:57:31.000 --> 00:57:36.599
welding, steering and suspensions. Called
today for all your trucking problems at

796
00:57:36.719 --> 00:57:40.079
nine O nine six five three five
seven one six. That's nine O nine

797
00:57:40.320 --> 00:57:45.559
six five three fifty seven sixteen.
That's Edge Manufacturing on the air because they

798
00:57:45.639 --> 00:57:52.559
acare. This segment is sponsored by
press Print, Southern California's best full service

799
00:57:52.679 --> 00:57:57.840
union printer and mailhouse. They offer
their lowest prices around with unmatched service and

800
00:57:57.920 --> 00:58:01.679
reliability and free delivery throughout So cal. Press Print can print anything from letterhead,

801
00:58:01.760 --> 00:58:06.679
business cards, and campaign literature to
mailers of any size, lon signs,

802
00:58:06.760 --> 00:58:09.039
banners, doorhangers, are just about
anything you might want. Press Print

803
00:58:09.119 --> 00:58:13.440
promises to save money for you,
your business, or your campaign. If

804
00:58:13.480 --> 00:58:16.559
you'd like to learn more, contact
Mike Croombrint at press Print seven one four

805
00:58:16.760 --> 00:58:22.360
three nine nine eight seven zero eight. Get the Union Bug. Do you

806
00:58:22.440 --> 00:58:25.880
want to learn and get answers to
questions not address in the mass media.

807
00:58:27.320 --> 00:58:30.199
Then you want to hear my show
business Game Chambers with me, Sarah Westall,

808
00:58:30.559 --> 00:58:35.639
I have conversations with thought leaders who
have the courage to address off limit

809
00:58:35.719 --> 00:58:38.880
topics, so you and your family
have the tools to make the right decisions.

810
00:58:39.280 --> 00:58:44.719
Join me Wednesdays at four pm on
ten fifty AM and one oh six

811
00:58:44.800 --> 00:58:49.719
point five FM. Right here at
casey AA, that station that leaves no

812
00:58:49.960 --> 00:58:57.719
listener behind. Case A Radio has
openings for one hour talk shows. If

813
00:58:57.760 --> 00:59:00.639
you want to host a radio show, now is the time. I'm make

814
00:59:00.719 --> 00:59:05.400
CACAA your flagship station. Our rates
are affordable and our services are second to

815
00:59:05.559 --> 00:59:08.880
nine. We broadcast to a population
of five million people plus. We stream

816
00:59:08.920 --> 00:59:14.239
and podcast on all major online audio
and video systems. If you've been thinking

817
00:59:14.320 --> 00:59:19.719
about broadcasting a weekly radio program on
real radio plus the Internet, contact our

818
00:59:19.800 --> 00:59:23.840
CEO at two eight one five nine
nine ninety eight hundred two eight one five

819
00:59:24.079 --> 00:59:28.440
nine nine ninety eight hundred. You
can skype your show from your home to

820
00:59:28.519 --> 00:59:31.920
our Redlands, California studio, where
our live producers and engineers are ready to

821
00:59:32.000 --> 00:59:37.800
work with you personally. A radio
program on CACAA is the perfect work from

822
00:59:37.840 --> 00:59:42.960
home avocation In these stressful times.
Just type CACAA radio dot com into your

823
00:59:43.000 --> 00:59:45.760
browser to learn more about hosting a
show on the best station in the nation,

824
00:59:46.000 --> 00:59:50.840
or call our CEO for details.
Two eight one, five nine nine

825
00:59:51.079 --> 00:59:59.199
ninety eight hundred. You're on board
caseyaa's Inland balk Express CACAA ten fifty am,

826
00:59:59.480 --> 01:00:08.480
the station that needs no less.
You're behind and be