1
00:00:01,120 --> 00:00:04,839
Speaker 1: How'd you like to listen to dot NetRocks with no ads? Easy?

2
00:00:05,360 --> 00:00:08,599
Become a patron For just five dollars a month you

3
00:00:08,640 --> 00:00:11,359
get access to a private RSS feed where all the

4
00:00:11,400 --> 00:00:14,599
shows have no ads. Twenty dollars a month will get

5
00:00:14,640 --> 00:00:18,440
you that and a special dot NetRocks patron mug. Sign

6
00:00:18,519 --> 00:00:22,359
up now at Patreon dot dot NetRocks dot com. Hey,

7
00:00:22,480 --> 00:00:26,280
Carl and Richard here with your twenty twenty four NDC schedule.

8
00:00:26,480 --> 00:00:29,199
Speaker 2: Will be at as many NDC conferences as possible this year,

9
00:00:29,239 --> 00:00:32,240
and you should consider attending no matter what. The Copenhagen

10
00:00:32,280 --> 00:00:35,840
Developers Festival happens August twenty sixth through the thirtieth. Tickets

11
00:00:35,880 --> 00:00:38,880
at Cphdevfest dot com.

12
00:00:39,119 --> 00:00:43,200
Speaker 1: NDC Porto is happening October fourteenth through the eighteenth. The

13
00:00:43,240 --> 00:00:49,200
early bird discount ends June fourteenth. Tickets at Ndcporto dot com.

14
00:00:49,240 --> 00:01:02,039
Speaker 2: And we'll see you there, we hope.

15
00:01:04,400 --> 00:01:08,599
Speaker 1: Hey, guess what it's dot net rocks again. I'm Carl

16
00:01:08,599 --> 00:01:11,920
Franklin and I'm Richard Campbell, and this is episode nineteen nineteen.

17
00:01:12,000 --> 00:01:14,760
And Buddy, I went and I looked up you know

18
00:01:14,879 --> 00:01:19,719
what happened in this year in history? So some significant

19
00:01:19,719 --> 00:01:23,799
events included the collapse of the Habsburg and German empires.

20
00:01:24,200 --> 00:01:26,920
Speaker 2: Yes, the end of World War One, end of World War.

21
00:01:26,840 --> 00:01:28,719
Speaker 1: One, the redrawing of the map of Europe at the

22
00:01:28,760 --> 00:01:33,719
Paris Peace Conference, the influence of Woodrow Wilson's fourteen points.

23
00:01:34,079 --> 00:01:36,120
Speaker 2: Interesting they call it the Paris Peace Conference. It was

24
00:01:36,120 --> 00:01:40,879
the Versiah Cords, and they were not They were not very.

25
00:01:40,680 --> 00:01:42,079
Speaker 1: Peaceful, not very peaceful.

26
00:01:42,159 --> 00:01:46,480
Speaker 2: Yeah, they were a very good setup for World War two, yep.

27
00:01:46,640 --> 00:01:49,719
Speaker 1: And sowing the seeds of World War two, growing demands

28
00:01:49,760 --> 00:01:54,159
for self determination by smaller nations, the growing involvement of

29
00:01:54,200 --> 00:01:57,799
the United States and European politics and trade all that. Yeah.

30
00:01:58,000 --> 00:02:00,599
Speaker 2: Yeah, one would argue this is the ess of the

31
00:02:00,680 --> 00:02:04,159
nation stage. Yeah, basically the model we've been functioned ever since.

32
00:02:04,359 --> 00:02:06,400
Speaker 1: I bet you didn't know Richard Campbell knew so much

33
00:02:06,400 --> 00:02:09,840
about history. O. Who am I kidding? You knew you knew?

34
00:02:14,400 --> 00:02:16,000
Go ahead and roll the music, because I got a

35
00:02:16,039 --> 00:02:26,199
pretty good better no framework, awesome? All right? What do

36
00:02:26,240 --> 00:02:30,520
you got? So you know, I'm cutting down on travel

37
00:02:30,800 --> 00:02:34,400
to conferences. At the end of the year, I'm not

38
00:02:34,479 --> 00:02:37,599
going outside the country. I'm not even sure that there

39
00:02:37,599 --> 00:02:41,080
are any more domestic conferences that I'm going to at

40
00:02:41,120 --> 00:02:43,840
the end of this year, but I still want to

41
00:02:43,879 --> 00:02:46,240
do some Blazer hands on training.

42
00:02:46,639 --> 00:02:46,840
Speaker 2: Right.

43
00:02:46,919 --> 00:02:50,680
Speaker 1: So what I've done is if you go to APPVNX

44
00:02:50,759 --> 00:02:55,000
dot com, slash training or nineteen nineteen dot po dot me,

45
00:02:55,919 --> 00:03:00,680
you'll see the training page which links to a form

46
00:03:00,719 --> 00:03:03,080
that you can fill out. I've decided I want to

47
00:03:03,120 --> 00:03:07,319
do these online classes on Mondays. Oh yeah, mostly Mondays

48
00:03:08,439 --> 00:03:10,680
from now until the end of the year, most of

49
00:03:10,680 --> 00:03:14,639
them except for you know, some week off here and there.

50
00:03:15,120 --> 00:03:17,400
And this is a form where you can select the

51
00:03:17,520 --> 00:03:21,360
dates that you want to attend. Each Monday, we'll have

52
00:03:21,400 --> 00:03:24,680
one starting at nine am Eastern and nine am Pacific,

53
00:03:24,840 --> 00:03:27,560
so that you know, if more people want to do

54
00:03:27,639 --> 00:03:30,479
in a Pacific time blah blah blah. But the deal

55
00:03:30,599 --> 00:03:33,120
is that once we hit that magic number of ten,

56
00:03:34,240 --> 00:03:36,479
then the class is going to happen, Right, So you

57
00:03:36,520 --> 00:03:40,240
should select all of the dates that you can possibly make,

58
00:03:41,000 --> 00:03:43,479
and then there's a greater chance that one of them.

59
00:03:43,360 --> 00:03:45,840
Speaker 2: Will will take But online classes.

60
00:03:45,639 --> 00:03:49,879
Speaker 1: Will online classes, but you know it's interactive. You'll be

61
00:03:49,919 --> 00:03:53,319
able to do the hands on work and even though

62
00:03:53,360 --> 00:03:55,680
we won't get to all the materials in one day,

63
00:03:56,080 --> 00:03:59,439
you still get all the materials right so you can

64
00:03:59,479 --> 00:04:00,639
go back on your own time.

65
00:04:01,159 --> 00:04:03,080
Speaker 2: Well in the class size sounds like assuming nice as

66
00:04:03,080 --> 00:04:05,240
small too, so people go be able to interact a lot.

67
00:04:05,280 --> 00:04:06,960
Speaker 1: Well, ten is the minimum. I haven't really set a

68
00:04:07,000 --> 00:04:09,479
maximum yet, but of course if worry about that problem

69
00:04:09,520 --> 00:04:12,439
when it happened, I'll worry about that problem. So that's

70
00:04:12,439 --> 00:04:15,120
what it is at Phoenix dot com slash training. Who's

71
00:04:15,120 --> 00:04:16,399
talking to us Richard crawbd to.

72
00:04:16,360 --> 00:04:19,519
Speaker 2: Comment on top of show nineteen sixteen, So just last

73
00:04:19,560 --> 00:04:21,199
one published. To be honest, this is the show we

74
00:04:21,199 --> 00:04:24,480
did with Mark Rendalls. We called how Simple is as

75
00:04:24,480 --> 00:04:28,360
Simple as Possible? And in finding conversation with Mark, as

76
00:04:28,360 --> 00:04:31,959
always he's a thoughtful guy as well as very funny.

77
00:04:31,959 --> 00:04:34,240
I think we spent most of the show laughing. And

78
00:04:34,360 --> 00:04:36,279
Rob King has this great comment. He said, once again

79
00:04:36,360 --> 00:04:39,360
another great episode which reaffirms my position that everything old

80
00:04:39,800 --> 00:04:42,839
is new again. I feel like there's a whole generation

81
00:04:42,879 --> 00:04:45,839
of developers coming up with only known JavaScript frontends and

82
00:04:45,879 --> 00:04:49,000
think the server is only for APIs or serving the JavaScript.

83
00:04:49,160 --> 00:04:51,360
I've started going back to a server side first approach

84
00:04:51,360 --> 00:04:53,199
of building apps, keeping get all in the server and

85
00:04:53,199 --> 00:04:56,000
then build it out towards the client only required. Yes,

86
00:04:56,160 --> 00:04:58,680
Lazer made this is really simple, and it's approached to

87
00:04:58,720 --> 00:05:02,759
adding interactivity as required. I've even converted a Blazer WASM

88
00:05:02,759 --> 00:05:05,399
app to pure service side because it's actually faster than

89
00:05:05,439 --> 00:05:08,639
the first page and the app doesn't need any interactivity

90
00:05:08,720 --> 00:05:12,800
beyond the forms. To that end, I've found that you

91
00:05:12,879 --> 00:05:16,759
can integrate htmx something that Mark talked about right seamless

92
00:05:16,759 --> 00:05:18,560
see into Blazer server and a built out of a

93
00:05:18,560 --> 00:05:20,920
little demo repo which I'll include in the show notes

94
00:05:21,959 --> 00:05:25,360
to Rob's GitHub repository. It's just a sort of hey,

95
00:05:25,399 --> 00:05:27,959
these two things will play together, Blazer and HGMX.

96
00:05:28,199 --> 00:05:29,040
Speaker 1: Yeah, there you go.

97
00:05:29,199 --> 00:05:31,519
Speaker 2: You know, it's just sort of a sort of recognition

98
00:05:31,600 --> 00:05:34,000
of yeah, we're kind of calling in a cycle. Like

99
00:05:34,040 --> 00:05:36,480
there's clearly cases for putting more load on the client,

100
00:05:36,480 --> 00:05:39,360
but one would argue we swung too far that way. Yes,

101
00:05:39,759 --> 00:05:42,600
although I know as we get into these conversations about

102
00:05:42,920 --> 00:05:45,319
large language models and machine learning so forth, there's now

103
00:05:45,360 --> 00:05:47,680
a big push about how do you get more of

104
00:05:47,680 --> 00:05:51,560
that workload on the client because folks are concerned now

105
00:05:51,560 --> 00:05:53,720
that too much of that workload is on the server.

106
00:05:53,959 --> 00:05:56,160
So I think it'll be some good talking points for

107
00:05:56,199 --> 00:05:56,800
today's show.

108
00:05:56,879 --> 00:06:00,720
Speaker 1: I'm really happy that a guy like Mark Render, who

109
00:06:01,000 --> 00:06:05,120
you know, go back and listen to his stuff before

110
00:06:05,160 --> 00:06:07,600
a Blazer came out. He was not really a service

111
00:06:07,680 --> 00:06:11,319
side guy at all, like very much in the in

112
00:06:11,439 --> 00:06:16,240
the composable c I CD pipeline kind of thing and

113
00:06:16,240 --> 00:06:19,800
and all you know JavaScript, and a guy like Mark

114
00:06:19,839 --> 00:06:23,040
comes out and says, you know what, I took Blazer

115
00:06:23,120 --> 00:06:25,560
server for a ride and I really like it.

116
00:06:25,639 --> 00:06:28,399
Speaker 2: Yeah, it's really great. Well, you know, and again his whole,

117
00:06:28,439 --> 00:06:30,959
that whole how simple is simple is getting back to

118
00:06:30,959 --> 00:06:32,319
the intention of what we're trying to do for the

119
00:06:32,319 --> 00:06:35,360
customer exactly right, Like I do think we get enamored

120
00:06:35,519 --> 00:06:38,720
of the technology. And again I brought this show and

121
00:06:38,759 --> 00:06:40,639
these comments out because we're about to go down a

122
00:06:40,680 --> 00:06:43,759
path about a whole other class of development, and I

123
00:06:43,800 --> 00:06:46,360
think it's good to push back on those fundamentals and say,

124
00:06:46,360 --> 00:06:48,160
what are we trying to deliver to customers?

125
00:06:48,279 --> 00:06:50,680
Speaker 1: Yep, no pressure pre shant, Yeah for sure.

126
00:06:51,560 --> 00:06:53,399
Speaker 2: So Rob, thank you so much for your comment and

127
00:06:53,439 --> 00:06:55,079
a copy of Music co Buy. It's on its way

128
00:06:55,120 --> 00:06:57,120
to you. And if you'd like a copy of Music

129
00:06:57,160 --> 00:06:58,800
co By, I write a comment on the website at

130
00:06:58,839 --> 00:07:01,000
dot at rocks dot com or on the facebooks. We

131
00:07:01,000 --> 00:07:02,639
publish every show there, and if you comment there and

132
00:07:02,680 --> 00:07:04,199
never read on the show, we'll send you a copy

133
00:07:04,199 --> 00:07:04,560
of Music to.

134
00:07:04,600 --> 00:07:06,879
Speaker 1: Go By, Music to Code by, Still Going Strong. I

135
00:07:07,120 --> 00:07:11,439
just sold another Flak collection, which is good, and I

136
00:07:11,480 --> 00:07:14,720
got inspired last night to do a new one.

137
00:07:15,000 --> 00:07:16,399
Speaker 2: How many how many have you made?

138
00:07:16,480 --> 00:07:17,720
Speaker 1: Is it like twenty one?

139
00:07:17,879 --> 00:07:20,439
Speaker 2: Twenty one? Okay, so you know you're digging around for

140
00:07:20,480 --> 00:07:21,680
another track? Huh, that's awesome.

141
00:07:21,680 --> 00:07:23,920
Speaker 1: Well, I was listening to it last night. I just

142
00:07:23,920 --> 00:07:24,560
decided to.

143
00:07:25,079 --> 00:07:26,199
Speaker 2: I don't know if you've heard it. Stuff.

144
00:07:26,279 --> 00:07:28,720
Speaker 1: It's pretty good. It is pretty good. Yeah, I've been

145
00:07:28,720 --> 00:07:31,360
listening for a long time. Yeah, I mean I started.

146
00:07:31,720 --> 00:07:35,560
I was listening to the more recent ones okay last

147
00:07:35,639 --> 00:07:37,079
night and I was like, yeah, you know, this is

148
00:07:37,120 --> 00:07:40,800
pretty good stuff. I think I got inspired to do

149
00:07:40,879 --> 00:07:41,439
another one.

150
00:07:41,759 --> 00:07:43,519
Speaker 2: Hey, you should meet that guy. He's pretty clever.

151
00:07:43,839 --> 00:07:47,000
Speaker 1: Yeah, so we'll see. I'm not announcing it now, I'm

152
00:07:47,040 --> 00:07:48,160
just announcing my intent.

153
00:07:48,439 --> 00:07:49,120
Speaker 2: You got it?

154
00:07:49,199 --> 00:07:49,560
Speaker 1: Some itch?

155
00:07:49,600 --> 00:07:51,680
Speaker 2: I like when you have an itch? Yeah, stuff happens

156
00:07:51,680 --> 00:07:54,120
when you have an itch. Yeah yeah.

157
00:07:54,199 --> 00:07:58,480
Speaker 1: Well, anyway, you can also follow us on ex Twitter.

158
00:07:58,519 --> 00:08:00,959
We've been there for years, but the cool kids are

159
00:08:00,959 --> 00:08:05,079
hanging out. I'm Astdon, I'm Carl Franklin at tech Hub dot.

160
00:08:04,920 --> 00:08:07,279
Speaker 2: Social, and I'm Rich Campbell at master do Social.

161
00:08:07,480 --> 00:08:10,959
Speaker 1: Send us a two. That's another way you can get

162
00:08:11,000 --> 00:08:15,360
a copy of music to code buy. Okay, let us

163
00:08:15,480 --> 00:08:21,519
introduce Prashant Boyar. He is six times AI MVP, three

164
00:08:21,600 --> 00:08:27,399
times Business Applications MVP, Microsoft Certified Trainer, speaker, author, and

165
00:08:27,480 --> 00:08:30,439
leading AI architect from the Washington, DC area of the

166
00:08:30,519 --> 00:08:33,879
United States. Currently, he works as an AI architect in

167
00:08:33,919 --> 00:08:37,360
the Office of the Chief Technology Officer at Applied Information

168
00:08:37,519 --> 00:08:41,320
Sciences That's AI at dot com. Braschant is on the

169
00:08:41,399 --> 00:08:46,600
leadership committee for AI Fest, AI and mL User Group

170
00:08:47,080 --> 00:08:51,360
and the Northern Virginia Data Platform User Group. As a

171
00:08:51,399 --> 00:08:55,759
renowned international speaker, he frequently presents a tech conferences. Additionally,

172
00:08:55,799 --> 00:08:59,919
Prashant was awarded the Antarctic Service Medal of the United

173
00:09:00,080 --> 00:09:05,399
States of America for his outstanding service in Antarctica. You know,

174
00:09:05,480 --> 00:09:07,559
I was going to tease you for not having enough

175
00:09:07,600 --> 00:09:10,759
stuff on your CV, but let's go right to Antarctica.

176
00:09:10,799 --> 00:09:11,399
Speaker 2: What was that?

177
00:09:13,240 --> 00:09:16,759
Speaker 3: So, first of all, thank you very much for having me.

178
00:09:17,840 --> 00:09:22,320
That was once in a lifetime experience for me. Like

179
00:09:23,080 --> 00:09:26,440
I grew up in India and two thousand and seven

180
00:09:26,559 --> 00:09:28,919
is the year I came to the United States for

181
00:09:29,080 --> 00:09:33,320
my master's degree and I got I was really fortunate

182
00:09:33,559 --> 00:09:38,000
that I got a scholarship in the university and as

183
00:09:38,080 --> 00:09:40,679
part of that scholarship, I got a chance to work

184
00:09:40,759 --> 00:09:44,360
with a lot of brilliant scientists. And those scientists were

185
00:09:45,000 --> 00:09:49,320
studying the costmes raise and everyone in that room except

186
00:09:49,320 --> 00:09:51,600
me was like really brilliant. So I had to, you know,

187
00:09:51,639 --> 00:09:55,639
fit in and I know, and you.

188
00:09:55,639 --> 00:09:58,240
Speaker 1: Might be a little unbiased towards yourself.

189
00:09:58,440 --> 00:10:01,480
Speaker 3: Now, Like if you if you talk to any experimental

190
00:10:01,519 --> 00:10:06,120
physics physicists or in theoretical physicists, not only they know

191
00:10:06,600 --> 00:10:09,480
software development really well, but they also know all this

192
00:10:09,639 --> 00:10:12,639
machine learning match and some advanced mathematics really well. So

193
00:10:12,639 --> 00:10:14,960
I have like after that, I have really high respect

194
00:10:15,039 --> 00:10:18,759
for all these physicists because for the work they do.

195
00:10:19,480 --> 00:10:22,000
And if you compare the salaries against that, you will

196
00:10:22,039 --> 00:10:22,639
be surprised.

197
00:10:22,720 --> 00:10:25,279
Speaker 2: Yeah, they don't get paid much for yes, as hard

198
00:10:25,279 --> 00:10:28,679
as they as they work. Antarctica is a great location

199
00:10:28,960 --> 00:10:32,200
for doing stuff like measuring comic cause of rays. You know,

200
00:10:32,279 --> 00:10:35,720
the atmosphere is a little thinner, and it's very very dark,

201
00:10:36,000 --> 00:10:39,080
and there's a minimum amount of noise, but it's and

202
00:10:39,399 --> 00:10:41,360
it's sad. That's like the best time to do that

203
00:10:41,399 --> 00:10:47,399
work is midwinter of Antarctica where there's no sunlight for months,

204
00:10:48,200 --> 00:10:50,240
so for the tough place to be for shann you

205
00:10:50,360 --> 00:10:51,200
hung out there.

206
00:10:51,320 --> 00:10:56,519
Speaker 3: Yes, So as part of that experiment, like we built,

207
00:10:56,759 --> 00:11:00,200
it was a joint collaboration WITHWEN multiple universities, and ASA

208
00:11:00,399 --> 00:11:04,639
was the main sponsor because typically NASA sponsors this, you know,

209
00:11:04,720 --> 00:11:06,960
give this huge grant to the unisity so that they

210
00:11:06,960 --> 00:11:10,840
can do research on their behalf. And as part of

211
00:11:10,879 --> 00:11:14,120
the research, you know, the best results to get is

212
00:11:14,159 --> 00:11:17,039
via satellite, but you know, sending any payload to satellite

213
00:11:17,080 --> 00:11:19,679
is very expensive. So they came up with an approach

214
00:11:19,720 --> 00:11:24,320
about how we send this really large duration balloons in

215
00:11:24,399 --> 00:11:30,000
Antarctic continent which fly around you know the south you know,

216
00:11:30,559 --> 00:11:33,679
South Pole for around twenty to eighty days, so not

217
00:11:33,679 --> 00:11:36,480
really twenty to thirty days, collect as much data as possible,

218
00:11:36,519 --> 00:11:39,000
and then we take years and years to analyze the data.

219
00:11:39,720 --> 00:11:43,720
And not many people know, but US government has a

220
00:11:43,720 --> 00:11:47,200
really big station there, like Unartic continent is huge and

221
00:11:47,799 --> 00:11:50,519
a lot of countries have stations, but US in particular

222
00:11:50,519 --> 00:11:53,279
has like huge stations over there. And the station I

223
00:11:53,360 --> 00:11:56,919
went to, it's called Macmurdo Station. So when I was

224
00:11:56,960 --> 00:12:02,519
there into like in the end of twenty eighteen, sorry,

225
00:12:02,559 --> 00:12:05,759
two thousand and eight, I was there for almost a

226
00:12:05,799 --> 00:12:09,080
couple of months. And that is like a typical summer

227
00:12:09,120 --> 00:12:12,120
time in Antarctica, and that's the time where they will

228
00:12:12,159 --> 00:12:15,360
have like around thousand people on the base and everybody

229
00:12:15,480 --> 00:12:18,559
will be there to conduct some kind of scientific experiment

230
00:12:19,120 --> 00:12:21,320
and to support those experiment. They also need a lot

231
00:12:21,360 --> 00:12:24,840
of people, like they need cooks, they need chefs, you know,

232
00:12:24,879 --> 00:12:27,240
they knew it, people who can like blow the roads

233
00:12:27,240 --> 00:12:29,360
and all this kind of stuff. And I met a

234
00:12:29,360 --> 00:12:32,440
lot of interesting people there who like, let's say a

235
00:12:32,440 --> 00:12:35,519
software engineering by trade. They work here nine months in

236
00:12:35,639 --> 00:12:38,200
US continent, and for three months they just go to

237
00:12:38,240 --> 00:12:40,200
Antarctica and work as a chef or as a cook

238
00:12:40,240 --> 00:12:43,200
because they just like it there and the pay is good.

239
00:12:43,120 --> 00:12:45,559
Speaker 1: And the base has good internet access. I hear yes,

240
00:12:45,600 --> 00:12:47,639
and I know that because I have a friend I

241
00:12:47,720 --> 00:12:51,679
went to grammar school with who is messaging me from Antarctica.

242
00:12:51,720 --> 00:12:54,480
I'm like, what you're aware, but he was at that

243
00:12:54,600 --> 00:12:59,440
base and yeah, so pretty cool. How in the summertime

244
00:12:59,519 --> 00:13:03,960
in Antaro would you think of it as bomby? Would

245
00:13:04,000 --> 00:13:07,200
you go out in shorts? What's the temperature?

246
00:13:07,480 --> 00:13:10,759
Speaker 3: So temperature is not that bad. You can go out

247
00:13:10,840 --> 00:13:13,200
and go and go out in shots only for ten

248
00:13:13,240 --> 00:13:14,440
to fifteen minutes, that's it.

249
00:13:15,000 --> 00:13:15,559
Speaker 1: Wow.

250
00:13:16,240 --> 00:13:20,000
Speaker 3: So like the during the summer time, you know there

251
00:13:20,039 --> 00:13:23,279
the temperature was in the range of you know, like

252
00:13:23,440 --> 00:13:28,480
minus let's say ten degrees celsius to plus five degrees celsius. Okay,

253
00:13:28,600 --> 00:13:30,360
but there used to be some drays where you know,

254
00:13:30,440 --> 00:13:32,879
it was like very very bad, Like we had like

255
00:13:32,919 --> 00:13:35,279
one day where it was minus twenty degrees celsius and

256
00:13:35,799 --> 00:13:39,000
things were pretty rough over there. But you have heated

257
00:13:39,080 --> 00:13:42,480
rooms and you know, like food is awesome, there are

258
00:13:42,519 --> 00:13:45,440
some recreational activities and you know, and that when you

259
00:13:45,480 --> 00:13:48,360
go to those kinds of places, you kind of realize, okay,

260
00:13:48,840 --> 00:13:52,039
how fast paced your life is back in like you know,

261
00:13:52,080 --> 00:13:54,919
you know, the civilized world, and how much of a

262
00:13:55,000 --> 00:13:57,039
time you got when you go to these kind of places.

263
00:13:57,120 --> 00:14:00,320
Speaker 1: Did you ever have to slice open a banta with

264
00:14:00,399 --> 00:14:03,799
your lightsaber to keep warm? No? No, that's the question.

265
00:14:05,039 --> 00:14:05,279
Speaker 2: I was.

266
00:14:05,279 --> 00:14:07,919
Speaker 3: I was fortunate enough. No, I went there on a

267
00:14:07,960 --> 00:14:09,039
good good time over.

268
00:14:08,840 --> 00:14:11,399
Speaker 2: There, Yeah, very good. You only had to smell them

269
00:14:11,399 --> 00:14:13,360
on the outside.

270
00:14:15,279 --> 00:14:18,440
Speaker 1: Oh my, sorry for that little diversion, but I just

271
00:14:18,480 --> 00:14:18,919
had to know.

272
00:14:18,960 --> 00:14:19,440
Speaker 2: It's great.

273
00:14:19,480 --> 00:14:23,639
Speaker 1: So let's talk about Microsoft co Pilot Studio. I know

274
00:14:24,240 --> 00:14:27,360
because I went to build with Richard, and we know

275
00:14:27,480 --> 00:14:32,559
that it's for building your own custom agents and you

276
00:14:32,600 --> 00:14:34,799
can connect them to a bunch of sources of data,

277
00:14:35,000 --> 00:14:38,080
and we know that you can hook them up into

278
00:14:38,559 --> 00:14:42,000
M three sixty five and all that. But what don't

279
00:14:42,039 --> 00:14:42,240
we know?

280
00:14:42,480 --> 00:14:45,840
Speaker 3: Okay, so let me go back to the history and

281
00:14:45,960 --> 00:14:50,960
start with Microsoft first, foray into the cornosential AI. Like

282
00:14:51,039 --> 00:14:53,759
if you remember in March twenty sixteen, I think it's

283
00:14:53,799 --> 00:14:57,120
the first time when Satya Nadela went onto the stage

284
00:14:57,159 --> 00:15:01,519
and he announced about Microsoft bought framework. Because still that time,

285
00:15:02,039 --> 00:15:05,639
there was no good story when it comes to building

286
00:15:05,720 --> 00:15:09,240
let's say, any kind of child bought or conversational experiences.

287
00:15:09,600 --> 00:15:13,519
And after maybe two or three, after a month after that,

288
00:15:14,279 --> 00:15:16,679
I know Microsoft is getting a lot of good price

289
00:15:16,759 --> 00:15:19,759
right now because of their partnership with open AI. But

290
00:15:19,840 --> 00:15:22,559
they also had their fair share of failures as well.

291
00:15:23,440 --> 00:15:26,759
I think In March twenty sixteen, they launched a Twitter

292
00:15:26,799 --> 00:15:32,960
bought called day oh yeah yes, yes yeah, and its

293
00:15:33,000 --> 00:15:36,279
long form was thinking about you And they launched on

294
00:15:36,360 --> 00:15:39,919
Twitter and that pot was from Microsoft Research and it

295
00:15:40,039 --> 00:15:42,879
was their first public foray into the world of AI.

296
00:15:43,759 --> 00:15:46,320
And people soon figure out on Twitter, you know, to

297
00:15:46,399 --> 00:15:53,039
teach that bot to say some nasty stuff and Microsoft yes, yes, yeah, Microsoft.

298
00:15:53,399 --> 00:15:57,639
Microsoft try to like you know again reteaching like good stuff,

299
00:15:57,639 --> 00:15:59,600
but within sixteen hours they had to take it down

300
00:16:00,320 --> 00:16:04,159
and they still now a lot of key executives at

301
00:16:04,200 --> 00:16:07,399
Microsoft talks about that particular failure saying like, hey, we

302
00:16:07,519 --> 00:16:10,519
learn a lot from that failure, like how what other

303
00:16:10,600 --> 00:16:14,600
things not to do with AI stuff? And so they

304
00:16:14,600 --> 00:16:17,799
continue to work on board framework for next three years.

305
00:16:18,279 --> 00:16:22,279
But board framework, like is very complex.

306
00:16:22,519 --> 00:16:25,120
Speaker 2: Yeah, we did a show, We did shows on bot framework,

307
00:16:25,159 --> 00:16:28,639
like you said, like twenty seventeen. It's hard to think

308
00:16:28,639 --> 00:16:31,799
about that now because the opening eye things overwhelmed everything. Yes,

309
00:16:32,440 --> 00:16:34,559
and they made a lot of sense, like the fact

310
00:16:34,559 --> 00:16:37,440
bot to me was totally logical that you could point

311
00:16:37,720 --> 00:16:39,559
a piece of software at a fact and it would

312
00:16:39,600 --> 00:16:41,639
give you sort of a natural language interface to it.

313
00:16:42,879 --> 00:16:45,759
Speaker 3: But still, like Microsoft struggle to get really good traction

314
00:16:45,879 --> 00:16:49,039
with a lot of customers and developers because it was

315
00:16:49,080 --> 00:16:53,519
extremely hard to get you know, your conversational experiences right.

316
00:16:53,960 --> 00:16:57,559
And then in November twenty nineteen they launch a low

317
00:16:57,559 --> 00:17:01,279
code version of it called Power Virtual Asian and the

318
00:17:01,440 --> 00:17:04,920
name Power itself can give you indication which product family

319
00:17:05,039 --> 00:17:07,839
belongs to. It was from a power platform. So they

320
00:17:07,880 --> 00:17:11,519
continued on that journey and then basically there was a

321
00:17:11,559 --> 00:17:15,839
time where the entire investment on board framework stopped and

322
00:17:15,920 --> 00:17:19,519
all the things like Microsoft was doing on Power virtualation side.

323
00:17:19,920 --> 00:17:23,160
They had decent success. But then the real success or

324
00:17:23,279 --> 00:17:28,720
real i'll call life changing moment for that particular product

325
00:17:28,799 --> 00:17:32,759
or group was Ignite last year. Fen went on stage

326
00:17:32,759 --> 00:17:35,880
and say, hey, Microsoft now is the co pilot company,

327
00:17:36,440 --> 00:17:38,319
and if you would like to create a low code,

328
00:17:38,319 --> 00:17:40,920
no code co pilots, then Copilot Studio is the product.

329
00:17:41,759 --> 00:17:44,839
And I know Microsoft get a lot of flak, especially

330
00:17:44,839 --> 00:17:49,279
their marketing department, for naming like weird products, but I

331
00:17:49,319 --> 00:17:51,279
think this time really nailed it.

332
00:17:52,079 --> 00:17:55,160
Speaker 2: To be clear, GitHub came up with the name, right,

333
00:17:55,799 --> 00:17:58,000
That's why it's such a good name, because Microsoft didn't

334
00:17:58,039 --> 00:17:58,359
think of it.

335
00:17:59,640 --> 00:18:02,640
Speaker 1: We have a history of making really long names that

336
00:18:02,920 --> 00:18:05,440
are very kind of academic sounding.

337
00:18:05,519 --> 00:18:06,799
Speaker 2: Well they're named by Kimmittee.

338
00:18:07,240 --> 00:18:10,240
Speaker 3: Yes, yeah, and someone I heard. I don't know how

339
00:18:10,279 --> 00:18:12,839
true it is, but this product was supposed to have

340
00:18:12,880 --> 00:18:15,880
another name. They already have a lot of printed name

341
00:18:16,000 --> 00:18:18,720
and banners on it, but just three one week or

342
00:18:18,720 --> 00:18:21,599
two weeks before Ignite, someone came up with this name

343
00:18:21,640 --> 00:18:23,880
and they have to, you know, do a lot of

344
00:18:24,079 --> 00:18:25,839
you know, last minute stuff.

345
00:18:26,079 --> 00:18:29,640
Speaker 2: There's been a lot of name twitching, right copilot for

346
00:18:29,720 --> 00:18:33,920
Microsoft three sixty five, Microsoft three sixty five copilot like it's.

347
00:18:33,720 --> 00:18:37,160
Speaker 1: It's you know, the AI thing has been sort of

348
00:18:37,200 --> 00:18:40,359
like a gold rush, hasn't it. Yes, like every company

349
00:18:40,440 --> 00:18:43,279
wants to get their put their stake in the ground

350
00:18:43,319 --> 00:18:45,920
and say we're doing AI or we're doing you know,

351
00:18:46,079 --> 00:18:51,759
generative AI. And I mean I don't I don't want

352
00:18:51,759 --> 00:18:55,599
our listeners who are dot developers to be thinking that, oh,

353
00:18:55,720 --> 00:19:00,559
development is gone. Microsoft is just focusing completely from now

354
00:19:00,599 --> 00:19:03,839
on on AI. Because if you went to build, there

355
00:19:03,880 --> 00:19:06,440
wasn't a whole lot about dot net, you know what

356
00:19:06,440 --> 00:19:08,839
I mean. There wasn't a whole lot of code, plumbing code.

357
00:19:08,839 --> 00:19:13,039
It was all focused on AI. But you know, we're,

358
00:19:13,279 --> 00:19:16,119
as Richard said in the beginning when he read that comment,

359
00:19:16,880 --> 00:19:21,519
we need to stay focused on delivering value to our customers.

360
00:19:21,960 --> 00:19:26,279
And the sort of AI thing just smacked this in

361
00:19:26,279 --> 00:19:28,519
the face and came out of nowhere and says, hey,

362
00:19:28,599 --> 00:19:32,799
guess what this is going to allow you developers to

363
00:19:32,920 --> 00:19:37,039
provide better value to your customers. So really interested to

364
00:19:37,119 --> 00:19:39,160
hear some scenarios about that, and you.

365
00:19:39,400 --> 00:19:41,680
Speaker 3: Write a call like, you know, just don't think that

366
00:19:41,759 --> 00:19:44,480
copilot Studio is the only product that I will need

367
00:19:44,519 --> 00:19:47,640
to use, you know, when I had to create custom copilots.

368
00:19:47,880 --> 00:19:51,079
There is another story in Azure world called AI Studio.

369
00:19:51,519 --> 00:19:54,119
So if you are a professional developer and you don't

370
00:19:54,160 --> 00:19:56,240
want to use let these drag and drop kind of

371
00:19:56,240 --> 00:19:58,559
tools or low code no code tools and you would

372
00:19:58,599 --> 00:20:04,559
have full control. Studio is your choice where since it's

373
00:20:04,599 --> 00:20:06,839
part of Azure, you know it has a really good

374
00:20:06,960 --> 00:20:10,319
pro developer story. You can be a dot net developer,

375
00:20:10,319 --> 00:20:12,559
you can be a Python developer, and you can use

376
00:20:12,599 --> 00:20:15,519
all the good services in Azure. And one other thing

377
00:20:15,599 --> 00:20:19,240
I like about AI service a Azure EI studio is

378
00:20:19,960 --> 00:20:23,119
the model selection. Like. One thing Microsoft has done which

379
00:20:23,480 --> 00:20:27,319
not many people know is Microsoft launched something called Model

380
00:20:27,319 --> 00:20:30,400
as a service where you can bring in some open

381
00:20:30,440 --> 00:20:32,920
source model as well and host those in Azure and

382
00:20:33,000 --> 00:20:35,079
use those in your Genia applications, so you don't have

383
00:20:35,119 --> 00:20:37,640
to go with let's say models from open Ai or

384
00:20:37,720 --> 00:20:40,000
models from let's say Microsoft for that matter. You can

385
00:20:40,119 --> 00:20:43,960
use Lama hugging pace. You know, those models are also available.

386
00:20:43,960 --> 00:20:45,440
So if you want to create a co pilot or

387
00:20:45,480 --> 00:20:48,319
agent using that, you can do that now in Asure.

388
00:20:48,519 --> 00:20:52,160
Speaker 1: But they still have GPT four GPT four oh dally

389
00:20:52,519 --> 00:20:57,359
all that stuff from AI Open ai. But all right,

390
00:20:57,480 --> 00:21:01,480
so wow, So that is really the fundamental difference between

391
00:21:01,519 --> 00:21:05,920
these two is that co pilot studio is maybe for

392
00:21:06,440 --> 00:21:10,359
draggy dropping business people and Azure AI studios more for us.

393
00:21:11,079 --> 00:21:14,400
Speaker 3: There is one fundamental difference which not many people tay

394
00:21:14,480 --> 00:21:17,759
pay attention. So when it comes to building a co

395
00:21:17,880 --> 00:21:20,759
pilot or any kind of a application, a lot of

396
00:21:20,799 --> 00:21:23,880
time people don't think about the UX part of it.

397
00:21:24,880 --> 00:21:27,880
I know. That's where a copilot studio shines really well.

398
00:21:27,920 --> 00:21:30,480
Like out of the box, it has really good capacity

399
00:21:30,519 --> 00:21:33,519
of you know, having a really good UX. Let's say

400
00:21:33,559 --> 00:21:36,480
I launch a co pilot for my existing enterprise app.

401
00:21:37,000 --> 00:21:40,160
But then that copilot should able to understand or behave

402
00:21:40,240 --> 00:21:43,960
like actual human right. The reason chad jipt got so

403
00:21:44,279 --> 00:21:47,839
popular so fast is because we can interact with it

404
00:21:48,359 --> 00:21:51,519
and you know, it can understand like at least what

405
00:21:51,519 --> 00:21:53,920
I'm trying to say maybe ninety to ninety percent of time,

406
00:21:53,960 --> 00:21:56,720
and it can give me good information back. And if

407
00:21:56,720 --> 00:21:59,759
you compare that with old version of the chatboards that

408
00:21:59,839 --> 00:22:02,079
more of these public facing website used to have. Those

409
00:22:02,160 --> 00:22:04,880
used to be like, oh the major retailers or bank

410
00:22:05,000 --> 00:22:06,920
used to have those, like used to be simple time

411
00:22:07,079 --> 00:22:10,000
st where the first thing I used to google that

412
00:22:10,039 --> 00:22:12,240
time is ourbing is what's the best way to get

413
00:22:12,279 --> 00:22:14,400
to an actual human agent rather than you know, going

414
00:22:14,440 --> 00:22:16,039
through their automated.

415
00:22:16,039 --> 00:22:18,000
Speaker 2: Response because that's what everybody does.

416
00:22:18,039 --> 00:22:21,680
Speaker 3: Yes, But now with this, after the generative AI and

417
00:22:21,720 --> 00:22:24,559
some of the advancement, you know, you can at least

418
00:22:24,559 --> 00:22:27,880
have some decent conversation. And that's where this co Pilot

419
00:22:27,920 --> 00:22:30,400
Studios shine where it has a lot of widgets, it

420
00:22:30,480 --> 00:22:34,519
has a lot of you know, building capabilities that if

421
00:22:34,599 --> 00:22:40,480
I would like to build a good conversational experience, Copilot

422
00:22:40,519 --> 00:22:43,200
Studio really shines as compared to Azure Eye Studio.

423
00:22:43,279 --> 00:22:45,400
Speaker 1: Can I take the agents that I make in copiled

424
00:22:45,440 --> 00:22:48,319
studio and move them over into Azure Ai studio and

425
00:22:48,480 --> 00:22:49,119
enhance them.

426
00:22:49,519 --> 00:22:53,160
Speaker 3: Right now, the create direct conversation story is not there.

427
00:22:53,799 --> 00:22:55,839
But what you can do is, let's say, if you

428
00:22:55,880 --> 00:22:59,279
have already an agent in Azure AI side, you can

429
00:22:59,279 --> 00:23:02,440
directly call that from a Copilot studio, so you can

430
00:23:02,480 --> 00:23:04,759
do integration. But there is no like Magic vand to

431
00:23:04,839 --> 00:23:05,680
convert those.

432
00:23:05,680 --> 00:23:06,599
Speaker 1: That's actually better.

433
00:23:06,960 --> 00:23:09,640
Speaker 2: Yes, well, it also reminds you that there's more than

434
00:23:09,640 --> 00:23:12,039
one team instead of Microsoft working on this stuff, right,

435
00:23:13,039 --> 00:23:15,319
you know, that's that's sort of the real reality is

436
00:23:15,319 --> 00:23:17,680
that there's lots of different places moving at once.

437
00:23:17,920 --> 00:23:21,119
Speaker 3: And I'm a big believer of Fusion Team where hey, developers,

438
00:23:21,160 --> 00:23:23,440
you focus on you know, doing really crazy stuff behind

439
00:23:23,440 --> 00:23:26,160
the scene, you know, make the magic, but also gave

440
00:23:26,200 --> 00:23:29,079
a lot of information worker or business users and ability

441
00:23:29,119 --> 00:23:31,960
to you know, take the advantage of what they know

442
00:23:32,039 --> 00:23:34,759
from the business and then you know, have an easy

443
00:23:34,799 --> 00:23:39,680
way to interact with the back end system. So so

444
00:23:39,759 --> 00:23:42,200
Copilo Studio also have that functionality way. Let's say I

445
00:23:42,240 --> 00:23:44,480
have my logic app I would like to call or

446
00:23:44,599 --> 00:23:46,880
I would like to call my EPI. I can do

447
00:23:46,960 --> 00:23:49,839
that directly. There is an STTP connector available or I

448
00:23:49,880 --> 00:23:54,400
can also invoke a power Automate cloud flow, and power

449
00:23:54,400 --> 00:23:56,720
Automate has more than twelve hundred out of the box

450
00:23:56,720 --> 00:23:58,759
connectors where you know, or I can build my custom

451
00:23:58,799 --> 00:24:01,599
connector as well, which then can you know, call my

452
00:24:01,640 --> 00:24:03,559
back end APIs and then can get the job done.

453
00:24:03,640 --> 00:24:06,759
Speaker 1: I have a quick story here. I got an email

454
00:24:07,000 --> 00:24:10,279
last week from somebody who said, you know, I see

455
00:24:10,319 --> 00:24:12,559
that you have a YouTube channel and you've got some

456
00:24:12,680 --> 00:24:16,480
great content there, and uh, you know, for a very

457
00:24:16,559 --> 00:24:20,240
low fee. Actually it was pretty good. I will, I will,

458
00:24:20,839 --> 00:24:28,000
I will generate SEO happy titles, descriptions, tags and stuff

459
00:24:28,000 --> 00:24:30,440
for your YouTube content. So I said, okay, sure, I'll

460
00:24:30,480 --> 00:24:33,559
try it. And what I got back was a RAR

461
00:24:33,799 --> 00:24:36,839
file full of text files for and a lot of

462
00:24:36,839 --> 00:24:41,480
them were for the dot net rocks videos. And they

463
00:24:41,599 --> 00:24:45,240
kept mentioning the dot net Rocks videos are simply a

464
00:24:45,279 --> 00:24:49,480
still shot with our podcast that plays on it. And

465
00:24:49,519 --> 00:24:52,079
this thing went on and on about how you will

466
00:24:52,200 --> 00:24:57,160
learn from these training videos, hands on experience and stuff.

467
00:24:59,039 --> 00:25:03,279
They got to the they got to the songs that

468
00:25:03,759 --> 00:25:07,519
my band has put up there, and they called, you know, Nachelle,

469
00:25:07,559 --> 00:25:10,319
who's our background singer, but she does sing on some

470
00:25:10,359 --> 00:25:12,960
of them, but on this one that I was singing

471
00:25:13,160 --> 00:25:17,119
like all of it lead singer Nachelle Rollins blah blah blah,

472
00:25:17,200 --> 00:25:20,480
this big, grandiose stuff, and I wrote it back. I'm

473
00:25:20,480 --> 00:25:23,039
not gonna say it is by her own back and say, dude,

474
00:25:23,079 --> 00:25:27,640
this is bullsh this is some AI generated crap that

475
00:25:27,720 --> 00:25:31,440
doesn't know a still video from you know, from a

476
00:25:31,640 --> 00:25:35,119
training video for crying out loud. So yeah, I won't

477
00:25:35,119 --> 00:25:40,960
be using their actually incorrect, factually incorrect and being sold

478
00:25:41,160 --> 00:25:44,599
as you know, the I don't know whatever.

479
00:25:45,359 --> 00:25:48,680
Speaker 2: So I mean, you say, quietly said Fusion development there,

480
00:25:48,680 --> 00:25:51,319
which is that power platform working with dot net dev

481
00:25:51,400 --> 00:25:53,519
and so forth. So you see a sort of same

482
00:25:53,640 --> 00:25:57,400
dynamic where we have domain experts, which I think is

483
00:25:57,440 --> 00:25:59,440
the big biggest challenge when you're building a lot of

484
00:25:59,480 --> 00:26:04,039
these language models, is you need a domain expert checking

485
00:26:04,079 --> 00:26:07,279
for factual accuracy, knows enough to know that is right

486
00:26:07,400 --> 00:26:10,519
or that is wrong. Yes, who actually sang on this track,

487
00:26:11,279 --> 00:26:13,880
as well as some coding knowledge to be able to

488
00:26:13,920 --> 00:26:14,920
pull the pieces together.

489
00:26:15,079 --> 00:26:17,480
Speaker 3: Yes, that is, that is correct. And one of the

490
00:26:17,480 --> 00:26:21,400
things is, you know a lot of people like really

491
00:26:21,400 --> 00:26:25,680
got excited after the rise of generative EI, but there

492
00:26:25,720 --> 00:26:29,680
are like legitimate you know application, A lot of organizations

493
00:26:29,680 --> 00:26:33,279
built which doesn't use generative AI at all. Like one

494
00:26:33,319 --> 00:26:36,799
of my customers, they did use Copilot Studio, and they

495
00:26:36,839 --> 00:26:40,160
did launch a custom Copilot for their public facing website,

496
00:26:40,759 --> 00:26:43,319
but they didn't use any of the generative AI features

497
00:26:43,359 --> 00:26:46,880
at all, Like, because there will be some instances where

498
00:26:46,960 --> 00:26:50,200
you don't want to give a sort of correct or

499
00:26:50,200 --> 00:26:53,400
incorrect answer, right, You want to get give a factually

500
00:26:53,440 --> 00:26:56,119
correct answer. And that are some of the things that

501
00:26:56,160 --> 00:26:59,119
are there in Copilot Studio where for some of the

502
00:26:59,160 --> 00:27:02,720
information you absolutely want to give the appropriate answers maybe

503
00:27:02,720 --> 00:27:06,039
from your own knowledge pace or maybe from your own

504
00:27:06,200 --> 00:27:09,519
database or you and that you can do. And there

505
00:27:09,519 --> 00:27:12,599
are certain things where you think, okay, I can use

506
00:27:12,640 --> 00:27:14,960
generative AI for some of the answers. You can do

507
00:27:15,000 --> 00:27:17,240
that segregation wasy very easily as well.

508
00:27:17,880 --> 00:27:22,240
Speaker 1: I'm just another story. I love telling these stories my

509
00:27:22,359 --> 00:27:24,359
Facebook feed. I don't know about you guys, if you're

510
00:27:24,400 --> 00:27:26,799
on Facebook at all, but every once in a while

511
00:27:26,839 --> 00:27:31,960
I scroll and I'm seeing these accounts that are something

512
00:27:32,119 --> 00:27:37,799
like tiny houses, right, and then this clearly generated beautiful

513
00:27:37,960 --> 00:27:42,240
scene of this wonderful house with multiple staircases and flowing

514
00:27:42,279 --> 00:27:45,000
waterfalls by a brook. And it's just like a place

515
00:27:45,000 --> 00:27:47,480
that you'd want to be, right And the caption is

516
00:27:47,599 --> 00:27:51,480
like peaceful place. It's not. Hey, this is where I

517
00:27:51,519 --> 00:27:56,319
am right now, it's not. And to get millions of

518
00:27:56,400 --> 00:28:02,599
views and hundreds of thousands of comments, it just makes

519
00:28:02,599 --> 00:28:07,319
me mad. Welcome to the new world.

520
00:28:07,680 --> 00:28:10,240
Speaker 2: Well, and again it's you used a bought to generate

521
00:28:10,279 --> 00:28:12,160
the content. Who's to say you didn't use a bought

522
00:28:12,240 --> 00:28:16,799
to generate the hits? Well, there you go, all right,

523
00:28:16,880 --> 00:28:20,240
So let's take a break, and when we come back,

524
00:28:20,279 --> 00:28:21,400
I've got a couple of questions.

525
00:28:21,480 --> 00:28:25,599
Speaker 1: Great, did you know there's a dot net on aws community?

526
00:28:26,279 --> 00:28:30,279
Follow the social media blogs, YouTube influencers, and open source

527
00:28:30,359 --> 00:28:34,039
projects and add your own voice. Get plugged into the

528
00:28:34,079 --> 00:28:38,759
dot net on aws community at aws dot Amazon dot com,

529
00:28:38,839 --> 00:28:43,400
slash dot net. Hey Carl, here are you maximizing your

530
00:28:43,440 --> 00:28:48,000
dot net applications potential? Rayguns tools not only catch errors,

531
00:28:48,200 --> 00:28:52,839
they provide actionable insights to optimize your code experience the

532
00:28:52,920 --> 00:28:56,559
power of detailed reports and real time monitoring to enhance

533
00:28:56,599 --> 00:29:00,839
your applications performance. Join thousands of dot next developers who

534
00:29:00,839 --> 00:29:04,559
trust Raygun to help them keep their applications running smoothly.

535
00:29:05,240 --> 00:29:08,960
Visit raygun dot com, slash dot net rocks. That's raygun,

536
00:29:09,240 --> 00:29:11,920
r A y g U N dot com, slash do

537
00:29:12,200 --> 00:29:15,319
tn E t r O c ks for your free

538
00:29:15,400 --> 00:29:16,480
fourteen day trial.

539
00:29:17,680 --> 00:29:20,920
Speaker 2: We're back. It's dot at rocks. I'mrogard Campbell. Let's call Franklin. Hey, Hey,

540
00:29:21,319 --> 00:29:23,960
I'm talking to our friend Forshett about the role of

541
00:29:24,680 --> 00:29:27,880
as your AI studio as well as the co pilot studio,

542
00:29:27,960 --> 00:29:29,400
because what's better than two?

543
00:29:31,759 --> 00:29:33,960
Speaker 1: And I like how you can use them together.

544
00:29:34,119 --> 00:29:36,680
Speaker 2: What's the call out for Patreon there, Carl.

545
00:29:36,960 --> 00:29:39,480
Speaker 1: Yeah, And if you don't want to hear those ads

546
00:29:39,559 --> 00:29:43,480
during the show, you can subscribe to our Patreon account

547
00:29:43,519 --> 00:29:45,279
for five bucks a month. You can get a feed

548
00:29:45,319 --> 00:29:47,920
that has no ads. So there you go.

549
00:29:48,359 --> 00:29:53,720
Speaker 2: Awesome, Forsham, you mentioned like the ux of a co

550
00:29:53,960 --> 00:29:56,039
pilot app. Isn't it just a textbox?

551
00:29:56,880 --> 00:30:00,039
Speaker 3: Not necessarily? Sometimes it can be text paused, but some

552
00:30:00,119 --> 00:30:03,279
times you may want to show a really nice adapt card.

553
00:30:03,480 --> 00:30:06,039
Sometimes you may have to show an image or maybe

554
00:30:06,079 --> 00:30:08,680
the user is looking for a video. So those kind

555
00:30:08,720 --> 00:30:11,759
of things like if I go with traditional board framework way,

556
00:30:12,480 --> 00:30:14,200
those kind of things are really hard to you know,

557
00:30:14,319 --> 00:30:19,200
get it right. And another thing way where Copilot Studio

558
00:30:19,279 --> 00:30:23,079
really shines is you know, like how I can support

559
00:30:23,079 --> 00:30:26,759
my copilot across multiple devices or multiple experience. That is

560
00:30:26,920 --> 00:30:29,000
like one of the biggest challenges I have seen with

561
00:30:29,079 --> 00:30:33,359
multiple conversational experiences. Like, yeah, like I build it, it's

562
00:30:33,359 --> 00:30:35,920
worse on my machine, or it works in the environment

563
00:30:35,920 --> 00:30:38,720
where I have deployed, But what about now the customer

564
00:30:38,799 --> 00:30:42,000
wants to use the same copilot on Microsoft Teams, or

565
00:30:42,039 --> 00:30:44,279
they want to use it on a public facing website,

566
00:30:44,400 --> 00:30:46,960
or the customer loves Slack they want to use it there.

567
00:30:47,000 --> 00:30:50,240
Like how I can you know, now redo everything so

568
00:30:50,279 --> 00:30:52,759
that you know, my custom copilot will work on that

569
00:30:52,839 --> 00:30:55,279
experience as well or on that service as well. So

570
00:30:55,359 --> 00:30:59,240
copilot Studio has a lot of these channels or services

571
00:30:59,279 --> 00:31:05,200
already prebuilt where I don't have to redesign my experiences.

572
00:31:05,279 --> 00:31:07,480
All I do is do some back end like changes

573
00:31:07,599 --> 00:31:11,000
in the studio itself, and the same copilot then can

574
00:31:11,039 --> 00:31:13,839
be accessible on let's say SLAG or Microsoft Teams.

575
00:31:14,240 --> 00:31:17,039
Speaker 1: In both of these things, I noticed that you know,

576
00:31:17,119 --> 00:31:20,960
you can connect it to your database. Yes, And apparently

577
00:31:21,359 --> 00:31:25,920
the coolest thing is you know how many reverse threaded

578
00:31:26,000 --> 00:31:32,359
left handed hemptas are left in inventory in Atlanta, Georgia warehouse, right,

579
00:31:32,759 --> 00:31:35,200
and it'll just tell you that kind of stuff. But

580
00:31:36,400 --> 00:31:41,559
the cooler things about generative AI for me are building

581
00:31:41,720 --> 00:31:47,440
models from you know, training and unreal data. And is

582
00:31:47,480 --> 00:31:52,960
that something that is not necessarily as valuable to regular,

583
00:31:53,039 --> 00:31:55,920
run of the mill businesses as it would be to say,

584
00:31:56,119 --> 00:31:58,839
you know, a chat gpt, which is the genie in

585
00:31:58,920 --> 00:32:00,359
the bottle that knows everything.

586
00:32:00,519 --> 00:32:03,799
Speaker 3: So, if I understand your question correctly, you're alluding more

587
00:32:03,839 --> 00:32:06,119
to you know. How about I have my own data

588
00:32:06,559 --> 00:32:09,119
and I would like to use that data into generative AI.

589
00:32:10,160 --> 00:32:12,039
That's one way of doing it. Or maybe I have

590
00:32:12,119 --> 00:32:14,279
my own data and I would like to generate my

591
00:32:14,400 --> 00:32:18,200
own model using that data because I want the answers

592
00:32:18,240 --> 00:32:20,640
in a specific format or in a specific way rather

593
00:32:20,680 --> 00:32:24,000
than what you know, Chat Jupiter gives me. So there

594
00:32:24,000 --> 00:32:25,519
are a couple of things you can do. And you

595
00:32:25,559 --> 00:32:28,359
may have heard this a lot. Very few companies these

596
00:32:28,440 --> 00:32:31,960
days are investing building their own model because it's very

597
00:32:31,960 --> 00:32:35,799
hard to compete with companies like Amazon, Google, or open

598
00:32:35,839 --> 00:32:39,920
AI or even Microsoft. So most common techniques which most

599
00:32:39,960 --> 00:32:43,000
of the enterprises are using these days is either you

600
00:32:43,079 --> 00:32:47,400
go with a rack pattern like retrieval augumented generation, where

601
00:32:47,480 --> 00:32:50,480
you bring your own data and then feed that context

602
00:32:50,559 --> 00:32:54,319
or data to your large language model and then get

603
00:32:54,359 --> 00:32:57,720
the inside. Second very common is not very common, Less

604
00:32:57,720 --> 00:33:01,799
common and more expensive out is fine tuning this models

605
00:33:02,599 --> 00:33:04,839
like where as part of fine tuning, then I provide

606
00:33:04,880 --> 00:33:08,119
series of prompts some training data. So then after a

607
00:33:08,200 --> 00:33:10,759
fine tuned model is there, then it will be responding

608
00:33:10,759 --> 00:33:13,480
in a specific way, and it will be also having

609
00:33:13,519 --> 00:33:17,400
a specific information that let's say the generic large language

610
00:33:17,400 --> 00:33:19,680
model or foundation on a large language model will not have.

611
00:33:19,960 --> 00:33:22,839
Speaker 1: Now, when you say that, do you mean cleaning the data,

612
00:33:23,000 --> 00:33:25,680
like taking out the things that aren't relevant, or do

613
00:33:25,720 --> 00:33:27,720
you mean when you say fine tuning, do you mean

614
00:33:28,240 --> 00:33:30,279
telling it what to ignore in the data.

615
00:33:30,400 --> 00:33:33,759
Speaker 3: Yes, so you have to basically provide custom instructions. You

616
00:33:33,759 --> 00:33:36,000
have to provide custom prompts, and you can also provide

617
00:33:36,000 --> 00:33:38,319
your own data as well. Where okay, these are the

618
00:33:38,440 --> 00:33:40,920
sete of things that you should be doing in this

619
00:33:40,960 --> 00:33:43,440
particular way. Don't include that or don't include these, so

620
00:33:43,480 --> 00:33:46,079
those kinds of additional things you can do. One thing

621
00:33:46,119 --> 00:33:48,680
I do want to highlight is fine tuning is not

622
00:33:48,880 --> 00:33:53,119
for everyone it's only you know, for companies where you

623
00:33:53,200 --> 00:33:56,599
know you're going to have large workload or your consumption

624
00:33:56,720 --> 00:33:59,559
is going to be very high. And let's say with

625
00:33:59,640 --> 00:34:03,160
this large language model and or this large language model

626
00:34:03,200 --> 00:34:07,000
will not work in its current capacity because under the hood,

627
00:34:07,000 --> 00:34:10,880
if you read the licensing documentation from both open ai

628
00:34:11,039 --> 00:34:14,760
or even from Microsoft on Azure side, not only you

629
00:34:14,760 --> 00:34:17,159
had to pay money to find tune the model, but

630
00:34:17,400 --> 00:34:19,760
once the model is fine tuned, you had to also

631
00:34:19,760 --> 00:34:22,159
pay Microsoft money or open am money to host that

632
00:34:22,199 --> 00:34:24,679
fine tune model. And then on the top of that

633
00:34:24,840 --> 00:34:27,360
the charges for your tokens that you will be exchanging.

634
00:34:28,159 --> 00:34:30,000
Speaker 2: It sounds like rags the way to go there.

635
00:34:30,119 --> 00:34:32,400
Speaker 1: Yes, oh you know what you say that, Richard, But

636
00:34:33,039 --> 00:34:36,440
we have some mutual friends in Europe, and one in

637
00:34:36,480 --> 00:34:43,280
particular whose company decided to do away with RAG in

638
00:34:43,320 --> 00:34:47,960
their company because the percentage of errors that came back

639
00:34:48,280 --> 00:34:54,480
was too high. And I just asked this person literally

640
00:34:54,800 --> 00:34:57,719
two weeks ago, has your opinion changed? And no, no

641
00:34:57,840 --> 00:35:01,360
it hasn't. It's still still or prone. So what do

642
00:35:01,360 --> 00:35:02,079
you think about that?

643
00:35:02,320 --> 00:35:06,280
Speaker 3: So RAG, you know, on theory, it looks simple, but

644
00:35:06,400 --> 00:35:08,480
one of the hardest challenge I have seen with the

645
00:35:08,559 --> 00:35:12,519
RAG is let's say I'm looking for a specific information

646
00:35:12,639 --> 00:35:15,800
from my company data. There is a limit on how

647
00:35:15,880 --> 00:35:18,000
much of the day, like how much of a token

648
00:35:18,079 --> 00:35:20,280
size or context size you can send to a large

649
00:35:20,320 --> 00:35:24,079
language model. So most challenging part with the RAG implementation

650
00:35:24,239 --> 00:35:27,639
is how I will make sure for that particular specific

651
00:35:27,679 --> 00:35:31,039
question or conversation I'm having, how I can get the

652
00:35:31,159 --> 00:35:35,000
least amount of data that is more relevant, like how

653
00:35:35,000 --> 00:35:37,599
I can get that, And that's where you know, then

654
00:35:37,679 --> 00:35:40,039
you have to have vector databases. That's where you need

655
00:35:40,079 --> 00:35:43,079
to have your ash eye search and those are like

656
00:35:43,159 --> 00:35:46,480
configuring those and getting those right are easier said than done.

657
00:35:46,960 --> 00:35:49,760
And let's say if you're let's say, if you the

658
00:35:49,800 --> 00:35:54,039
example you gave Carl is in case they're saying less errorrates, sorry,

659
00:35:54,039 --> 00:35:56,119
more errorrates on RAG, maybe you know they will go

660
00:35:56,159 --> 00:36:00,800
with fine tuning because there they have the build capacity

661
00:36:01,159 --> 00:36:03,039
in terms of you know, fine tune the model. Because

662
00:36:03,079 --> 00:36:05,239
fine tuning is not everyone's cup of tea, you know,

663
00:36:05,320 --> 00:36:08,400
to have certain skills. And also they may be have

664
00:36:08,480 --> 00:36:11,880
a use case where the cost will justify, you know,

665
00:36:12,119 --> 00:36:16,360
aware or the cost the amount they had to put

666
00:36:16,360 --> 00:36:18,559
in the fine tuning will justify in terms of the

667
00:36:19,239 --> 00:36:20,760
errate they're getting on the right side.

668
00:36:20,960 --> 00:36:24,599
Speaker 1: Yeah, how much would it cost for the rag of

669
00:36:24,639 --> 00:36:27,000
the Franklin Brothers band to get the lead singer? Right?

670
00:36:28,079 --> 00:36:28,599
Speaker 3: I don't know.

671
00:36:33,039 --> 00:36:35,039
Speaker 1: Those are the questions you're gonna be asking, though, you know,

672
00:36:35,079 --> 00:36:39,079
we want we want accuracy because sometimes you can ask

673
00:36:39,119 --> 00:36:42,920
a question if you're if you're looking for information, right,

674
00:36:43,039 --> 00:36:47,119
real information. Typically you're asking a very specific question. Yes,

675
00:36:47,159 --> 00:36:49,880
the answer could be yes, or it could be no, yes,

676
00:36:50,400 --> 00:36:52,280
or the answer could be, you know, one of a

677
00:36:52,440 --> 00:36:57,639
hundred possibilities, and that is usually a fundamental question that

678
00:36:57,840 --> 00:37:01,119
business decisions will be made on because if it wasn't,

679
00:37:02,000 --> 00:37:03,760
what's the point of rag.

680
00:37:04,000 --> 00:37:04,280
Speaker 2: Right.

681
00:37:05,039 --> 00:37:07,519
Speaker 1: We need to ask these fundamental questions, and then you

682
00:37:07,559 --> 00:37:09,719
need to base decisions on See, we have to be really,

683
00:37:09,760 --> 00:37:12,920
really sure that that answer is correct.

684
00:37:12,960 --> 00:37:15,599
Speaker 2: And one of these challenges I remember from you know,

685
00:37:16,039 --> 00:37:19,199
decades ago, where you only have to generate a report

686
00:37:19,239 --> 00:37:22,840
once that has inaccurate numbers, and everybody questions the numbers

687
00:37:22,840 --> 00:37:25,719
forever after that. Yeah, that's right, So you know, you

688
00:37:25,840 --> 00:37:27,840
can't afford to lose that confidence.

689
00:37:28,760 --> 00:37:32,559
Speaker 3: And that's where you know, a lot of companies, especially

690
00:37:32,599 --> 00:37:35,280
after the rise of Jenny I especially the big companies

691
00:37:35,320 --> 00:37:38,719
where if they get out any wrong information or incorrect information,

692
00:37:38,800 --> 00:37:41,400
if there is like a significant financial penalty against it,

693
00:37:42,400 --> 00:37:44,599
the first thing they are doing is cleaning up their house,

694
00:37:44,679 --> 00:37:47,360
like cleaning the data. Like what if your data you

695
00:37:47,360 --> 00:37:49,760
have multiple versions of the data. One is saying good thing,

696
00:37:49,800 --> 00:37:53,239
but another saying is bad thing, and then somehow your

697
00:37:53,280 --> 00:37:56,039
AI search or whatever other two you're using picked up

698
00:37:56,039 --> 00:37:59,159
the bad information and then use that to the for

699
00:37:59,280 --> 00:37:59,960
you rag pattern.

700
00:38:00,039 --> 00:38:03,760
Speaker 1: I remember Gary Short and Seth Warez both saying that

701
00:38:04,280 --> 00:38:06,719
the majority of their job, and this was maybe ten

702
00:38:06,800 --> 00:38:11,559
years ago, was cleaning data, you know, just to get

703
00:38:11,679 --> 00:38:16,000
more accurate. And so the ability the fine tuning, as

704
00:38:16,039 --> 00:38:20,760
you say, to omit things that are outliers or standard

705
00:38:20,760 --> 00:38:22,880
deviations that are way high or something.

706
00:38:23,880 --> 00:38:26,159
Speaker 2: And we now up here in Canada we have a

707
00:38:26,199 --> 00:38:30,039
core precedent around this with Air Canada. So Air Canada

708
00:38:30,119 --> 00:38:35,800
deployed a chat bot using an LM for providing advice

709
00:38:35,920 --> 00:38:41,039
for travel, and a fellow used that tool to figure

710
00:38:41,039 --> 00:38:43,199
out if he could change a flight for it was

711
00:38:43,239 --> 00:38:46,599
a bereavement. Fares are called sover Yeah, a funeral or

712
00:38:46,639 --> 00:38:48,440
something like that. This is a great story. Yeah, And

713
00:38:48,480 --> 00:38:51,159
the software said, yeah, no problem, you can change that.

714
00:38:51,320 --> 00:38:53,599
And so then he went to actually change it and

715
00:38:53,639 --> 00:38:55,880
they said, no, you can't change that, said, well, the

716
00:38:55,920 --> 00:38:58,440
bot told me, and then he was forced to buy

717
00:38:58,480 --> 00:39:00,480
the ticket the last minute spend even more so forth.

718
00:39:00,559 --> 00:39:03,880
But he's now gone through the courts and the court ruled, hey,

719
00:39:03,960 --> 00:39:07,519
that you're using that chatbot as if it was an employee.

720
00:39:07,880 --> 00:39:12,119
You have the same liabilities to that employee. So if

721
00:39:12,119 --> 00:39:15,039
your employee gave incorrect information, you'd be held accountable for it.

722
00:39:15,199 --> 00:39:16,880
Your software, you're also held accountable.

723
00:39:16,920 --> 00:39:20,119
Speaker 1: This is a great time to be alive in the

724
00:39:20,199 --> 00:39:23,960
you know, seeing the real genesis of AI taking off

725
00:39:23,960 --> 00:39:26,159
and getting more and more accurate in the in these

726
00:39:26,239 --> 00:39:28,880
kinds of stories really shape it well.

727
00:39:28,880 --> 00:39:31,880
Speaker 2: And it's I'm happy to have those case laws in

728
00:39:31,920 --> 00:39:35,400
place because it's for me as the architects, sitting down

729
00:39:35,440 --> 00:39:37,719
with leadership saying we want to use this technologies, like

730
00:39:38,039 --> 00:39:40,840
here is the risk we are assuming, right, right, And

731
00:39:40,920 --> 00:39:44,280
so you know that's just why you spend money on testing.

732
00:39:44,679 --> 00:39:47,719
Speaker 1: And that was a yes or no answer, right, It was.

733
00:39:47,719 --> 00:39:48,960
Speaker 2: A little more complicated than.

734
00:39:48,840 --> 00:39:51,440
Speaker 1: That, Can I get a refund? Yes or something along

735
00:39:51,440 --> 00:39:53,199
those lines or something along those lines.

736
00:39:53,039 --> 00:39:54,760
Speaker 2: And yeah, you're right. It pretty much breaks down it

737
00:39:54,800 --> 00:39:55,599
exactly that, Carl.

738
00:39:55,679 --> 00:39:57,840
Speaker 3: Yeah, and that's the reason. Like some of the customers

739
00:39:57,840 --> 00:40:00,559
I work with, they are pretty large customers, and one

740
00:40:00,599 --> 00:40:03,760
thing they're doing is hey, EI is great, but nobody

741
00:40:03,760 --> 00:40:06,800
should be launching anything because it going through some governance

742
00:40:06,800 --> 00:40:10,639
committee where they will be evaluating what exactly you're doing.

743
00:40:11,039 --> 00:40:13,679
You know, what technology you're using, whether it's going to

744
00:40:13,679 --> 00:40:17,320
be for internal users or for your customers, because if

745
00:40:17,320 --> 00:40:19,639
you do any mistake, and let's say if it's for

746
00:40:19,760 --> 00:40:22,840
your outside customers, you know, the implication of that can

747
00:40:22,880 --> 00:40:23,559
be tremendous.

748
00:40:23,800 --> 00:40:27,599
Speaker 2: Yes, now, definitely the motion at this moment here we are,

749
00:40:27,719 --> 00:40:29,719
you know to a latter half of twenty twenty four

750
00:40:31,000 --> 00:40:34,920
is all internal apps. It's an hr bot, Yes right.

751
00:40:34,920 --> 00:40:37,920
It's those kinds of things where you're only upsetting the

752
00:40:37,960 --> 00:40:40,079
employee when you lie to them about how many vacation

753
00:40:40,159 --> 00:40:40,679
days they have.

754
00:40:41,320 --> 00:40:43,960
Speaker 3: And again that's a good point. You got what, Richard,

755
00:40:44,000 --> 00:40:46,840
because vacation days is something that you shouldn't be relying

756
00:40:46,880 --> 00:40:49,239
on general to AI. You tell you, it should be

757
00:40:49,280 --> 00:40:53,039
coming straight from your employee handbook, where where you know

758
00:40:53,079 --> 00:40:55,599
you have full control over what information is getting from

759
00:40:55,599 --> 00:40:57,559
there and what information is surfaced.

760
00:40:57,199 --> 00:40:59,599
Speaker 2: If this isn't a project that makes sense Pshan, what

761
00:40:59,679 --> 00:41:02,760
prout checks do like that seems like exactly, I'm not

762
00:41:02,800 --> 00:41:05,280
going to read the handbook. Yes, I'm going to pester

763
00:41:05,559 --> 00:41:07,840
HR for an answer, yes, and they're going to yell

764
00:41:07,880 --> 00:41:09,960
at me. So we make the bot to solve that

765
00:41:10,119 --> 00:41:11,719
problem and you.

766
00:41:11,679 --> 00:41:14,519
Speaker 3: Can solve it easily. Now with Copilot Studio along with

767
00:41:14,599 --> 00:41:18,159
Azure Ai Studio as well, but co Pilot Studio, a

768
00:41:18,159 --> 00:41:21,119
lot of companies are also using it for customer centric

769
00:41:21,159 --> 00:41:25,039
BOTE outward bots like Business to Customer as well. And

770
00:41:25,159 --> 00:41:28,320
one of the reasons they are doing it because some

771
00:41:28,360 --> 00:41:32,440
of the red teaming and other stuff the co Pilot

772
00:41:32,480 --> 00:41:35,880
Studio team has done because not peny people understand when

773
00:41:35,880 --> 00:41:39,119
you launch anything external, especially at childbot like Microsoft learn

774
00:41:39,199 --> 00:41:42,320
It in twenty sixteen, where people will try to get

775
00:41:42,320 --> 00:41:44,719
some nasty stuff from it. And if you launch a

776
00:41:44,719 --> 00:41:47,880
public facing website chadbot, you know, anyone can start interacting

777
00:41:47,880 --> 00:41:51,159
with it and start asking some stuff that they shouldn't

778
00:41:51,159 --> 00:41:53,559
be asking or your company shouldn't be giving responsor to that.

779
00:41:53,800 --> 00:41:56,760
Speaker 2: So business bring us back to outward facing is not

780
00:41:57,000 --> 00:41:58,519
there yet. It's too high.

781
00:41:58,400 --> 00:42:01,400
Speaker 3: Risk y's high risk. Yes, but if you know, if

782
00:42:01,400 --> 00:42:03,599
you use the right tools, or if you know you

783
00:42:03,639 --> 00:42:06,159
know you have a good team in house, you can

784
00:42:06,199 --> 00:42:08,960
definitely use some of these tools where you can have

785
00:42:09,000 --> 00:42:11,920
a more confidence, you know, launching this outward facing application

786
00:42:12,000 --> 00:42:12,320
as well.

787
00:42:12,320 --> 00:42:14,559
Speaker 2: I just debate that anyone has a good team at

788
00:42:14,559 --> 00:42:15,079
this point.

789
00:42:15,400 --> 00:42:18,440
Speaker 1: What are the things that I think would help And

790
00:42:18,519 --> 00:42:23,280
it's definitely almost impossible with large language models that are

791
00:42:23,320 --> 00:42:26,000
trained on language right and try to predict the next

792
00:42:26,039 --> 00:42:30,559
word in their answer, but citing references. Right, if that

793
00:42:32,239 --> 00:42:35,599
Air Canada bot, you know, when it said yes you

794
00:42:35,639 --> 00:42:38,119
can get a refund and here's a link to the

795
00:42:38,159 --> 00:42:43,280
policy that you can read and decide for yourself, right,

796
00:42:44,800 --> 00:42:48,199
then it might have been a bit more accurate, and

797
00:42:48,400 --> 00:42:51,079
not not necessarily accurate, but it would have allowed the

798
00:42:51,199 --> 00:42:53,960
user of that bot to make a more informed decision.

799
00:42:54,079 --> 00:42:54,320
Speaker 2: Yep.

800
00:42:54,480 --> 00:42:58,199
Speaker 3: And I think KYL, like Microsoft co Pilot studior team

801
00:42:58,239 --> 00:43:01,360
should hire KRL because now he's letting all their good

802
00:43:01,360 --> 00:43:04,880
features indirectly via all these success stories. But yes, that's

803
00:43:04,880 --> 00:43:08,559
one of the advantage where this co pilor studio giates

804
00:43:08,639 --> 00:43:11,119
is if you are going with the rat pattern and

805
00:43:11,199 --> 00:43:13,719
let's say you're using your knowledge based then while it

806
00:43:13,800 --> 00:43:18,360
generate the information, it does give you information about the citation,

807
00:43:18,519 --> 00:43:21,480
like what are the source of this information and so

808
00:43:21,519 --> 00:43:23,079
that you can click on it and then you can

809
00:43:23,199 --> 00:43:25,599
verify Okay, this information looks great or I can then

810
00:43:25,639 --> 00:43:28,400
you know, use it with more confidence and then I'll

811
00:43:28,440 --> 00:43:29,480
go on with the next stuff.

812
00:43:29,559 --> 00:43:33,159
Speaker 1: That's great. It's so good to hear. Yeah, ask chat

813
00:43:33,199 --> 00:43:38,400
GPT for a reference. Sometimes get that and on lessly.

814
00:43:38,480 --> 00:43:40,800
It doesn't know the way, it's bunchet. It's a very

815
00:43:40,840 --> 00:43:42,079
difficult thing to deal with. Yah.

816
00:43:43,280 --> 00:43:45,400
Speaker 2: Yeah, I think this is where they it's just where

817
00:43:45,400 --> 00:43:47,480
they use the term grounding. I don't know why this

818
00:43:47,559 --> 00:43:48,639
term was necessary.

819
00:43:48,880 --> 00:43:51,199
Speaker 3: Yeah, they use the grounding. Yes, Like, okay, we use

820
00:43:51,239 --> 00:43:54,159
this information from grounding. It's not like something the lllum

821
00:43:54,199 --> 00:43:56,800
made up and this is the source of it or

822
00:43:56,840 --> 00:43:57,320
proof of it.

823
00:43:57,400 --> 00:44:00,840
Speaker 2: Can we just always be grounded like thing? Right?

824
00:44:00,920 --> 00:44:02,960
Speaker 3: Like really like the people who have kids, you know,

825
00:44:03,159 --> 00:44:05,239
grounding me is totally different in their context.

826
00:44:05,280 --> 00:44:07,440
Speaker 1: Yeah right, true enough. Or the people who like to

827
00:44:07,440 --> 00:44:09,599
walk barefoot out in the lawn every day because it

828
00:44:09,639 --> 00:44:12,079
helps their magnetism or something.

829
00:44:12,960 --> 00:44:15,199
Speaker 2: Walking barefoot and grass is pretty good. Man. You don't

830
00:44:15,199 --> 00:44:16,159
have nice.

831
00:44:16,719 --> 00:44:19,880
Speaker 1: It's nice. It's a nice feeling, but it's all it

832
00:44:19,920 --> 00:44:23,559
needs to be. There's some weird science that is associated

833
00:44:23,559 --> 00:44:28,639
with it would say not soccer, but okay, pseudoscience anyway.

834
00:44:29,239 --> 00:44:34,440
Speaker 2: Now I appreciate that. I don't want software speculating on

835
00:44:34,480 --> 00:44:37,559
my van. I prefer it to say I don't know.

836
00:44:37,760 --> 00:44:40,400
Speaker 3: Yes, and those kinds of things you can configure in

837
00:44:40,440 --> 00:44:43,519
co pilot studio, like you can have Okay, I'm launching

838
00:44:43,519 --> 00:44:45,920
this copilot and it's going to take care of only

839
00:44:45,960 --> 00:44:48,960
four or five things. Anything extra, people like, anyone ask

840
00:44:49,320 --> 00:44:50,920
it can just come back with I don't know or

841
00:44:51,119 --> 00:44:53,119
I'm not allowed to talk about that. Go through this

842
00:44:53,239 --> 00:44:55,360
link or talk to this person and then and do this.

843
00:44:56,039 --> 00:44:59,840
Speaker 1: In chat GPT's defense, it does that, you know, when

844
00:44:59,840 --> 00:45:02,199
it doesn't have an answer, it says you should call

845
00:45:02,239 --> 00:45:05,920
the manufacturer or you know, contact the whatever the authority.

846
00:45:06,159 --> 00:45:10,280
Speaker 3: One thing people don't understand is chigibt is a custom experience,

847
00:45:10,400 --> 00:45:13,639
Like you're not interacting with an LM directly like as

848
00:45:13,679 --> 00:45:17,239
an API. It's a customer application. A lot of guard rails,

849
00:45:17,320 --> 00:45:20,639
like Opena has put into it. Now they're launching a

850
00:45:20,679 --> 00:45:22,519
memory as well, so a lot of things are going.

851
00:45:22,559 --> 00:45:24,599
It's a custom app, so yes, if you have a

852
00:45:24,599 --> 00:45:27,000
team that can build a custom app like that, then

853
00:45:27,079 --> 00:45:29,119
yes you will get similar answer. But let's say if

854
00:45:29,119 --> 00:45:32,159
I'm a developer, I'm calling this LLM directly. You know

855
00:45:32,320 --> 00:45:34,920
most of this LM don't have a memory as well,

856
00:45:34,960 --> 00:45:37,719
Like the calls are stateless, so those kinds of things

857
00:45:37,760 --> 00:45:39,039
it may not say, and it may just give you

858
00:45:39,159 --> 00:45:41,840
completely like completely incorrect answer as well.

859
00:45:41,920 --> 00:45:44,400
Speaker 2: Yeah, and I guess that that brings me back to

860
00:45:44,440 --> 00:45:46,519
this whole. Like the HR bot makes sense is if

861
00:45:46,559 --> 00:45:48,800
you can ground it on. I know you're not going

862
00:45:48,840 --> 00:45:51,719
to read the user they employee hand books, so I'm

863
00:45:51,719 --> 00:45:53,639
going to turn it into the ability for you to

864
00:45:53,679 --> 00:45:56,519
ask questions and it's going to reference accurately. And when

865
00:45:56,559 --> 00:45:59,920
I'm not when my certainty is low in the SAW

866
00:46:00,159 --> 00:46:02,840
where let software certainty low, it's like you should talk

867
00:46:02,880 --> 00:46:05,000
to HR about that. Because from an HR perspective, if

868
00:46:05,000 --> 00:46:07,199
I can take eighty percent of the stupid questions and

869
00:46:07,320 --> 00:46:10,000
just answer them so I'm only dealing with the twenty

870
00:46:10,039 --> 00:46:12,599
percent that are more complex, I'm pretty happy. Like that's

871
00:46:12,639 --> 00:46:13,320
a good outcome.

872
00:46:13,360 --> 00:46:15,039
Speaker 3: You can use for HR, you can use it for

873
00:46:15,079 --> 00:46:17,840
your IT support as well, because it has connectors where

874
00:46:18,239 --> 00:46:20,320
it can talk to your own knowledge base and that

875
00:46:20,360 --> 00:46:21,639
knowledge base can be in service.

876
00:46:21,719 --> 00:46:24,199
Speaker 2: Now yeah, no, I just put I have an automated

877
00:46:24,199 --> 00:46:26,000
responder just says you try turning it off and turning

878
00:46:26,000 --> 00:46:26,599
it back on again.

879
00:46:26,599 --> 00:46:32,000
Speaker 3: Like said, yeah, my favorite code is be smart restart.

880
00:46:32,159 --> 00:46:36,679
Speaker 1: You know there you go. Nice? Nice works for phones too. Yep, yes,

881
00:46:37,239 --> 00:46:41,239
all the time. And Pat and Dwayne Patrick Kines Douain

882
00:46:41,320 --> 00:46:44,400
Laflat from security this week say you should be restarting

883
00:46:44,400 --> 00:46:49,039
your phone at least once a week at least if not,

884
00:46:49,280 --> 00:46:53,320
you know, three or four times a week. Yeah, all right,

885
00:46:53,400 --> 00:46:55,519
So where do we go from here? Is there more

886
00:46:55,639 --> 00:46:58,639
that we haven't discussed? Like how is it expensive? How

887
00:46:58,639 --> 00:46:59,440
do we get started?

888
00:46:59,519 --> 00:47:03,559
Speaker 3: Okay, so cost wise, even though it's a part of

889
00:47:03,639 --> 00:47:07,159
Power Platform, its cost is a little bit different. It's

890
00:47:07,239 --> 00:47:10,920
just like ash. Everything is based on the consumption. So

891
00:47:10,960 --> 00:47:13,840
the minimum pricing point is you had to bot. You

892
00:47:13,880 --> 00:47:17,400
had to buy a package for twenty five thousand messages,

893
00:47:17,679 --> 00:47:21,360
which cost two hundred dollars. And again this is a

894
00:47:21,400 --> 00:47:24,039
retail price, like if you have an enterprise agreement, you

895
00:47:24,039 --> 00:47:26,679
may be getting some discount on that, but that brings

896
00:47:26,760 --> 00:47:28,840
up a question is what exactly a message is? Right?

897
00:47:28,960 --> 00:47:33,039
A message is basically anything like when I ask anything

898
00:47:33,519 --> 00:47:36,320
to my copilot, if that's going to trigger any kind

899
00:47:36,320 --> 00:47:39,320
of response or any kind of AI agent a call

900
00:47:39,400 --> 00:47:42,880
behind the scene, that constitute as a one message. And

901
00:47:42,920 --> 00:47:46,880
if I'm using generative AI features, then just for the

902
00:47:46,920 --> 00:47:49,880
sake of simplicity, what Microsoft has done is any answer

903
00:47:49,920 --> 00:47:51,880
that has generative AI in it, it is just going

904
00:47:51,880 --> 00:47:54,480
to count as the two messages. So if I ask

905
00:47:54,519 --> 00:47:57,599
a question you know how many PTOs are there, and

906
00:47:57,639 --> 00:48:00,519
I'm using some of the generative EI features and then

907
00:48:00,559 --> 00:48:03,519
that come back, then it just counts as the two messages.

908
00:48:03,719 --> 00:48:06,079
Speaker 1: Now it's two hundred dollars a month for twenty five

909
00:48:06,119 --> 00:48:08,679
thousand messages a month. Does that mean that if I

910
00:48:08,719 --> 00:48:12,440
only use one thousand messages or you know, less than that,

911
00:48:12,559 --> 00:48:14,639
am I only going to get charge point zero zero

912
00:48:14,760 --> 00:48:16,599
eight cents per message?

913
00:48:16,719 --> 00:48:19,719
Speaker 3: No? No, you have to, Like the minimum is two hundred.

914
00:48:19,760 --> 00:48:21,239
Speaker 1: That's what you have, whether you use it or not.

915
00:48:21,480 --> 00:48:26,320
Speaker 3: Yes, And that's one story which is really not good

916
00:48:26,679 --> 00:48:30,519
in my opinion right now, because you buy that license

917
00:48:30,519 --> 00:48:33,639
for per tenant. Okay, but what if I'm working for

918
00:48:33,719 --> 00:48:36,480
a really big organization where I have thousands and thousands

919
00:48:36,480 --> 00:48:40,440
of employee and ten departments are there. Every department has

920
00:48:40,440 --> 00:48:42,880
their own copilot and how exactly we're going to manage

921
00:48:42,880 --> 00:48:46,239
the pricing because those twenty five thousand message I can

922
00:48:46,280 --> 00:48:47,960
be exhausting in one hour or two hours.

923
00:48:48,039 --> 00:48:51,199
Speaker 2: Yeah sure, yeah, yeah, it feels like it's beta pricing,

924
00:48:51,360 --> 00:48:54,000
Like it's just pricing because you're doing a beta test project,

925
00:48:54,039 --> 00:48:57,920
not actually deployment level yet. And so now I'm starting

926
00:48:57,920 --> 00:49:01,920
to feel like the relationship between copile studio and as

927
00:49:01,960 --> 00:49:05,119
your AI studio is the same relationship between power platform

928
00:49:05,239 --> 00:49:06,639
and visual studio.

929
00:49:06,360 --> 00:49:08,960
Speaker 1: Correct, right, Yeah, that makes sense to me, Richard.

930
00:49:09,159 --> 00:49:12,880
Speaker 2: So if you're the front end stuff, the place to experiment,

931
00:49:13,280 --> 00:49:16,760
the price to prototype copilot studio when you want to

932
00:49:16,800 --> 00:49:21,079
get into precision and tuning and more skilled people AI

933
00:49:21,199 --> 00:49:23,960
studio knowing that anything you build an AI studio you

934
00:49:24,039 --> 00:49:27,920
can surface in copilot. Yeah, and all right, I dig this.

935
00:49:27,920 --> 00:49:30,400
This is huge in development. Like you said, we did

936
00:49:30,400 --> 00:49:34,000
that show with vishuas earlier this year where we talked

937
00:49:34,000 --> 00:49:37,400
about that same dynamic between power platform and dot net right.

938
00:49:37,239 --> 00:49:39,440
Speaker 3: That the and I know yeah, which Fus is a

939
00:49:39,519 --> 00:49:42,719
huge advocate of future development. In fact, he's the one

940
00:49:42,719 --> 00:49:45,119
who taught me all this concept, and I agree with

941
00:49:45,159 --> 00:49:48,280
it because the more and more I work with enterprise customers,

942
00:49:48,480 --> 00:49:52,840
like there's always the case where developers, like good developers

943
00:49:52,880 --> 00:49:56,320
are a rare commodity. Yeah, and you'll agree that, you know,

944
00:49:56,440 --> 00:49:58,599
it's very hard to find good developers, Like how about

945
00:49:58,639 --> 00:50:02,000
then we create some tools where you need developers only

946
00:50:02,039 --> 00:50:04,519
when that tool cannot do certain things for you, right.

947
00:50:04,559 --> 00:50:07,119
Speaker 2: Right, But the other part is like I never look

948
00:50:07,159 --> 00:50:09,000
at the power platform folks, and I guess it's the

949
00:50:09,039 --> 00:50:12,079
copilot studio folks. As these are the lessers, these are

950
00:50:12,119 --> 00:50:15,159
the domain experts. They're actually going to save you time

951
00:50:15,519 --> 00:50:19,000
because they know more about the problem space than you do. Right, Like,

952
00:50:19,239 --> 00:50:22,119
my struggle with really great developers is how much time

953
00:50:22,159 --> 00:50:25,199
they spend having to learn the domain. Where if you

954
00:50:25,320 --> 00:50:28,800
have a good domain expert relationship going there, you can

955
00:50:28,880 --> 00:50:31,039
focus more on breaking those bits down and go, okay,

956
00:50:31,039 --> 00:50:33,559
well that's where some custom code needs to live. That's

957
00:50:33,599 --> 00:50:35,719
there is no easy solution there. We need to build something,

958
00:50:35,800 --> 00:50:37,840
take a spike, you know, those kinds of things. Like

959
00:50:37,920 --> 00:50:41,119
that's a really constructive way to work quickly.

960
00:50:41,320 --> 00:50:44,960
Speaker 3: And I'm seeing like so I'm talking about this conversation

961
00:50:45,039 --> 00:50:49,280
AI or Microsoft Board framework since twenty seventeen and since

962
00:50:49,400 --> 00:50:52,960
last Ignite. I'm saying like just a huge interest from

963
00:50:52,960 --> 00:50:56,599
developers as well, like pro developers thinking about or are

964
00:50:56,679 --> 00:50:59,599
eager to learn about co pilot studio. Like a couple

965
00:50:59,639 --> 00:51:02,159
of weeks before, I was at dev Intersection or next

966
00:51:02,239 --> 00:51:05,800
Gen EI conference in Vegas and I did session and

967
00:51:05,840 --> 00:51:10,280
a workshop on copilot Studio and its attendance was more

968
00:51:10,360 --> 00:51:12,840
than some of the popular you know, pro developer sessions

969
00:51:12,960 --> 00:51:14,760
or confront or workshops were there.

970
00:51:15,920 --> 00:51:20,360
Speaker 2: Yeah, that's great, but any Yeah, I appreciate this because

971
00:51:20,360 --> 00:51:22,920
you start now, you start thinking about rethinking the UX

972
00:51:22,960 --> 00:51:26,880
of an application, a lot less buttons and sliders and

973
00:51:26,960 --> 00:51:30,639
a lot more texts and most of the screen taken

974
00:51:30,719 --> 00:51:35,599
up with it presenting information to you based on your requests.

975
00:51:35,679 --> 00:51:37,599
Speaker 3: Right, And it doesn't have to be only this. It

976
00:51:37,639 --> 00:51:40,239
can be. Yes, for some things you'll be using this,

977
00:51:40,440 --> 00:51:44,440
but for some more traditional or more things where you

978
00:51:44,519 --> 00:51:47,119
definite need to have you like proper UI with multiple

979
00:51:47,119 --> 00:51:49,920
buttons and multiple sliders, yes you'll have that, but this

980
00:51:50,000 --> 00:51:52,559
can be your beginning point where I'm every day I'm

981
00:51:52,559 --> 00:51:54,960
starting my work. I'm just coming here, you know, doing

982
00:51:55,000 --> 00:51:57,159
my stuff. I asking copilot to do a bunch of

983
00:51:57,199 --> 00:52:00,559
stuff for me. And I always give the example, you know,

984
00:52:00,639 --> 00:52:04,400
Javis in Iron Man. That's a good example. Hey, there

985
00:52:04,400 --> 00:52:07,039
are certain things if Javis can do. Let Javis handle that.

986
00:52:07,119 --> 00:52:09,400
And I'm going to focus more on the complicated stuff

987
00:52:09,400 --> 00:52:10,039
behind the scene.

988
00:52:11,559 --> 00:52:16,840
Speaker 2: Yeah, I'm waiting for office to bet on this tap. Right,

989
00:52:16,920 --> 00:52:19,559
Like you think about immediately think in terms of the

990
00:52:19,599 --> 00:52:22,599
old Alan Cooper of the Sovereign app. Right, what's the

991
00:52:22,639 --> 00:52:25,719
first app you go to? Now? For me? It's still Outlook,

992
00:52:26,079 --> 00:52:27,239
you know. That's why I'm sad.

993
00:52:29,880 --> 00:52:32,639
Speaker 1: Six five threads one for me.

994
00:52:35,559 --> 00:52:37,639
Speaker 2: There's also teams, like a lot of folks, that's their

995
00:52:37,679 --> 00:52:41,840
starting point. And you start thinking about the LLM interface

996
00:52:41,880 --> 00:52:44,119
to this, how you would start your day? I mean

997
00:52:44,599 --> 00:52:48,559
right away, your first step in either app is triage. Yes, right,

998
00:52:48,639 --> 00:52:51,880
what's the most important thing? Right? What do I need

999
00:52:51,920 --> 00:52:52,400
to deal with?

1000
00:52:52,480 --> 00:52:52,760
Speaker 1: Now?

1001
00:52:53,079 --> 00:52:54,800
Speaker 2: You know, where are the crises and so forth? Like

1002
00:52:54,880 --> 00:52:58,719
you could see a good agentic type tooling in this,

1003
00:52:59,400 --> 00:53:02,360
helping this or through those priorities. I mean, I would

1004
00:53:02,440 --> 00:53:04,800
argue the first triage is how many of these has

1005
00:53:04,840 --> 00:53:06,639
the bot already responded to?

1006
00:53:07,000 --> 00:53:07,159
Speaker 3: MM?

1007
00:53:07,239 --> 00:53:10,079
Speaker 2: Hmm right, I've not only sorted these out. Here's the

1008
00:53:10,079 --> 00:53:12,039
ones I think. Here the answers are, are you okay

1009
00:53:12,079 --> 00:53:14,280
with this? It's like yeah, yes, and that that that

1010
00:53:14,280 --> 00:53:16,000
that that that okay? What are the ones I need

1011
00:53:16,039 --> 00:53:18,800
a trioge like my wife could my life could be

1012
00:53:18,840 --> 00:53:19,320
way better?

1013
00:53:19,519 --> 00:53:24,119
Speaker 3: Yeah. I think that story is still evolving with Copilot

1014
00:53:24,159 --> 00:53:27,960
for Microsoft sixty five. Uh, and they do have one

1015
00:53:28,840 --> 00:53:31,440
capability there where in case you would like to extend

1016
00:53:31,440 --> 00:53:33,639
the functionality of out of the box co pilot there

1017
00:53:34,039 --> 00:53:37,039
you can use Copilot Studio and its license is already

1018
00:53:37,079 --> 00:53:41,119
covered as part of your Microsoft Pay sixty five Copilot license,

1019
00:53:41,719 --> 00:53:43,039
so I can listen.

1020
00:53:43,159 --> 00:53:46,880
Speaker 2: Prashan, I'm an old grizzled Microsoft observer. At this point,

1021
00:53:47,320 --> 00:53:52,280
you're not grizzled. I've come to appreciate disper he said

1022
00:53:52,320 --> 00:53:56,480
on a podcast. You will know a technology is being

1023
00:53:56,519 --> 00:54:00,239
taken serious by Microsoft when they bet their mark key

1024
00:54:00,280 --> 00:54:06,760
products on it. Yes, right, when that shows up in Outlook. Okay,

1025
00:54:06,760 --> 00:54:10,599
they're committed, like, no two ways about it. That's why

1026
00:54:10,639 --> 00:54:13,920
we bought into Active Acts or you know that whole

1027
00:54:13,960 --> 00:54:17,760
control surface. That's why we bought into Ola, and it's

1028
00:54:17,760 --> 00:54:22,119
also why we were suspect of WPF when Office never

1029
00:54:22,199 --> 00:54:25,360
took on WPF. It wasn't until Visual Studio took out WPF.

1030
00:54:25,400 --> 00:54:29,239
They're like, okay, right, Like that's there's a reason to

1031
00:54:29,280 --> 00:54:31,599
take the cues this way as to how far down

1032
00:54:31,639 --> 00:54:34,199
the path these products really are. Like I feel like

1033
00:54:34,239 --> 00:54:38,239
these are still experimental products because the Microsoft seems to

1034
00:54:38,280 --> 00:54:40,360
feel these things are still expound products.

1035
00:54:40,360 --> 00:54:42,320
Speaker 1: Well, they may be working on them, but they haven't

1036
00:54:42,320 --> 00:54:45,440
surfaced yet. Yeah, I have a question for you, totally

1037
00:54:45,480 --> 00:54:49,119
off topic, well not really. Apple's coming out very soon

1038
00:54:49,239 --> 00:54:51,880
with probably out by now, the new iPhone that has

1039
00:54:52,079 --> 00:54:59,360
AI Apple Intelligence not artificial at all, and I kind

1040
00:54:59,360 --> 00:55:04,239
of I feel like I get it. They're basically taking

1041
00:55:04,320 --> 00:55:09,559
the SII experience and exposing you know, GPT four oh

1042
00:55:09,679 --> 00:55:13,840
on the back end so that Siri will become more intelligent.

1043
00:55:14,880 --> 00:55:17,280
And I just wonder if you know anything about that,

1044
00:55:17,400 --> 00:55:20,400
because I, you know, if that is as easy as

1045
00:55:20,400 --> 00:55:22,239
it sounds. Man, I'm going to get in my car,

1046
00:55:22,320 --> 00:55:25,119
turn on car Play, and just start talking to my

1047
00:55:25,280 --> 00:55:27,440
car and it's going to be night rider.

1048
00:55:28,599 --> 00:55:30,880
Speaker 2: You know, what do you think for Shadow? I have thoughts.

1049
00:55:31,199 --> 00:55:33,760
Speaker 3: So I'm not an Apple person. I use most of

1050
00:55:33,800 --> 00:55:37,400
the Android stuff. But after saying, like some of the

1051
00:55:37,440 --> 00:55:42,800
demos from open Ei recently, the demo ware, the latest

1052
00:55:42,880 --> 00:55:45,280
version of GPT four oh is multimodel, and then we

1053
00:55:45,360 --> 00:55:48,159
can have multiple conversation and we can also change the

1054
00:55:48,199 --> 00:55:51,519
context in between. I think we are getting there as

1055
00:55:51,559 --> 00:55:55,119
long as they figure out how to run this LLLMS

1056
00:55:55,239 --> 00:55:58,039
or all these models cheaply on the devices. Yeah, I

1057
00:55:58,039 --> 00:56:00,400
think we are there, like maybe one or two years.

1058
00:56:00,599 --> 00:56:03,239
Once they put some SA safeguards in it. And let's

1059
00:56:03,239 --> 00:56:05,519
say I'm driving a car and if that LM can

1060
00:56:05,599 --> 00:56:08,239
execute on my car's hardware, we should be there.

1061
00:56:08,559 --> 00:56:12,199
Speaker 1: Are you saying that the LLLM stuff is on the phone,

1062
00:56:12,239 --> 00:56:14,960
they're not just calling an API out to GPT to

1063
00:56:16,159 --> 00:56:16,840
do the work.

1064
00:56:17,039 --> 00:56:20,000
Speaker 3: Yeah, Like once that gets mainstream, like where we will

1065
00:56:20,000 --> 00:56:22,920
have more and more this small language model is doing

1066
00:56:22,920 --> 00:56:26,360
all this stuff, but they're actually executing on the local stuff,

1067
00:56:26,440 --> 00:56:28,920
we'll have like much better experience, and we'll also see

1068
00:56:28,960 --> 00:56:32,079
more and more push from these companies as well, because.

1069
00:56:32,079 --> 00:56:34,519
Speaker 1: I don't know if that's the case now. It seems

1070
00:56:34,559 --> 00:56:37,239
to me that they're just calling out too with an API,

1071
00:56:37,360 --> 00:56:38,000
but that.

1072
00:56:38,159 --> 00:56:39,920
Speaker 3: Is a very expensive right now, Like if you look

1073
00:56:39,960 --> 00:56:44,440
at yes, they can do it for a few months

1074
00:56:44,480 --> 00:56:47,679
or a few years, but that's not really long sustainable

1075
00:56:48,239 --> 00:56:52,800
business model because all these lllms require really bf infrastructure.

1076
00:56:53,599 --> 00:56:56,679
That's the reason why I think. I don't please don't

1077
00:56:56,760 --> 00:56:58,840
quote me on this, but I heard even though openan

1078
00:56:58,880 --> 00:57:01,119
AA is making so much money, like billions of dollars,

1079
00:57:01,639 --> 00:57:03,800
they're still not profitable because they had to spend a

1080
00:57:03,800 --> 00:57:06,559
lot of money behind the scene on the infrastructure to support.

1081
00:57:06,159 --> 00:57:07,679
Speaker 2: That far and far more.

1082
00:57:07,920 --> 00:57:09,639
Speaker 1: Yes, what was your thoughts, Richard?

1083
00:57:09,760 --> 00:57:11,880
Speaker 2: I think Apple is late to the AI game and

1084
00:57:11,920 --> 00:57:13,880
needed to announce something, and the fact that they keep

1085
00:57:13,920 --> 00:57:15,639
pushing back the release of it is proof they don't

1086
00:57:15,639 --> 00:57:19,800
know what they just committed to. Ooh, and they're panicking.

1087
00:57:20,960 --> 00:57:26,280
You know, we expect Apple to come late with a

1088
00:57:26,320 --> 00:57:29,480
refined product that has clearly not happened. The fact that

1089
00:57:29,480 --> 00:57:34,119
they've already pushed the dates back shows they they said

1090
00:57:34,119 --> 00:57:36,960
they were going to jump before they jumped, and now

1091
00:57:37,079 --> 00:57:40,519
they're struggling because it is a very unrefined product. It

1092
00:57:40,599 --> 00:57:43,119
is not asked Apple asque at all, but they could

1093
00:57:43,159 --> 00:57:45,920
not afford to not say something good way to hear,

1094
00:57:46,280 --> 00:57:48,039
so they had You know, they only do one show

1095
00:57:48,039 --> 00:57:50,559
a year. They do that show in the WWDCS in June.

1096
00:57:51,039 --> 00:57:52,679
Were you really going to wait till twenty twenty five?

1097
00:57:52,760 --> 00:57:55,440
You couldn't, so you said it, and now you're trying

1098
00:57:55,440 --> 00:57:58,360
to make it come true. That's not very apple asque,

1099
00:57:58,559 --> 00:58:01,360
but all the hallmarks are there. The only reason we

1100
00:58:01,440 --> 00:58:03,760
might give a pass is because Zappli you don't do that.

1101
00:58:04,199 --> 00:58:05,400
It looks like they did this time.

1102
00:58:05,639 --> 00:58:08,320
Speaker 1: Well that feels like the end of our show, Prashant.

1103
00:58:08,400 --> 00:58:12,280
This has been very enlightening and I appreciate you putting

1104
00:58:12,360 --> 00:58:16,840
up with my digressions and you know, little voyages into

1105
00:58:16,960 --> 00:58:20,199
stories and things. But man, this is just chocolate block

1106
00:58:20,280 --> 00:58:22,360
full of great meaty goodness.

1107
00:58:22,519 --> 00:58:24,920
Speaker 3: So thank you, thank you, thanks for having me. And

1108
00:58:25,239 --> 00:58:28,559
it was really it was my pleasure, you know, pleasure

1109
00:58:28,599 --> 00:58:29,920
on all of my side for.

1110
00:58:30,000 --> 00:58:33,800
Speaker 1: This fantastic all right, and we'll talk to you next

1111
00:58:33,840 --> 00:58:57,840
time on dot net rocks. Dot net Rocks is brought

1112
00:58:57,840 --> 00:59:01,280
to you by Franklin's Net and produce by Pop Studios,

1113
00:59:01,679 --> 00:59:05,719
a full service audio, video and post production facility located

1114
00:59:05,719 --> 00:59:08,639
physically in New London, Connecticut, and of course in the

1115
00:59:08,719 --> 00:59:12,599
cloud online at pwop dot com.

1116
00:59:12,800 --> 00:59:14,920
Speaker 4: Visit our website at d O T N E t

1117
00:59:15,159 --> 00:59:19,199
R O c k s dot com for RSS feeds, downloads,

1118
00:59:19,320 --> 00:59:22,960
mobile apps, comments, and access to the full archives going

1119
00:59:23,039 --> 00:59:26,280
back to show number one, recorded in September two.

1120
00:59:26,119 --> 00:59:29,119
Speaker 1: Thousand and two. And make sure you check out our sponsors.

1121
00:59:29,280 --> 00:59:32,320
They keep us in business. Now go write some code,

1122
00:59:32,639 --> 00:59:33,360
See you next time.

1123
00:59:34,320 --> 00:59:36,079
Speaker 3: You got Jamal Vans

1124
00:59:38,199 --> 00:59:42,920
Speaker 2: And