Unlocking AI Innovations in Construction with Marcus Turner

1
00:00:00,677 --> 00:00:06,957
One of my most favorite quotes that a guest said was, is that,

2
00:00:06,957 --> 00:00:11,197
that basically construction people come together as strangers and leave as enemies.

3
00:00:18,637 --> 00:00:23,317
And it's a pretty strong statement. And, you know, we, I would like,

4
00:00:23,337 --> 00:00:24,337
I love to see less of that.

5
00:00:24,417 --> 00:00:29,077
And I would love to see people kind of working on processes and projects and things like that.

6
00:00:30,017 --> 00:00:33,217
Because, I mean, we don't want to just go to work and just know it's a frustrating

7
00:00:33,217 --> 00:00:36,277
experience. We kind of want to go to work and have some relief.

8
00:00:37,137 --> 00:00:42,177
And I think that it would be, you know, construction is sometimes a very long

9
00:00:42,177 --> 00:00:44,957
process because it's going to be years being with these people.

10
00:00:45,157 --> 00:00:49,697
So I think having more of a collaborative environment would be beneficial.

11
00:00:50,277 --> 00:00:55,277
Welcome to the EBFC show, the easier, better for construction podcast.

12
00:00:55,517 --> 00:01:00,717
I'm your host, Felipe Engineer Manriquez. this show is all about the business of construction.

13
00:01:01,637 --> 00:01:05,057
Today's show is also sponsored by the Lean Construction Institute.

14
00:01:05,417 --> 00:01:10,797
LCI is working to lead the building industry and transforming its practices and culture.

15
00:01:11,017 --> 00:01:15,217
Its vision is to create a healthy and thriving industry that delivers outstanding

16
00:01:15,217 --> 00:01:18,057
project outcomes every time for everyone.

17
00:01:18,297 --> 00:01:22,077
Check the show notes for more information. Now, to the show.

18
00:01:22,237 --> 00:01:24,537
Welcome to the show, Marcus Turner.

19
00:01:24,897 --> 00:01:31,637
Marcus, I am so excited to have you on the show that I even woke up early to

20
00:01:31,637 --> 00:01:34,577
make sure that I didn't miss you and thank you so much for coming on the show,

21
00:01:34,737 --> 00:01:36,837
ladies and gentlemen, this is going to be a treat.

22
00:01:37,217 --> 00:01:42,737
Marcus Turner is the one person that you need to be following on LinkedIn and social media.

23
00:01:42,957 --> 00:01:50,757
He is my secret weapon as to how I've become so able to use AI to enhance my

24
00:01:50,757 --> 00:01:54,937
workflows, my lifestyle, and all the things that I want to get done to make

25
00:01:54,937 --> 00:01:57,157
my life easier, better, and faster.

26
00:01:57,677 --> 00:02:02,037
Marcus, please introduce yourself so people can know why you're such a gangster

27
00:02:02,037 --> 00:02:06,817
when it comes to AI, artificial intelligence, and all things tech.

28
00:02:08,132 --> 00:02:11,632
Thank you for that, Felipe. You definitely have taken some things that I've

29
00:02:11,632 --> 00:02:14,532
showed you and definitely ran with it in ways that I never imagined.

30
00:02:14,812 --> 00:02:18,972
And I am so proud of all the things that you've accomplished where I'm learning

31
00:02:18,972 --> 00:02:20,332
stuff from you at this point.

32
00:02:20,372 --> 00:02:23,012
And we're sharing information back and forth.

33
00:02:23,272 --> 00:02:25,932
It's really cool because we can text each other like, look what I found,

34
00:02:25,992 --> 00:02:27,352
look what I found in the sky. Awesome.

35
00:02:28,332 --> 00:02:33,112
My name is Marcus Turner. I am part of the second half of Constructor.

36
00:02:33,252 --> 00:02:36,952
If anybody here who follows the show remembers Brittany Campbell Turner.

37
00:02:37,452 --> 00:02:42,972
She is my wife and we're on the journey, working together, trying to figure

38
00:02:42,972 --> 00:02:45,532
out how to make the construction industry more interesting.

39
00:02:45,872 --> 00:02:53,312
How I stumbled upon AI and working with it was actually around 2020,

40
00:02:53,752 --> 00:02:59,352
yeah, 2020 when we were working on marketing efforts, as far as trying to get

41
00:02:59,352 --> 00:03:00,612
our communication out there.

42
00:03:00,812 --> 00:03:04,832
And we were messing with what I didn't know at the time, but I actually got

43
00:03:04,832 --> 00:03:08,792
to learn through Clubhouse. Everybody remembers being on Clubhouse in 2020?

44
00:03:09,092 --> 00:03:15,172
Yeah, I remember. So, but what I learned was, is that I was using it through

45
00:03:15,172 --> 00:03:19,972
this one platform that was trying to help us work on our marketing messages.

46
00:03:20,432 --> 00:03:24,992
And I was really frustrated during the time of trying to figure out like,

47
00:03:24,992 --> 00:03:28,712
Like, what do I need to say to get it to look the way that I wanted to do?

48
00:03:28,792 --> 00:03:30,692
I don't understand how to get AI to do that.

49
00:03:31,032 --> 00:03:35,212
And then I was like, so what is this system using? I don't understand why the

50
00:03:35,212 --> 00:03:38,372
software is even needed in the first place.

51
00:03:39,192 --> 00:03:44,692
And I started learning about GPT-3. And I started learning about prompt engineering.

52
00:03:45,172 --> 00:03:48,692
And so I realized I didn't really need that software product anymore.

53
00:03:48,912 --> 00:03:52,252
So I kind of like dished that product off to the side. And I was like,

54
00:03:52,292 --> 00:03:57,292
oh, okay, this GPT-3 thing seems to be kind of more worth it,

55
00:03:57,312 --> 00:03:58,352
trying to mess around with that.

56
00:03:58,492 --> 00:04:01,332
I had some results and actually I walked away from it.

57
00:04:01,612 --> 00:04:05,152
And then, like a lot of people, Chet GPT came around.

58
00:04:06,872 --> 00:04:11,192
And I got excited. And I was like, how factual is this information?

59
00:04:11,452 --> 00:04:16,412
Trying to figure stuff out. So during that time, it came out during the same

60
00:04:16,412 --> 00:04:20,512
time college finals were happening. And I had some friends who were professors.

61
00:04:21,592 --> 00:04:26,472
So, I had some friends who were professors who were very kind to me to actually

62
00:04:26,472 --> 00:04:31,592
give me some of their finals and some papers and gave me some ideas to come up with.

63
00:04:31,712 --> 00:04:36,412
And I was just like, I can see you totally. If I was a college student, I can use this to cheat.

64
00:04:39,852 --> 00:04:45,152
Get that edge. Cool, transparent. And so, I asked my college friends because

65
00:04:45,152 --> 00:04:50,392
I really wanted to dive into how accurate was the information and what is the

66
00:04:50,392 --> 00:04:51,352
acceptable level. problem.

67
00:04:52,585 --> 00:04:57,485
And they were like, this is actually not that bad. This is actually pretty good.

68
00:04:57,685 --> 00:05:01,885
And then they showed me some answers that they came up with and the way that

69
00:05:01,885 --> 00:05:05,045
the AI solved the problem. I got to the right solution.

70
00:05:05,285 --> 00:05:08,745
The steps that I did was not necessarily something that they would teach,

71
00:05:08,945 --> 00:05:11,005
but I still got to the same solution.

72
00:05:11,325 --> 00:05:14,625
And then so we had some conversations back and forth.

73
00:05:15,525 --> 00:05:18,725
How are you going to prepare for the next semester? Stuff like that.

74
00:05:18,985 --> 00:05:22,865
Do you think this is going to replace students, teachers? Do you think it's going to help teachers?

75
00:05:23,685 --> 00:05:28,365
And a lot of them actually came with it with them from a very positive mindset.

76
00:05:29,005 --> 00:05:33,325
One of the teachers, one of the professors I was talking to was a communication professor.

77
00:05:33,905 --> 00:05:37,805
So there's a lot of written papers and things like that. And he was like,

78
00:05:37,885 --> 00:05:40,425
you know, I'll just probably do a lot more like speaking.

79
00:05:40,825 --> 00:05:44,805
So how do you present yourself and things like that? So more real world type

80
00:05:44,805 --> 00:05:47,565
of stuff instead of just like looking at papers every single time.

81
00:05:47,565 --> 00:05:53,285
Because a lot of, I mean, you can have as much AI detection software out there in the world.

82
00:05:53,525 --> 00:05:56,725
It doesn't mean that all someone has to do is just running through two different

83
00:05:56,725 --> 00:06:00,625
AI programs or something like that. And then all of a sudden, Hey, I got a paper.

84
00:06:01,685 --> 00:06:04,485
So for all the listeners, you just heard how to hack through.

85
00:06:05,085 --> 00:06:07,605
It's that easy to beat the AI detection.

86
00:06:08,785 --> 00:06:11,985
Yeah. And it's always going to be a, like a forever game, you know,

87
00:06:11,985 --> 00:06:15,445
going back and for it so and i think

88
00:06:15,445 --> 00:06:18,665
so that was a i think that was part of it and then another

89
00:06:18,665 --> 00:06:21,705
part was talking to one that was a math

90
00:06:21,705 --> 00:06:25,525
teacher i actually learned that it's not as good with

91
00:06:25,525 --> 00:06:28,285
an engineering professor friend of mine it wasn't as

92
00:06:28,285 --> 00:06:31,045
good with math and as i thought

93
00:06:31,045 --> 00:06:33,965
it was so we tried some stuff and i

94
00:06:33,965 --> 00:06:36,945
was like okay so ai is not fully ready to

95
00:06:36,945 --> 00:06:40,485
be kind of this all-encompassing thing

96
00:06:40,485 --> 00:06:43,225
where we can just ask a question to it and then give us a

97
00:06:43,225 --> 00:06:46,125
jarvis-like answer and everybody can be tony stark iron man

98
00:06:46,125 --> 00:06:49,945
you know so it's got

99
00:06:49,945 --> 00:06:53,885
some limitations i mean i'll i've even asked it like basic finance equations

100
00:06:53,885 --> 00:06:57,785
and it's good at telling you like what's the definition of the equation and

101
00:06:57,785 --> 00:07:01,485
it it'll lie to you and tell you that it's calculating correctly and then if

102
00:07:01,485 --> 00:07:08,145
you run the the math yourself on the side you're like hey math doesn't math and you call it math.

103
00:07:08,365 --> 00:07:11,145
It doesn't math math doesn't math and then it

104
00:07:11,145 --> 00:07:14,125
it'll apologize to you and then run it

105
00:07:14,125 --> 00:07:17,305
again and like magically on the second time it usually gets

106
00:07:17,305 --> 00:07:22,645
it right usually not always yeah check check ai's math people if you're relying

107
00:07:22,645 --> 00:07:28,625
on ai math you're gonna be disappointed yeah i i think one of the things that

108
00:07:28,625 --> 00:07:33,925
i find i don't use it as much for math anymore but what i one of the things

109
00:07:33,925 --> 00:07:36,365
i I actually use a lot for it is telling it,

110
00:07:36,385 --> 00:07:39,265
asking it code, especially with chat GPT.

111
00:07:39,565 --> 00:07:43,425
Sometimes you can get it to solve a math problem by showing code.

112
00:07:44,245 --> 00:07:47,305
It's actually a lot more reliable.

113
00:07:48,679 --> 00:07:52,519
I think they have some, they have some Python scripts that they can kind of

114
00:07:52,519 --> 00:07:56,199
run in the background, at least for a chat, for chat GPT.

115
00:07:56,679 --> 00:08:00,419
We're getting to the AIs after that, but that is something that they actually

116
00:08:00,419 --> 00:08:02,139
do a really good job with. So.

117
00:08:02,559 --> 00:08:06,539
Yeah, I absolutely do. I agree. And there's even some with the GPT store.

118
00:08:06,579 --> 00:08:11,319
Now you can create your own GPT and force it into those environments where you're

119
00:08:11,319 --> 00:08:14,199
calling up like Python to do some things.

120
00:08:14,199 --> 00:08:20,259
Things and if you're if you're running software that's like good for math like python super good.

121
00:08:20,799 --> 00:08:24,999
Your odds for correct answers are going to be very high but the large language

122
00:08:24,999 --> 00:08:29,379
model in itself is like you're asking it to do something that it was not intended

123
00:08:29,379 --> 00:08:33,399
to do even though there have been emergent things happening with open ai which

124
00:08:33,399 --> 00:08:36,259
i want to get into on the show but before we get into all that marcus thank

125
00:08:36,259 --> 00:08:40,019
you for introducing yourself and i think the other thing you forgot

126
00:08:40,099 --> 00:08:43,299
to mention i'm just going to remind you marcus and i are both nerds

127
00:08:43,299 --> 00:08:47,239
so just in case you didn't know and it didn't come across like you

128
00:08:47,239 --> 00:08:50,639
heard him he's like talking to teachers he's not in school anymore he's been

129
00:08:50,639 --> 00:08:54,919
out of school for a long time and he still talks to teachers so just go to show

130
00:08:54,919 --> 00:08:58,639
like that's uh things we have in common like we're both you know electrically

131
00:08:58,639 --> 00:09:05,019
uh focused and we both are not in school but still learning every single day which is why i love

132
00:09:05,099 --> 00:09:09,239
getting those text messages from you at 11 o'clock at night because it's going

133
00:09:09,239 --> 00:09:14,219
to be something good it's going to be good so i wanted to ask you like just

134
00:09:14,219 --> 00:09:24,379
starting off what is something that surprised you recently with ai i haven't gotten into this one yet.

135
00:09:26,039 --> 00:09:32,999
So this is a lot i think the thing that actually caught my interest was this

136
00:09:32,999 --> 00:09:35,159
thing called ai AI scientist.

137
00:09:36,079 --> 00:09:37,139
And...

138
00:09:38,758 --> 00:09:41,878
There's a version of it that I could talk about a little bit earlier,

139
00:09:42,118 --> 00:09:45,658
but I think going into this one was pretty deep.

140
00:09:45,998 --> 00:09:55,678
So AI scientists is created by a lot of some people that are scholar researchers and things like that.

141
00:09:55,818 --> 00:10:00,238
And basically, they just wanted to do they're like, well, you know,

142
00:10:00,258 --> 00:10:04,718
there's this process that we do to write these academic research articles and things like that.

143
00:10:04,718 --> 00:10:11,098
But what we also want to do is take the research articles and start adding them

144
00:10:11,098 --> 00:10:18,718
together and trying to see if AI can find some ideas and help us find some solutions from it.

145
00:10:18,938 --> 00:10:27,298
So that's the big, carry audacious goal of theirs is that now we could use AI

146
00:10:27,298 --> 00:10:31,518
to do all this research and this research thing can now be executed by AI.

147
00:10:31,518 --> 00:10:36,038
And now we can find some other things that we didn't find before because we

148
00:10:36,038 --> 00:10:39,938
just didn't necessarily think about it in the way that an AI would.

149
00:10:40,058 --> 00:10:46,238
I mean, I wouldn't say a way that AI would, but like we can't iterate as fast

150
00:10:46,238 --> 00:10:52,238
and basically using AI to be able to be the assistant to help us be able to do that.

151
00:10:52,538 --> 00:10:56,418
Now we can be able to do some things a lot better than we normally could before. for.

152
00:10:56,878 --> 00:11:03,058
So that has caught my interest as of late.

153
00:11:04,338 --> 00:11:12,658
And I think the other thing that has caught my interest just recently is seeing

154
00:11:12,658 --> 00:11:21,018
AI as far as being able to kind of create your own small apps on the side.

155
00:11:22,545 --> 00:11:28,285
And it could be like a calculator. It can be like, hey, I can actually be,

156
00:11:28,445 --> 00:11:33,305
hey, I just want to create an interactive lesson plan, stuff like that.

157
00:11:33,345 --> 00:11:37,305
Little small stuff that you'd ever really think of. Oh, I don't have a PDF highlighter.

158
00:11:37,665 --> 00:11:39,685
Oh, I got to buy a PDF highlighter.

159
00:11:40,445 --> 00:11:44,145
Well, why can't I just ask AI to help me build a PDF highlighter?

160
00:11:44,625 --> 00:11:49,085
So now I can highlight the PDF and send it back to the person that I was talking

161
00:11:49,085 --> 00:11:54,225
to. and i spent no dollars on that kind of awesome.

162
00:11:57,545 --> 00:12:00,205
That's so cool and we had a i've done

163
00:12:00,205 --> 00:12:03,105
a couple ai talks and you've done a couple ai talks as

164
00:12:03,105 --> 00:12:06,545
well in the construction space and the

165
00:12:06,545 --> 00:12:10,325
last one i did one i had a researcher from the university of austin texas in

166
00:12:10,325 --> 00:12:13,545
the audience and we had a there's a section at the end of the presentation where

167
00:12:13,545 --> 00:12:17,945
i I just open up a prompt window and I ask the audiences anything they want

168
00:12:17,945 --> 00:12:23,025
to see so they could practice prompting with me because I've got years of experience

169
00:12:23,025 --> 00:12:24,445
now prompting thanks to you.

170
00:12:24,825 --> 00:12:29,345
And so we asked it about research. And the interesting thing is I found out

171
00:12:29,345 --> 00:12:33,245
a couple of things like construction research right now is almost nothing on AI.

172
00:12:33,485 --> 00:12:37,705
And the reason for that is because it takes researchers with peer review and

173
00:12:37,705 --> 00:12:42,565
the process they have to go through about two years on the fast side and more

174
00:12:42,565 --> 00:12:44,345
than two years on the typical side.

175
00:12:44,705 --> 00:12:50,745
So we're about to see a wave starting next year and the year after on lots of

176
00:12:50,745 --> 00:12:54,325
research published about AI in the design and construction space.

177
00:12:54,485 --> 00:12:58,225
It's coming, but it's not, there's almost nothing there yet because all the

178
00:12:58,225 --> 00:13:00,845
papers are in review and going through the process.

179
00:13:01,105 --> 00:13:05,145
And so the researcher asked like, I've got this topic I'm thinking about.

180
00:13:06,018 --> 00:13:10,558
And through prompting, we took that topic and then had, we used GPT and CLAUD,

181
00:13:10,658 --> 00:13:15,258
create iterations of things that they can study that would be interesting based

182
00:13:15,258 --> 00:13:17,318
on what their objectives were, the research.

183
00:13:18,178 --> 00:13:22,738
And they were just amazed at how fast it did it.

184
00:13:23,138 --> 00:13:27,198
And she said, her comment was like, what you just did on the screen could sometimes

185
00:13:27,198 --> 00:13:31,338
take us like three months or four months to get to the same end result.

186
00:13:31,638 --> 00:13:35,898
And you just did it in like 20 seconds. And it's only because I'm a slow typer.

187
00:13:37,218 --> 00:13:39,918
Yeah yeah if i would have voice to texted it it would

188
00:13:39,918 --> 00:13:42,778
have been like 10 seconds yeah and

189
00:13:42,778 --> 00:13:45,658
i actually obviously i really can't wait for the the

190
00:13:45,658 --> 00:13:48,338
40 voice to text feature to come out so i can kind

191
00:13:48,338 --> 00:13:51,018
of dialogue with the ai a little bit more but i use it

192
00:13:51,018 --> 00:13:54,218
now and i absolutely love it and i think you know

193
00:13:54,218 --> 00:13:57,378
i think that's gonna probably be the way that a lot of people start interacting with

194
00:13:57,378 --> 00:14:02,658
it a little bit because even today it was weird for me like a little while ago

195
00:14:02,658 --> 00:14:07,058
but i have friends who sent me voice memos instead of actual text passages which

196
00:14:07,058 --> 00:14:11,578
is pretty dope but i didn't even think that was like when i first got it i was

197
00:14:11,578 --> 00:14:14,298
like i used to be a voice about it was like.

198
00:14:16,618 --> 00:14:20,758
I know i have a i have a couple friends that consistently send me voice memos,

199
00:14:22,038 --> 00:14:26,498
and i'm still not used to it like you know you're you just get we're just creatures

200
00:14:26,498 --> 00:14:31,638
of habit marcus you get used to seeing characters in your little box in the

201
00:14:31,638 --> 00:14:34,338
message and that's what you want to you you expect and you get weird stuff.

202
00:14:34,398 --> 00:14:38,778
Like I have a friend that sends me videos through WhatsApp, which I really love.

203
00:14:38,818 --> 00:14:42,018
Like those are so cool to get short videos and WhatsApp.

204
00:14:42,118 --> 00:14:44,598
So there's, I mean, there's different things that people use and,

205
00:14:45,240 --> 00:14:51,120
Just, I like, of all my friends that message me, I like it that you spice it up. Keep it spicy.

206
00:14:52,340 --> 00:14:55,480
I'll start sending them some more voice memos to everyone. Yeah.

207
00:14:55,620 --> 00:14:57,740
Yeah. Let's see. Let's start a trend. Voice memos.

208
00:14:58,320 --> 00:15:01,360
So I want to get into unpacking a couple of things. Everyone,

209
00:15:01,620 --> 00:15:03,060
this is going to be a great episode.

210
00:15:03,160 --> 00:15:07,000
You're going to want to listen to the show two times, unpacking a little bit

211
00:15:07,000 --> 00:15:10,440
about AI and some of the current trends in design and construction.

212
00:15:10,660 --> 00:15:13,900
All right. All right. So Marcus, as an observer in the construction industry,

213
00:15:14,060 --> 00:15:17,320
you've had a lot of experience in your prior career.

214
00:15:17,920 --> 00:15:21,740
What has surprised you about how buildings get built?

215
00:15:22,700 --> 00:15:27,600
I think to me, it depends on the organization.

216
00:15:28,420 --> 00:15:31,140
And I think that part bothers me.

217
00:15:32,420 --> 00:15:36,700
How much upfront work an organization wants to do.

218
00:15:36,700 --> 00:15:44,980
And I think the other thing that bothers me a lot is when you don't do the upfront

219
00:15:44,980 --> 00:15:49,460
work and some companies use this term,

220
00:15:49,560 --> 00:15:52,280
not all do, but value engineering.

221
00:15:53,820 --> 00:16:00,000
And it really is, we, I'm not going to say the word because,

222
00:16:00,120 --> 00:16:01,900
you know, it's a podcast, we effed up.

223
00:16:02,460 --> 00:16:10,540
And we are now going to figure out what can we cut from the process that we

224
00:16:10,540 --> 00:16:14,820
think it's not as, wouldn't impact the project as much.

225
00:16:15,400 --> 00:16:20,000
And sometimes just having the upfront conversation would be it.

226
00:16:20,340 --> 00:16:23,960
So I think those two spaces or those two things are.

227
00:16:25,616 --> 00:16:29,376
Because my background for was working at a nuclear power plant.

228
00:16:29,536 --> 00:16:34,356
And one of the things that we did consistently is we brought all the parties together.

229
00:16:34,936 --> 00:16:40,036
We're going to be working on the project or the process or whatever the maintenance activity was.

230
00:16:40,756 --> 00:16:44,736
And they were pretty extreme with it. So this will go down to even changing of a light bulb.

231
00:16:45,836 --> 00:16:49,916
So who are the people you need to call, sit down, have a work package put together,

232
00:16:50,056 --> 00:16:55,656
have a plan scheduled around the plan. and you execute the changing of the light bulb.

233
00:16:56,496 --> 00:17:00,836
But there is some consistency within that.

234
00:17:00,916 --> 00:17:05,916
So you kind of know what to expect and you kind of set a pattern and you set a standard.

235
00:17:06,476 --> 00:17:11,256
You know, those who are sports fans, you know, some people like their teams to play hard.

236
00:17:11,676 --> 00:17:17,176
And even though somebody, let's say, I'm a basketball fan, so I'm just going

237
00:17:17,176 --> 00:17:21,596
to say, contesting a shot after the whistle has been blown, stuff like that.

238
00:17:21,596 --> 00:17:25,496
Like there's a standard to play and you try to set the standard of play so you

239
00:17:25,496 --> 00:17:30,016
can be consistent so when those times when you are You know,

240
00:17:30,016 --> 00:17:32,176
you're tired you're fatigued and things like that.

241
00:17:32,316 --> 00:17:36,796
Everyone has that routine kind of built in So when something actually happens

242
00:17:36,796 --> 00:17:40,656
that needs to happen, you just kind of do it instinctively as a team or as a group.

243
00:17:41,016 --> 00:17:50,056
So I think I think that was really good and I think that when I I looked into

244
00:17:50,056 --> 00:17:54,536
construction, had conversations with construction professionals and things like that.

245
00:17:54,696 --> 00:18:00,456
That was the one thing that I saw that was kind of lacking was the lack of camaraderie and teamwork.

246
00:18:02,165 --> 00:18:07,425
On Brittany's podcast, one of the, I think one of my most favorite quotes that

247
00:18:07,425 --> 00:18:14,365
a guest said was, is that, that basically construction people come together

248
00:18:14,365 --> 00:18:15,845
as strangers and leave as enemies.

249
00:18:23,185 --> 00:18:27,985
And it's a pretty strong statement. And you know, we, I would like,

250
00:18:28,005 --> 00:18:29,005
I love to see less of that.

251
00:18:29,005 --> 00:18:33,125
And I would love to see people kind of working on processes and projects and

252
00:18:33,125 --> 00:18:37,245
things like that, because, I mean, we don't want to just go to work and just

253
00:18:37,245 --> 00:18:38,385
know it's a frustrating experience.

254
00:18:38,625 --> 00:18:40,985
We kind of want to go to work and have some relief.

255
00:18:41,965 --> 00:18:46,905
And I think that it would be, you know, construction is sometimes a very long

256
00:18:46,905 --> 00:18:49,685
process because it's going to be years being with these people.

257
00:18:49,885 --> 00:18:54,685
So I think having more of a collaborative environment would be beneficial.

258
00:18:55,385 --> 00:18:59,025
That is so, so good. Good. That's such a good quote.

259
00:18:59,345 --> 00:19:04,385
I was on a job one time, and I also like to remember the things people say,

260
00:19:04,485 --> 00:19:08,305
and I even will write down. So I put it to commit it to memory.

261
00:19:08,905 --> 00:19:14,585
And one of the things that popped into mind as you were saying that is that there was a contractor.

262
00:19:15,125 --> 00:19:19,645
It's a mixed job. It's like, it's like a complicated job design build.

263
00:19:19,645 --> 00:19:26,005
And one of the general contractors in charge said something to the effect of,

264
00:19:26,065 --> 00:19:31,065
we've decided what we're going to do. We're no longer asking for your permission.

265
00:19:31,665 --> 00:19:34,805
And they said, tough shit, we're doing it.

266
00:19:35,165 --> 00:19:40,125
And that was like their quote. That was in response to, I remember,

267
00:19:40,305 --> 00:19:43,305
a trade contractor was asking a question.

268
00:19:43,585 --> 00:19:48,085
It's related to design. And that, that response was like, as if the designer,

269
00:19:48,225 --> 00:19:52,465
you know, was, was present, which they weren't at the time, but this is the

270
00:19:52,465 --> 00:19:55,025
type of the environment is such that.

271
00:19:55,945 --> 00:19:59,885
People get pitted against each other, even though they're trying to build something together.

272
00:20:00,485 --> 00:20:04,465
Yeah. And that's even, that's a design build. I mean, so like people are taking

273
00:20:04,465 --> 00:20:09,825
hard lines, even in design build, which is supposed to be like the solution to the hard bid.

274
00:20:09,965 --> 00:20:13,845
It's supposed to solve all the problems of the hard bid, just like construction

275
00:20:13,845 --> 00:20:18,285
management was supposed to solve all the problems of the hard bid, you know, before that.

276
00:20:18,425 --> 00:20:21,445
And we keep iterating. Now we've got to fast forward to today.

277
00:20:21,705 --> 00:20:24,525
We have integrated project delivery. It's supposed to solve all those problems.

278
00:20:24,525 --> 00:20:27,625
But we can unpack that and talk about that on another show.

279
00:20:28,025 --> 00:20:31,925
So I want to go back into AI. And this is a true story. But the audience,

280
00:20:32,025 --> 00:20:35,465
you don't know that in order to make this podcast happen today.

281
00:20:36,345 --> 00:20:43,145
Marcus and I both had to use AI in different ways just to get all of our tech to work.

282
00:20:43,385 --> 00:20:47,085
And we were, I was commenting to him as we were getting set up that,

283
00:20:47,105 --> 00:20:49,485
oh, Zoom has all these new features in place.

284
00:20:49,605 --> 00:20:53,685
Like we could share our screens at the same time. It's like I didn't get the

285
00:20:53,685 --> 00:20:56,625
memo on the release notes Zoom for that happening.

286
00:20:56,665 --> 00:20:59,225
We are recording this in Zoom because it's just easy.

287
00:20:59,745 --> 00:21:03,865
And I've experimented. If you're a fan of the show, we're in season five.

288
00:21:04,485 --> 00:21:09,965
I've used four different platforms throughout the years to make these shows possible.

289
00:21:10,225 --> 00:21:13,705
And I started with Zoom and I'm back on Zoom after five years.

290
00:21:13,965 --> 00:21:16,385
But I still experiment and I try stuff.

291
00:21:17,265 --> 00:21:21,185
And I'm super glad that Marcus also does the same. We learn to keep,

292
00:21:21,205 --> 00:21:25,865
we keep sharp when we come to like a black screen or like stuff that doesn't

293
00:21:25,865 --> 00:21:27,785
work. It doesn't, nothing can stop the show.

294
00:21:28,345 --> 00:21:33,465
Yeah. Yeah, definitely. It's crazy. Like sometimes when you run into problems

295
00:21:33,465 --> 00:21:39,945
and this was actually a quote, I think one of the things that a friend said to me recently,

296
00:21:40,145 --> 00:21:43,525
and I think this, this is actually something I actually use in my talks. now.

297
00:21:44,185 --> 00:21:49,965
We get so used to doing something a certain way that we forget that we can just ask AI.

298
00:21:51,075 --> 00:21:56,855
Or the solution to whatever our problem is, because we've been so trained to

299
00:21:56,855 --> 00:21:59,695
do it a certain way, just based off of our habits.

300
00:22:00,395 --> 00:22:05,215
That is so true. That is so true. I was roped into something recently,

301
00:22:05,375 --> 00:22:10,455
Marcus, where somebody had gotten an assignment and this person does not use AI.

302
00:22:10,635 --> 00:22:13,695
They're aware that AI exists, but they haven't played with it at all.

303
00:22:13,755 --> 00:22:17,035
And the company they work for does not encourage them to play with it.

304
00:22:17,035 --> 00:22:23,795
And I could tell because of how you've helped me retrain my brain with prompt engineering,

305
00:22:24,095 --> 00:22:28,795
that my ability to use all of my lean thinking to turn my thinking around to

306
00:22:28,795 --> 00:22:34,835
what's possible and to focus on the outcome that I want changes how I talk to people.

307
00:22:35,275 --> 00:22:39,435
So when I was talking to this individual, I said, because they asked me, like, can you help me?

308
00:22:39,555 --> 00:22:44,715
Oh, I need some lean help is what they said. And so for me, I confirmed that

309
00:22:44,715 --> 00:22:47,115
they actually wanted coaching and that everything's possible.

310
00:22:47,315 --> 00:22:52,655
It's a wide open because sometimes as a coach, you could be asked a question

311
00:22:52,655 --> 00:22:55,695
and they don't really want your help. They just want you to listen to their process.

312
00:22:55,935 --> 00:22:59,455
This was a case where they actually wanted coaching. So I said,

313
00:22:59,495 --> 00:23:03,135
this is what you said, this is your state, this is what's going to happen if

314
00:23:03,135 --> 00:23:05,775
you go down your path, and then here's an alternative,

315
00:23:06,075 --> 00:23:09,955
if you change two things, you can now do this, what you thought was going to

316
00:23:09,955 --> 00:23:12,195
take two years, and what had taken two years.

317
00:23:12,695 --> 00:23:15,655
You now can get done in less than four weeks.

318
00:23:15,655 --> 00:23:19,855
Weeks and at first they were like no can't

319
00:23:19,855 --> 00:23:22,655
be done and i just walked them through like it's just

320
00:23:22,655 --> 00:23:26,095
thinking and like the way that you prompt give some

321
00:23:26,095 --> 00:23:30,135
context this is what you want i have to do a little algebra to calculate the

322
00:23:30,135 --> 00:23:34,295
time and my algebra was pretty good because they had the they already had the

323
00:23:34,295 --> 00:23:43,115
answer their their process as is takes two years the revised process us four weeks. It was awesome.

324
00:23:43,775 --> 00:23:47,475
And now it's happening. And I asked the question just to close the loop.

325
00:23:47,535 --> 00:23:49,335
I said, what's your first step?

326
00:23:50,328 --> 00:23:54,128
And I made them like, tell me what the first step was. And I was like, nope, that's not it.

327
00:23:55,248 --> 00:23:59,108
They went right back into the process that they used for two years.

328
00:24:00,248 --> 00:24:03,928
And I was like, nope, I'm going to sit here and I'm going to wait until you

329
00:24:03,928 --> 00:24:06,868
execute step one to make sure that it starts.

330
00:24:06,928 --> 00:24:09,608
And then I said, I'm going to see you in a couple of days.

331
00:24:09,908 --> 00:24:13,328
And the first thing I'm going to say is, are you on to step two yet?

332
00:24:13,708 --> 00:24:19,808
And I'm just going to be on you like until we get done and we close it out in four weeks.

333
00:24:19,948 --> 00:24:22,688
In reality, they got six weeks to get it done. So we got a two-week buffer.

334
00:24:22,868 --> 00:24:27,148
So I give it two weeks for them to go back and make some mistakes.

335
00:24:27,288 --> 00:24:29,668
It's okay. It's okay. So I want to pack.

336
00:24:30,388 --> 00:24:34,128
Let's go back to, let's stay on AI, Marcus. And what are some of the current

337
00:24:34,128 --> 00:24:39,568
trends in design and construction that, you know, people like me that I'm out

338
00:24:39,568 --> 00:24:41,348
there nerding out with it all the time,

339
00:24:41,408 --> 00:24:43,088
but most of the audiences I've polled

340
00:24:43,088 --> 00:24:46,628
people and you've had talks with construction folks, and it's a mix.

341
00:24:46,768 --> 00:24:50,408
If you're going to an AI talk, you're going to get more than average people.

342
00:24:50,508 --> 00:24:57,168
When I talk to the everyday construction partner, they're not using AI at all. What's your take?

343
00:24:58,308 --> 00:25:02,368
I would say it's still very early days. It still feels very early days.

344
00:25:02,508 --> 00:25:04,448
Most people, they've heard of AI.

345
00:25:05,428 --> 00:25:10,428
Even some companies have even implemented AI solutions where they're using,

346
00:25:10,548 --> 00:25:17,428
you can access GPT-4.0 or Claude or Gemini, which is Google's product.

347
00:25:19,048 --> 00:25:25,588
So, but the thing is, is that a lot of people still use it or there's different

348
00:25:25,588 --> 00:25:30,648
levels of it is what one is, I've used it and I didn't like it.

349
00:25:31,008 --> 00:25:35,688
And typically those are the people who are kind of using it as a search engine.

350
00:25:37,108 --> 00:25:43,648
And it's not a search engine. I think one of the consistent themes that I do

351
00:25:43,648 --> 00:25:48,008
in my talks is, you know, you got to think of it as you're talking to an assistant or an intern.

352
00:25:48,488 --> 00:25:50,568
What are the steps that you're going to guide an intern with?

353
00:25:50,788 --> 00:25:53,848
How are you going to walk through the process with them? What's the end goal?

354
00:25:54,068 --> 00:25:54,968
What are some of the pitfalls?

355
00:25:55,388 --> 00:25:58,488
What are some of the constraints that you have on your question?

356
00:25:58,488 --> 00:26:06,668
So, and then the layer two of that is I would always say that from another thing

357
00:26:06,668 --> 00:26:09,508
I would tell people is that give it a role.

358
00:26:10,754 --> 00:26:18,334
You are a construction engineer or you are a commercial licensed architect.

359
00:26:18,994 --> 00:26:21,774
Licensed architect would probably get you some, they'll probably say,

360
00:26:21,834 --> 00:26:23,354
I'm not a licensed architect, blah, blah, blah, blah.

361
00:26:23,814 --> 00:26:28,134
But if you get it to role play in that realm, it can get you start thinking

362
00:26:28,134 --> 00:26:29,934
about what an architect would think about.

363
00:26:30,554 --> 00:26:35,154
Maybe if you just, even if you have like a document and maybe you have a perspective

364
00:26:35,154 --> 00:26:39,054
that you have on the document, but you want to understand another partner's

365
00:26:39,054 --> 00:26:40,094
perspective, right? Right.

366
00:26:40,234 --> 00:26:48,234
So those who do personas and companies, they understand they have a list of

367
00:26:48,234 --> 00:26:49,134
what those personas are.

368
00:26:49,314 --> 00:26:54,174
Well, based off a list of these personas, what would you think if I said this

369
00:26:54,174 --> 00:26:57,594
and you posted the document? Give me a debate back.

370
00:26:58,194 --> 00:27:01,834
You can get a list of things and now you can find out some of the objections

371
00:27:01,834 --> 00:27:06,114
or some of the concerns that somebody would have based off of a document or

372
00:27:06,114 --> 00:27:07,694
a plan or something like that. Scott.

373
00:27:07,894 --> 00:27:12,374
So I think just playing those two things help a lot.

374
00:27:12,534 --> 00:27:17,754
And then the third one is what I would always say is in most of your prompts,

375
00:27:17,994 --> 00:27:22,034
explaining in a step by step way or think step less things step by step.

376
00:27:22,714 --> 00:27:27,354
If you could do those things where it can kind of give you a breakdown of the

377
00:27:27,354 --> 00:27:29,514
process of what it's showing.

378
00:27:31,883 --> 00:27:37,963
It's still a prediction model but it's a prediction model based off the world

379
00:27:37,963 --> 00:27:46,423
what i mean by that is if it basically predicts the next token so the cow jumps over the moon you know,

380
00:27:48,683 --> 00:27:54,903
i'll play along okay so you know it's it's stuff like that so it you know romeo

381
00:27:54,903 --> 00:27:57,343
and juliet perfect Perfect.

382
00:27:58,503 --> 00:28:04,663
Julieta. Yes. But, you know, so it's playing off of that.

383
00:28:05,023 --> 00:28:09,883
And what you want to do is build it context. And if you build more context and

384
00:28:09,883 --> 00:28:13,343
you build in constraints and you build in role playing and you tell it to think

385
00:28:13,343 --> 00:28:19,063
step by step, you will get better results out of what your answers are.

386
00:28:19,063 --> 00:28:22,223
Instead of saying okay tell

387
00:28:22,223 --> 00:28:25,043
me the top five blah blah blah

388
00:28:25,043 --> 00:28:29,083
blah blah just like how you would use google give me the latest blah blah blah

389
00:28:29,083 --> 00:28:37,423
blah blah blah and that's just not what a large language model using just chat

390
00:28:37,423 --> 00:28:43,963
gpt or using just google gemini by itself that's not what they're capable of.

391
00:28:44,283 --> 00:28:49,563
They don't have access to the search engines and they don't have access to those type of things.

392
00:28:49,963 --> 00:28:54,683
Now, if I said you can give an A, now you can give it a tool to use,

393
00:28:54,903 --> 00:28:57,023
then it would use that tool.

394
00:28:57,283 --> 00:28:59,263
Then it would synthesize some of that stuff.

395
00:28:59,583 --> 00:29:03,663
And that's what you're seeing. I know, I'm assuming most people still use Chrome today.

396
00:29:04,143 --> 00:29:08,143
That's what you're, or that's what you're seeing within Chrome now that you're

397
00:29:08,143 --> 00:29:11,643
seeing AI generated stuff at the top for a little bit.

398
00:29:11,723 --> 00:29:17,503
And it's giving you a summary or synopsis, and now you're seeing the links that are followed behind it.

399
00:29:18,145 --> 00:29:23,465
Bing does the same thing now, and their Bing chat feature, those who are using,

400
00:29:23,725 --> 00:29:26,765
and they're using, Bing chat is using GPT 4.0.

401
00:29:27,105 --> 00:29:30,985
So if you prefer 4.0, I'm not saying that you should go to Bing,

402
00:29:31,125 --> 00:29:35,825
but if you prefer 4.0, you can use Bing chat to be able to do so.

403
00:29:36,165 --> 00:29:39,825
Yeah, and even Amazon is using Rufus to synthesize all the reviews.

404
00:29:40,505 --> 00:29:43,725
So if you go to, and I learned about this recently,

405
00:29:43,785 --> 00:29:46,705
I was on Amazon looking at cameras of all things, of course,

406
00:29:46,705 --> 00:29:51,305
because podcasters what my next purchase is going to be another camera and it

407
00:29:51,305 --> 00:29:55,965
would take like some of these cameras have you know thousands of reviews thousands

408
00:29:55,965 --> 00:30:00,825
because products people love the products they talk about it so rufus for amazon

409
00:30:00,825 --> 00:30:03,405
is now giving you a like a one paragraph,

410
00:30:04,045 --> 00:30:07,645
here's what most of the reviews say helpful not

411
00:30:07,645 --> 00:30:11,505
helpful like not like and it's like a very consistent but people

412
00:30:11,505 --> 00:30:14,545
look at in the reviews that it's synthesizing for

413
00:30:14,545 --> 00:30:17,505
you so you don't have to look at 10 000 reviews you

414
00:30:17,505 --> 00:30:20,505
can and it's a pretty good summary i've used

415
00:30:20,505 --> 00:30:23,785
it a few times with some different types of products that i didn't

416
00:30:23,785 --> 00:30:26,625
have a lot of experience with and i found it to be very helpful

417
00:30:26,625 --> 00:30:29,625
and when it first turned on it was terrible because uh

418
00:30:29,625 --> 00:30:34,905
you know like all things ai like it starts really bad but they've got to expose

419
00:30:34,905 --> 00:30:40,465
it to humans to help it train and learn and then it gets better very fast like

420
00:30:40,465 --> 00:30:44,285
the iterations and the speed of improvements like I've been experimenting with

421
00:30:44,285 --> 00:30:47,805
GPT on an app on my phone, talking to it.

422
00:30:48,245 --> 00:30:54,785
And we talked about this, Marcus. I had a consulting call about Scrum, probably about...

423
00:30:55,694 --> 00:30:59,874
You know, almost a year ago. And the voice to text was already a thing.

424
00:31:00,414 --> 00:31:04,354
And so I took the call, I was on the road and it was like a one hour call.

425
00:31:04,454 --> 00:31:06,314
I didn't record the call. I didn't have the client's permission.

426
00:31:06,514 --> 00:31:09,654
I didn't ask. I'm in California. So we have to always ask if you're going to

427
00:31:09,654 --> 00:31:12,554
record somebody, you got to ask their permission to be legal and compliant.

428
00:31:12,874 --> 00:31:15,034
So I just said, let me just do this live. I'm not going to record.

429
00:31:15,574 --> 00:31:19,374
And then after the call was over, I turned on GPT.

430
00:31:19,474 --> 00:31:24,514
And at that time, I think it was three, five voice to text, the audio version,

431
00:31:24,514 --> 00:31:26,914
the earlier version, not the version that's out now.

432
00:31:27,354 --> 00:31:30,974
And I just had a conversation with it. I said, I just had a consulting call.

433
00:31:31,154 --> 00:31:35,874
I want you to act as a coach to me, to coach me, the consultant,

434
00:31:35,994 --> 00:31:37,414
to make me a better consultant.

435
00:31:37,834 --> 00:31:41,434
I'm going to tell you what I talked about from memory. I want you to ask me

436
00:31:41,434 --> 00:31:45,774
questions for things that are not clear as I'm sharing you this narrative of

437
00:31:45,774 --> 00:31:51,154
what happened and then coach me to do better to help the person the next time we talk.

438
00:31:51,334 --> 00:31:55,894
I want definitive things. So we had this conversation, I'm driving and I'm just

439
00:31:55,894 --> 00:31:59,374
talking and I talked to it for 45 minutes,

440
00:31:59,594 --> 00:32:05,314
but coaching call was an hour, but I would talk to AI for 45 minutes while I

441
00:32:05,314 --> 00:32:08,754
was driving and, you know, hands-free. So it's like this going through my car.

442
00:32:09,314 --> 00:32:11,834
And at the end of all that, I got a transcript.

443
00:32:12,514 --> 00:32:16,914
So I could go back and see like, Oh, do this better.

444
00:32:17,494 --> 00:32:20,654
Listen here. Like I gave me suggestions, like you did.

445
00:32:20,694 --> 00:32:23,954
And it even gave me good things too. So it's, you know, a lot of the things,

446
00:32:23,994 --> 00:32:28,274
Marcus, when you're working with people and you ask for, for help or feedback,

447
00:32:28,634 --> 00:32:31,814
everybody always wants to give you constructive criticism, like right away.

448
00:32:32,254 --> 00:32:36,894
And so I appreciated the large language model, reinforcing the good coaching

449
00:32:36,894 --> 00:32:42,514
things that I was doing, like, you know, listening first, asking clarifying

450
00:32:42,514 --> 00:32:44,994
questions. It's that I did a really good job of doing that.

451
00:32:45,674 --> 00:32:51,334
And I was just super impressed. And since then, I've continued to refer back

452
00:32:51,334 --> 00:32:58,694
to that occasionally and continue enhancing and improving my ability to be more effective coaching.

453
00:32:58,994 --> 00:33:03,474
I think now today, if you're listening to the show and you haven't practiced

454
00:33:03,474 --> 00:33:07,434
with AI, just think about people like Marcus and I that have been doing this

455
00:33:07,434 --> 00:33:09,774
for almost half a decade now.

456
00:33:10,764 --> 00:33:14,124
And if you're not, you can catch up. You can still start to learn.

457
00:33:14,344 --> 00:33:18,104
There's really no reason why you can't learn something. And Marcus,

458
00:33:18,184 --> 00:33:19,364
you've got stuff that you want to share.

459
00:33:20,104 --> 00:33:25,044
So I don't know if you've got more trends that you want to unpack or if you want to start sharing.

460
00:33:25,444 --> 00:33:30,624
I can actually follow up with another story that's in line with all of this, right?

461
00:33:31,244 --> 00:33:35,444
So working with, so Brittany is a part of the Respect for People Committee.

462
00:33:35,604 --> 00:33:40,484
So those who were at the LCI last event know that it was recorded.

463
00:33:40,764 --> 00:33:45,864
So, one of the things that we wanted to do was get a comprehensive definition

464
00:33:45,864 --> 00:33:49,064
of what respect for people actually means.

465
00:33:49,424 --> 00:33:52,884
So, we recorded the event. People spoke.

466
00:33:53,664 --> 00:33:59,304
And one of the things that we wanted to do was, one, anonymize everything that's there.

467
00:33:59,684 --> 00:34:02,624
And then, two, figure out what their stories were.

468
00:34:02,784 --> 00:34:07,804
So, we wanted to summarize each person's story. So we, well,

469
00:34:07,924 --> 00:34:13,084
this is what I did, was summarize each one's story and then from there,

470
00:34:13,164 --> 00:34:17,364
we looked at everyone's story and said, okay, based off of these stories,

471
00:34:17,544 --> 00:34:19,804
this is what our current working definition is.

472
00:34:20,684 --> 00:34:25,524
Okay, now, based off of these stories that you now have, that are now summaries,

473
00:34:25,724 --> 00:34:30,884
take these stories and now add that to the, what do you think the new definition

474
00:34:30,884 --> 00:34:33,364
should be based off of the stories that were provided? it.

475
00:34:33,684 --> 00:34:39,124
So, and then what I usually do during those type of processes is I don't never

476
00:34:39,124 --> 00:34:41,624
ask for just one based off of those.

477
00:34:41,684 --> 00:34:45,064
I ask for multiple because I want different, different things,

478
00:34:45,164 --> 00:34:48,364
different perspectives and different things to be able to be shown to,

479
00:34:48,404 --> 00:34:49,704
to kind of go through that process.

480
00:34:49,844 --> 00:34:52,064
So I asked for five, just simply five.

481
00:34:52,784 --> 00:34:56,724
And so as a base off of these, give us, do five definitions.

482
00:34:56,904 --> 00:34:59,784
And it gave us, it gave us five different definitions.

483
00:35:00,224 --> 00:35:08,464
Now I'm not a part of the committee but i do of the odd assist things such as that and.

484
00:35:09,909 --> 00:35:12,989
So that's what, that's what I ended up doing for the, for that time.

485
00:35:13,189 --> 00:35:18,989
And then another thing that I did off of that was based off of each of one of

486
00:35:18,989 --> 00:35:24,169
the stories, because everyone had to, the reception of it was pretty high.

487
00:35:24,309 --> 00:35:30,209
It got ranked as far as the LCI scores, as far as meeting participations and

488
00:35:30,209 --> 00:35:31,969
things like that. So that was pretty good.

489
00:35:32,189 --> 00:35:36,249
And one of the things that I, feedback was, is we want more programming,

490
00:35:36,409 --> 00:35:43,649
and we wish the section was longer and so i said based off of the stories that

491
00:35:43,649 --> 00:35:49,609
were shared and people seem to enjoy those let's come up with curriculum based

492
00:35:49,609 --> 00:35:53,929
off these stories and so i what i did was is i categorized.

493
00:35:54,669 --> 00:35:58,809
Asked it to categorize a few of the stories to say what's similar to each other

494
00:35:58,809 --> 00:36:05,049
and then build a program off of that now what the respect for people committee

495
00:36:05,049 --> 00:36:08,089
does after that with the information and things like that.

496
00:36:08,169 --> 00:36:10,449
That's up to them as far as what they want to do with it.

497
00:36:10,609 --> 00:36:16,969
But I did try to start doing the lesson planning process of this.

498
00:36:17,409 --> 00:36:23,489
And going back to my professor friend, he actually now uses AI for both of them actually do.

499
00:36:23,589 --> 00:36:28,709
They both use AI for lesson planning, which is a big time suck for teachers.

500
00:36:28,929 --> 00:36:34,109
Those who are educators definitely try to see how, what ways you can be able to do that.

501
00:36:34,229 --> 00:36:37,449
And that was definitely something that they ended up doing throughout their

502
00:36:37,449 --> 00:36:41,229
process is being able to use AI as a subsistence to build out a lesson plan.

503
00:36:41,369 --> 00:36:42,549
So it's a lot easier for themselves.

504
00:36:43,309 --> 00:36:47,609
Yeah, that's a great story. And even the AI talk that I was giving where I talked

505
00:36:47,609 --> 00:36:51,069
about the researcher asking questions at the end, I tell people in the beginning,

506
00:36:51,189 --> 00:36:57,089
like, this entire presentation was started with me having a conversation with

507
00:36:57,089 --> 00:37:00,589
GPT-3, 5 in the beginning.

508
00:37:00,669 --> 00:37:05,049
And then we later, 4-0 came out, GPT-4 came out later.

509
00:37:05,049 --> 00:37:08,049
And so it started a year ago and built

510
00:37:08,049 --> 00:37:11,169
it in three five including the slides the

511
00:37:11,169 --> 00:37:14,269
flow the timing it's a two-hour interactive presentation

512
00:37:14,269 --> 00:37:20,589
and then we updated it with four gpt4 and we had feedback we had sessions we

513
00:37:20,589 --> 00:37:24,689
did a repeat so we had in our community in california the community said we

514
00:37:24,689 --> 00:37:29,229
want this ai presentation again like they wanted a replay live and so we did

515
00:37:29,229 --> 00:37:32,389
it two more times so So four times total, Sacramento,

516
00:37:32,789 --> 00:37:37,369
San Francisco, and then Sacramento Bay, and then Peninsula Bay Area.

517
00:37:38,009 --> 00:37:42,949
And in the second, the version, the second and third, second and fourth time,

518
00:37:43,009 --> 00:37:45,269
I'm sorry, third and fourth time, got to get my math right,

519
00:37:45,369 --> 00:37:49,729
third and fourth time, one of the feedback comments we got from the audience,

520
00:37:49,809 --> 00:37:54,889
this is one of the owners that builds over a billion dollars a year in hospitals

521
00:37:54,889 --> 00:37:56,569
here in California, said,

522
00:37:56,789 --> 00:37:58,729
you know, the presentation was very positive.

523
00:37:58,729 --> 00:38:02,189
And I'm not surprised, because you asked AI to help you make it.

524
00:38:02,249 --> 00:38:05,349
It gave a very positive spin. What about the pitfalls?

525
00:38:06,300 --> 00:38:10,980
And so in version, the third and fourth version, we've now balanced the,

526
00:38:10,980 --> 00:38:14,360
the down, the scary parts. I'm just going to call it the scary parts of AI.

527
00:38:15,660 --> 00:38:21,900
And one of the people, we had a electrical contractor, JP, shout out to JP was

528
00:38:21,900 --> 00:38:23,860
in the, he went to all the sessions.

529
00:38:24,080 --> 00:38:27,020
So every time it was there, he he's, he's seen all of it.

530
00:38:27,040 --> 00:38:35,120
And he commented to us like number one, shockingly how informative it is and like how we present it.

531
00:38:35,120 --> 00:38:37,940
And then the second thing was he's like

532
00:38:37,940 --> 00:38:40,820
and he mentions it every time now because i've seen him like at two more events

533
00:38:40,820 --> 00:38:45,380
since then he says every time he thinks about the updated version that shows

534
00:38:45,380 --> 00:38:51,120
the dark side of ai he said it gives him nightmares like while he's awake every

535
00:38:51,120 --> 00:38:59,360
time it's it's scary because the the dark side it can go to the worst parts of humanity

536
00:38:59,620 --> 00:39:02,660
the things that it can do and people always they

537
00:39:02,660 --> 00:39:05,380
will assign the dark sides of

538
00:39:05,380 --> 00:39:08,400
ai like to the ai itself and they forget that there's

539
00:39:08,400 --> 00:39:13,860
people prompting it engaging with it getting it to do these things and so there

540
00:39:13,860 --> 00:39:17,720
are some like scary stuff about it but it was interesting to see like how we

541
00:39:17,720 --> 00:39:25,520
took the same thing and we iterated with a a new version of it and made it totally balanced.

542
00:39:26,180 --> 00:39:29,580
For people, but it took like, I'm, you know, I call it human in the loop.

543
00:39:29,660 --> 00:39:34,160
I didn't, I didn't make that up, but that's a, in the space people talk about,

544
00:39:34,220 --> 00:39:37,900
don't forget the person inside, like you and you said, you did all this stuff

545
00:39:37,900 --> 00:39:40,400
with the respect for people task force.

546
00:39:40,440 --> 00:39:44,720
And then you kicked it back to the humans in the loop that are responsible for

547
00:39:44,720 --> 00:39:47,480
implementing, you know, where they go with that content. Yeah.

548
00:39:48,106 --> 00:39:51,986
That's going to be up to them. So it's not, it just, it's not automatically

549
00:39:51,986 --> 00:39:53,886
like doing something on its own.

550
00:39:54,306 --> 00:39:58,726
Even though with, even though there are now auto AIs and, you know,

551
00:39:58,726 --> 00:40:01,966
that type of stuff where we could talk about if you want to,

552
00:40:01,966 --> 00:40:05,266
you know, some of the scarier things that can happen.

553
00:40:05,446 --> 00:40:09,006
I mean, there's some, I mean, I've messed around with auto GPT,

554
00:40:09,146 --> 00:40:13,186
which was supposed to complete tasks for you based off of the questions that you ask.

555
00:40:13,226 --> 00:40:17,646
And what I realized was it ran up my API credits and it never completed the

556
00:40:17,726 --> 00:40:21,446
task it's so you know it's

557
00:40:21,446 --> 00:40:24,286
a good yeah for me it let me know that it's a long way

558
00:40:24,286 --> 00:40:26,986
to go from there some people try to so you know i looked on

559
00:40:26,986 --> 00:40:29,826
twitter and saw some use cases and i was like okay let me try that you can use

560
00:40:29,826 --> 00:40:34,506
case i tried it didn't work and i'm just like you know this is a great experiment

561
00:40:34,506 --> 00:40:38,106
i think it's a great thought experiment you can probably take bits and pieces

562
00:40:38,106 --> 00:40:42,706
from this and be able to do some stuff but you know for those who are super

563
00:40:42,706 --> 00:40:48,066
super deep into ai and agents and doing these things,

564
00:40:48,266 --> 00:40:50,826
there's some things I can definitely tell you to look into.

565
00:40:51,046 --> 00:40:54,766
I'm not going to go into detail about all of them, but Crew.ai,

566
00:40:55,086 --> 00:40:59,306
Autogen, and Llama Index,

567
00:40:59,666 --> 00:41:07,026
and Langchain all have agent frameworks that you can build upon that you can

568
00:41:07,026 --> 00:41:09,326
make these agents string into each other.

569
00:41:09,926 --> 00:41:18,446
Why is this important? Well, if I was giving an agent, if I was giving one AI

570
00:41:18,446 --> 00:41:21,426
one thing to be able to do the task,

571
00:41:21,746 --> 00:41:25,226
now the AI can critique itself using another AI.

572
00:41:25,446 --> 00:41:30,846
And now another agent, whereas we were talking about role-playing and why role-playing

573
00:41:30,846 --> 00:41:36,006
was very important, now it's role-playing different perspectives to be able

574
00:41:36,006 --> 00:41:37,186
to create some of these things.

575
00:41:37,926 --> 00:41:42,866
Now, I can actually transition this into what I was going to show Felipe.

576
00:41:43,006 --> 00:41:48,166
And one of the reasons why this show came about is I was excited about some

577
00:41:48,166 --> 00:41:52,446
AI features, and I wanted to show Felipe some of the things and some of the

578
00:41:52,446 --> 00:41:53,786
variables that I went down.

579
00:41:54,126 --> 00:42:01,486
And one of these things is showing how an AI agent can actually work for people

580
00:42:01,486 --> 00:42:07,486
to be able to do that. And so that's a great, good transition into what's,

581
00:42:07,486 --> 00:42:08,966
what I'm going to share my screen upon.

582
00:42:09,486 --> 00:42:13,126
Yeah, let's do it. Now, Marcus shares a screen just for all you know,

583
00:42:13,166 --> 00:42:17,506
while he's finagling the share button, just the big green button on the bottom.

584
00:42:17,766 --> 00:42:24,446
I tried to get him on the show for a while. And now that this topic finally unlocked the key.

585
00:42:25,086 --> 00:42:28,066
So now we're here. So, and you all get to learn with me at the same time.

586
00:42:28,106 --> 00:42:30,866
Like I have not seen what he's going to share. I'm seeing this for the first time myself.

587
00:42:33,006 --> 00:42:36,766
We hear all the time that building enclosure is the number one problem in construction.

588
00:42:37,166 --> 00:42:41,026
And when the task of the building enclosure is presented with so much pressure,

589
00:42:41,146 --> 00:42:43,506
it's something most of us would want to avoid.

590
00:42:45,544 --> 00:42:50,084
Until now, there has never been a course to teach someone how to manage a building enclosure.

591
00:42:50,424 --> 00:42:54,104
Building enclosure is a problem for every company, and the solution is often

592
00:42:54,104 --> 00:42:59,624
to ask someone within to try to find training or provide training to their teams to avoid these issues.

593
00:43:00,064 --> 00:43:03,604
Field Verified's Enclosure Management course is the answer to that problem.

594
00:43:03,864 --> 00:43:07,744
We learned the hard way. We've learned through failure. We've learned from others.

595
00:43:08,404 --> 00:43:12,624
This course, our team, will help those people who are assigned to manage the

596
00:43:12,624 --> 00:43:16,084
building enclosure be successful the first time so that they're spared some

597
00:43:16,084 --> 00:43:18,744
of the suffering and hardships that all of us went through.

598
00:43:19,044 --> 00:43:21,824
There's so much opportunity for a high return on investment,

599
00:43:22,084 --> 00:43:25,864
not just when an individual attends and all of the skills that they'll achieve,

600
00:43:26,064 --> 00:43:30,004
but when a team participates together they can raise their level of performance

601
00:43:30,004 --> 00:43:32,644
by knowing what each other can do, what to expect,

602
00:43:32,864 --> 00:43:36,204
and they can raise the level of performance of that company as their careers

603
00:43:36,204 --> 00:43:38,644
excel and the company elevates along with them.

604
00:43:39,224 --> 00:43:42,984
The training is built on the science of how people learn. We capture all of

605
00:43:42,984 --> 00:43:45,324
the components that are necessary to master a skill.

606
00:43:45,524 --> 00:43:48,824
There's a hands-on component to the training where we're building a masonry

607
00:43:48,824 --> 00:43:53,204
mock-up, a metal panel mock-up, we're doing EFIS, and even installing a window

608
00:43:53,204 --> 00:43:54,704
that later you get to test.

609
00:43:54,944 --> 00:43:58,484
The hands-on training gives us the opportunity to translate the lines on paper

610
00:43:58,484 --> 00:44:00,084
to actual construction in the field.

611
00:44:00,584 --> 00:44:03,704
We always laugh because after the training, people come up and say,

612
00:44:03,844 --> 00:44:05,644
this isn't just about building enclosure.

613
00:44:05,984 --> 00:44:08,344
There's a lot of leadership in here. There are a lot of people skills.

614
00:44:08,584 --> 00:44:13,144
That's because those skills are necessary to be successful in building enclosure and in construction.

615
00:44:13,564 --> 00:44:17,184
We'd love to see you get started right away. We have lots of options available.

616
00:44:17,404 --> 00:44:21,444
They can come to us in person at Field Verified in Phoenix, or we can bring

617
00:44:21,444 --> 00:44:25,784
the entire operation to your facility and train up to 25 people at a time.

618
00:44:25,884 --> 00:44:29,524
And we've created an online training that's available for people who might not

619
00:44:29,524 --> 00:44:30,824
be able to attend in person.

620
00:44:30,964 --> 00:44:35,044
Go to the website and contact us. We're gonna start by understanding what problems

621
00:44:35,044 --> 00:44:38,984
you're currently experiencing and develop a specific program that's right for

622
00:44:38,984 --> 00:44:41,704
the needs of your company. Yeah.

623
00:44:43,664 --> 00:44:48,324
Everyone's seen a preview about how me and Felipe will talk back and forth now.

624
00:44:48,504 --> 00:44:54,084
So Felipe, is there like, okay, so I'm going to ask you a question and then

625
00:44:54,084 --> 00:44:56,664
we can kind of go with it and I'll just type it in. Right. Okay.

626
00:44:57,820 --> 00:45:03,900
So this is kind of using, Storm is made by Stanford, which is what they ended

627
00:45:03,900 --> 00:45:06,060
up doing with creating a research assistant,

628
00:45:06,380 --> 00:45:10,200
which is a whole bunch of AI agents that were playing a role,

629
00:45:10,440 --> 00:45:14,120
chained together using a tool. And the tool is the internet.

630
00:45:14,500 --> 00:45:16,840
And it's a few things. One is called Archives.

631
00:45:17,280 --> 00:45:22,000
One's actually called, they actually use GitHub as well to gather a lot of information

632
00:45:22,000 --> 00:45:27,760
and and be able to create Wikisite argument articles based for research.

633
00:45:28,980 --> 00:45:33,320
So Felipe, is there anything that you would, they asking me to,

634
00:45:34,100 --> 00:45:40,060
there's a limitation to how long the topic could be, but can you send me a topic

635
00:45:40,060 --> 00:45:43,360
of something that you may be interested in? And then we can just kind of go from there.

636
00:45:43,640 --> 00:45:50,760
Yeah, I'm interested in a Scrum. So let's ask it about how industries are using

637
00:45:50,760 --> 00:45:53,020
Scrum in their management practices.

638
00:45:54,220 --> 00:45:56,640
And for those of you watching the show, you should know that I love Scrum.

639
00:45:56,740 --> 00:46:01,340
And I have a dedicated playlist just to all the Scrum episodes that happen.

640
00:46:01,440 --> 00:46:04,200
So I think I just forced this one to the playlist.

641
00:46:07,060 --> 00:46:14,800
Okay. So do you want me to elaborate or should I just kind of ask that question and kind of go with it?

642
00:46:15,380 --> 00:46:19,200
Yeah, I think let's just go with it without too much elaboration, see what we got. Okay.

643
00:46:20,240 --> 00:46:23,420
Oh, I'm just going to say you have to elaborate. I'm just going to do the same

644
00:46:23,420 --> 00:46:25,320
thing. Just type right here. Okay.

645
00:46:25,980 --> 00:46:28,400
And I'll change that. I want to understand.

646
00:46:29,760 --> 00:46:33,400
There you go. So he's taking that prompt in for those of you just listening.

647
00:46:33,520 --> 00:46:38,180
I want to understand how industries are using scrum management practices is the elaboration.

648
00:46:38,280 --> 00:46:41,400
So definitely check this out on YouTube. If you're only listening to the audio

649
00:46:41,400 --> 00:46:43,820
version of the podcast, love you people.

650
00:46:43,920 --> 00:46:47,660
And while you're there, it's not going to hurt you to hit the like button and

651
00:46:47,660 --> 00:46:50,820
leave us a comment. Marcus and I will definitely be watching the comments on

652
00:46:50,820 --> 00:46:52,400
the show and answering your questions.

653
00:46:52,440 --> 00:46:56,460
So if you've got something and you want a deeper dive on, hit us up in the comments

654
00:46:56,460 --> 00:46:59,080
either on YouTube or wherever you're listening to this podcast.

655
00:47:01,007 --> 00:47:03,627
You know, it takes about three minutes to be able to generate the article.

656
00:47:04,847 --> 00:47:09,667
While I do that, I can just sit here and wait for that.

657
00:47:10,407 --> 00:47:15,627
I'm not sure. Yeah. I guess your editor will probably edit, cut this part out

658
00:47:15,627 --> 00:47:17,387
until I get to the actual answer.

659
00:47:17,547 --> 00:47:20,167
Unless we talk about, unless we talk about something interesting.

660
00:47:20,727 --> 00:47:24,687
Okay. And then otherwise, yeah. So like shout out to editor Geronimo.

661
00:47:24,847 --> 00:47:28,707
Thank you for making the show so beautiful. Super appreciate what you're doing.

662
00:47:28,707 --> 00:47:33,067
And if you're listening, this is how we'll know if Geronimo actually listens to the show.

663
00:47:37,727 --> 00:47:42,267
If not, people are going to just get free banter. It's all right. Yeah.

664
00:47:42,707 --> 00:47:47,767
This is a free banter. So, yeah, like as fast as AI is, I could tell this the

665
00:47:47,767 --> 00:47:48,807
first time I'm seeing Storm.

666
00:47:49,407 --> 00:47:52,967
It's got the prompt says it's generating article may take up to three minutes.

667
00:47:52,967 --> 00:47:57,207
And then it says start identifying different perspectives for the researching

668
00:47:57,207 --> 00:47:58,687
of the topic steps one of four.

669
00:47:59,127 --> 00:48:03,627
So it looks like it's doing a multi-step thing. So if I was to guess what's

670
00:48:03,627 --> 00:48:07,167
happening behind the scenes, I would sense that it's, you know,

671
00:48:07,167 --> 00:48:08,587
searching what it's been trained on.

672
00:48:08,667 --> 00:48:11,387
Like you said, it's, it's trained in a couple of different areas.

673
00:48:11,527 --> 00:48:13,907
What is, what are the limitations of where it's trained? Can it go to,

674
00:48:13,947 --> 00:48:17,187
does this go to the active internet now or is it going to?

675
00:48:17,307 --> 00:48:20,907
Yes, this is going to the active internet now. So one of the things I can do

676
00:48:20,907 --> 00:48:26,467
is actually show some articles that I actually did using the downloaded version.

677
00:48:26,647 --> 00:48:28,247
So this is completely open sourced.

678
00:48:29,147 --> 00:48:31,907
The one I was showing before is a web UI.

679
00:48:32,687 --> 00:48:35,607
So one of the things I can do is just kind of click one of these things.

680
00:48:35,767 --> 00:48:42,987
Since we talk about IPD for a little bit of this show, I have to click one based off of that.

681
00:48:43,187 --> 00:48:48,427
So I'm just going to go ahead and click this real quick and kind of scroll through it.

682
00:48:48,427 --> 00:48:54,447
Um pretty quickly and so as you can see there's like a table of contents there's

683
00:48:54,447 --> 00:48:58,767
links so these actually i don't know if you can see the bottom left let's see

684
00:48:58,767 --> 00:49:00,247
it cited sources got references.

685
00:49:01,447 --> 00:49:04,607
So you know which is really good references

686
00:49:04,607 --> 00:49:15,087
that's impressive yeah it's uh 19 19 references hot dog yeah and you can kind

687
00:49:15,087 --> 00:49:18,327
of take this and then kind of read the read over this and then and dive back

688
00:49:18,327 --> 00:49:22,347
in to go deeper into your research to be able to figure out some more things, right?

689
00:49:22,487 --> 00:49:24,947
Because this is creating an article, it's creating information.

690
00:49:25,867 --> 00:49:30,147
From there, you can actually see what questions were being asked.

691
00:49:31,156 --> 00:49:34,876
Oh my God. Like, hold on. As you were scrolling, I did, I did read one sentence

692
00:49:34,876 --> 00:49:37,016
about the challenges right there.

693
00:49:37,096 --> 00:49:39,976
Challenges implementing AI within IPD frameworks.

694
00:49:40,176 --> 00:49:46,976
It's, it's saying that it's going to be hard because an obstacle is traditional

695
00:49:46,976 --> 00:49:51,796
methods and manual repetitive tasks that affect.

696
00:49:51,796 --> 00:49:54,556
Effect and it was like as i was reading it

697
00:49:54,556 --> 00:49:57,776
i thought that was pretty spot on where

698
00:49:57,776 --> 00:50:01,316
it's talking about the significant challenges you know

699
00:50:01,316 --> 00:50:04,596
people are so triggered and traumatized marcus in construction

700
00:50:04,596 --> 00:50:07,476
like there is no shortage of trauma

701
00:50:07,476 --> 00:50:10,656
written about how design and construction projects come

702
00:50:10,656 --> 00:50:13,796
together and even in this you know 19 reference

703
00:50:13,796 --> 00:50:17,656
article thing on ai with ipd it's

704
00:50:17,656 --> 00:50:20,436
it's warning the the reader which is us

705
00:50:20,436 --> 00:50:23,436
now that it's going to be hard because of these

706
00:50:23,436 --> 00:50:28,996
traditional methods that are really difficult and a lot of the tasks are really

707
00:50:28,996 --> 00:50:33,476
manual there's not a lot of automation and process improvement in the construction

708
00:50:33,476 --> 00:50:38,816
space and i had uh you know our mutual friend sean graystone on the show a couple

709
00:50:38,816 --> 00:50:41,696
seasons ago and he talked about the way that we

710
00:50:41,736 --> 00:50:44,696
build today is based on methods from before

711
00:50:44,696 --> 00:50:47,416
the civil war and if you go if

712
00:50:47,416 --> 00:50:51,916
you some if a construction historian went back in time to the civil war to see

713
00:50:51,916 --> 00:50:57,196
how projects get done and they they came back to 2024 it would look exactly

714
00:50:57,196 --> 00:51:01,936
the same like the same processes some of the the companies change and the names

715
00:51:01,936 --> 00:51:05,176
change but the the roles and the functions and the processes are

716
00:51:05,216 --> 00:51:07,576
still there intact the same.

717
00:51:07,776 --> 00:51:11,976
And it's, it creates like that environment, like you said earlier about how

718
00:51:11,976 --> 00:51:16,256
from that quote, people, strangers come together and they, after working together

719
00:51:16,256 --> 00:51:17,956
for a couple of years, they leave as enemies.

720
00:51:18,536 --> 00:51:22,216
Like it creates a system that just turns people against each other sometimes.

721
00:51:22,416 --> 00:51:25,916
And you know, there's crazy people like myself and you and other people,

722
00:51:25,996 --> 00:51:30,296
Brittany, for sure, that we think we don't have to fight with each other to get stuff done.

723
00:51:30,416 --> 00:51:33,256
We can actually work together. There is another way.

724
00:51:34,469 --> 00:51:40,109
There's very few stories of dysfunctional sports teams that somehow won a championship in the end.

725
00:51:41,789 --> 00:51:43,049
I don't know of any.

726
00:51:44,929 --> 00:51:49,009
So what's this down here at the bottom now? So if you wanted to kind of understand

727
00:51:49,009 --> 00:51:52,529
the thought process of what was the type of questions that are being asked.

728
00:51:52,829 --> 00:51:57,849
Right. So you can see what the answers are based off of what the and this is

729
00:51:57,849 --> 00:52:00,609
kind of using the A.I.'s common knowledge.

730
00:52:00,609 --> 00:52:07,709
And it's asking it basically playing a role being able to do that and then they

731
00:52:07,709 --> 00:52:11,429
even it'll even show like where it says hey i don't have enough information

732
00:52:11,429 --> 00:52:15,869
to answer this question right so you're going through this so you're seeing

733
00:52:15,869 --> 00:52:19,089
what the thought process is there right and then,

734
00:52:19,729 --> 00:52:22,669
construction manager there's different questions that are being asked as

735
00:52:22,669 --> 00:52:25,389
well so i think and then what's the

736
00:52:25,389 --> 00:52:28,149
other role ai specialist so they created the ai

737
00:52:28,149 --> 00:52:31,329
so it created it three roles an ipd expert a basic

738
00:52:31,329 --> 00:52:34,529
effect writer construction manager and an ai specialist

739
00:52:34,529 --> 00:52:39,029
so those are the roles that the ai generated from itself based off the question

740
00:52:39,029 --> 00:52:44,909
that i asked based off of one sentence so that's impressive yeah but it's chained

741
00:52:44,909 --> 00:52:48,349
together a lot of different prompts to be able to do that so you're chaining

742
00:52:48,349 --> 00:52:52,269
together a whole bunch of different prompts within ai and,

743
00:52:52,989 --> 00:52:56,229
so it's being able to understand what the question is to be able to create the

744
00:52:56,229 --> 00:53:00,009
prompts to be able to generate the personas that are being needed so this is

745
00:53:00,009 --> 00:53:02,409
kind of like an auto persona generator as well.

746
00:53:05,239 --> 00:53:09,299
And let's try to see if we can go back to okay

747
00:53:09,299 --> 00:53:12,899
that is still going okay i'm glad

748
00:53:12,899 --> 00:53:16,179
we showed this so and this is

749
00:53:16,179 --> 00:53:19,059
real time so so people can understand that like

750
00:53:19,059 --> 00:53:22,079
hey you know things don't always show up how they're supposed to

751
00:53:22,079 --> 00:53:25,119
i'm still going to let it run in the background to try to because i'm

752
00:53:25,119 --> 00:53:28,639
curious to see what the answer is but you're going

753
00:53:28,639 --> 00:53:31,879
through all those roles and so it's got four personas

754
00:53:31,879 --> 00:53:35,919
in the article about ipd that you did previously yeah

755
00:53:35,919 --> 00:53:38,799
and the as marcus is scrolling if you're if you're not watching this

756
00:53:38,799 --> 00:53:41,779
video live like the questions and answers are

757
00:53:41,779 --> 00:53:44,619
significant like this could be you know any

758
00:53:44,619 --> 00:53:47,899
one of these questions and answers in these personas could have

759
00:53:47,899 --> 00:53:50,979
been your session that you've done and in an ai interface

760
00:53:50,979 --> 00:53:54,559
like a online version of gpt or claude anthropics

761
00:53:54,559 --> 00:53:58,119
claude or gemini or pick your favorite and but

762
00:53:58,119 --> 00:54:00,919
then it keeps going so it gets a question it gets

763
00:54:00,919 --> 00:54:04,079
an answer and then the that persona that agent

764
00:54:04,079 --> 00:54:07,039
asks another question so there's a follow-up so

765
00:54:07,039 --> 00:54:09,879
it keeps having that dialogue and then the output that we saw

766
00:54:09,879 --> 00:54:14,459
when marcus first scrolled was like the the ready-built article and the article

767
00:54:14,459 --> 00:54:19,839
is a synthesis of all these four personas and in this example asking questions

768
00:54:19,839 --> 00:54:28,159
how many questions roughly do we see in this persona like four five i think it's about one two three.

769
00:54:29,399 --> 00:54:32,239
So three yeah three big questions and answers

770
00:54:32,239 --> 00:54:36,139
on the specialist and then you know it does it's doing the same thing for the

771
00:54:36,139 --> 00:54:40,679
other agents i think that's that's that's a lot like to do and keep it all together

772
00:54:40,679 --> 00:54:44,879
with memory so this is one of the things that in the early days of ai it didn't

773
00:54:44,879 --> 00:54:50,379
have memory who to really have like consistent dialogues with, and now, you.

774
00:54:50,862 --> 00:54:54,702
Here's an example where it's created an environment and it's using,

775
00:54:54,882 --> 00:54:59,762
you know, it's having four different conversations, probably almost at the same time.

776
00:54:59,962 --> 00:55:04,422
And then it's reading all those conversations simultaneously and creates this

777
00:55:04,422 --> 00:55:08,182
synthesized output based on these four different conversations.

778
00:55:08,382 --> 00:55:12,222
All started with one prompt, one input from a human.

779
00:55:12,682 --> 00:55:16,542
Yeah. And I think that's the nice thing.

780
00:55:16,542 --> 00:55:20,922
And if those were down the rabbit hole of AI agents and working with different

781
00:55:20,922 --> 00:55:25,882
AI personas to get it to collaborate and come up with, you know,

782
00:55:25,882 --> 00:55:29,142
document generation, that's a whole other use case.

783
00:55:29,262 --> 00:55:32,762
Probably the use case that people are going to mainly feature is it's going

784
00:55:32,762 --> 00:55:39,262
to be some type of AI to document generation or AI to research assistant at this point.

785
00:55:39,262 --> 00:55:44,122
Yeah, we talked about this in the LCI AI talk that we give here in California,

786
00:55:44,242 --> 00:55:46,442
that for content creation,

787
00:55:46,842 --> 00:55:52,542
this is a definite use case that if you don't know how to talk to it to do this

788
00:55:52,542 --> 00:55:54,482
type of stuff, then welcome.

789
00:55:54,562 --> 00:55:57,802
Now, you know, it can be done. This is one of the early cases,

790
00:55:57,902 --> 00:56:03,022
like a lot of people are using even Copilot in Word, in email.

791
00:56:03,182 --> 00:56:06,822
So some people, their companies have turned it on to help them,

792
00:56:06,822 --> 00:56:11,142
you know, write letters or to, you know, create content, you know,

793
00:56:11,162 --> 00:56:14,702
based on something you want to communicate because it is so good at writing.

794
00:56:14,822 --> 00:56:17,942
And it's, you know, typically grammatically correct.

795
00:56:18,402 --> 00:56:23,942
It's logical. And then there's a, we did this in the updated version.

796
00:56:23,942 --> 00:56:26,942
Version there's a test that they

797
00:56:26,942 --> 00:56:29,962
they run with not the not the turing test to

798
00:56:29,962 --> 00:56:32,742
see if like you can trick a person into thinking you're not

799
00:56:32,742 --> 00:56:35,482
talking to a robot or a computer but there's

800
00:56:35,482 --> 00:56:38,862
another test where they show like just average intelligence of an

801
00:56:38,862 --> 00:56:46,422
adult and gpt 40 has surpassed average intelligence of adult so you'll see as

802
00:56:46,422 --> 00:56:50,142
you talk to it and i i found this you know talking to it for years now like

803
00:56:50,142 --> 00:56:53,642
i've grown up with gpt like it's a you know like a member of my family like

804
00:56:53,642 --> 00:56:57,202
Like it started off as like a toddler, maybe a little.

805
00:56:57,222 --> 00:57:00,642
It started off when it got released a little bit beyond a toddler.

806
00:57:00,942 --> 00:57:05,542
And now it's like talking to like a 30-year-old.

807
00:57:05,882 --> 00:57:10,022
And in some instances, like with programming, in programming,

808
00:57:10,102 --> 00:57:13,222
it's like talking to a 20-year veteran software programmer.

809
00:57:13,762 --> 00:57:18,682
So the experience in some of the domains is very high. Very high.

810
00:57:19,902 --> 00:57:24,182
Trying to see. Okay. Okay, so actually, it's actually still going.

811
00:57:24,182 --> 00:57:26,822
It's on step four. Yeah, but it's on step four now, so it's right in the article.

812
00:57:27,782 --> 00:57:31,702
Yeah, we're almost there. The next thing I want to be able to show you,

813
00:57:31,742 --> 00:57:35,882
I don't want to go over it for this one yet, but I can show a different one

814
00:57:35,882 --> 00:57:37,902
where I can go to my articles.

815
00:57:39,484 --> 00:57:43,284
And I love how this article, if you ever talk to a researcher at the bottom

816
00:57:43,284 --> 00:57:45,804
of every research, at the bottom of every good research paper,

817
00:57:45,984 --> 00:57:50,024
it always ends with some nonsense about more research is necessary.

818
00:57:51,304 --> 00:57:56,044
So, like, I think it's so funny. I'm just calling out all the researchers.

819
00:57:56,124 --> 00:57:58,584
And I've had a few academics on the program.

820
00:57:58,984 --> 00:58:04,044
I'm calling you all out that you're planting seeds in every one of your works to get more works.

821
00:58:04,764 --> 00:58:09,004
So, yeah, you're baiting everybody. everybody, like research is needed.

822
00:58:09,624 --> 00:58:13,644
I mean, some implementation and then some research on the implementation would

823
00:58:13,644 --> 00:58:16,164
be needed. I think that would be actually a really good thing.

824
00:58:16,684 --> 00:58:22,964
One of the things, so there's this one that we're, for those who don't know,

825
00:58:23,064 --> 00:58:26,604
one of the interests of ours is exploring and

826
00:58:26,604 --> 00:58:30,004
we built a company around it was exploring blockchain

827
00:58:30,004 --> 00:58:33,064
and construction and builder pay

828
00:58:33,064 --> 00:58:36,584
is one of the solutions or build a chain is one of the solutions to

829
00:58:36,584 --> 00:58:39,544
be able to do that but and britney's written

830
00:58:39,544 --> 00:58:43,484
one paper that actually went to parliament parliament in

831
00:58:43,484 --> 00:58:48,304
the uk as far as their thoughts and ideas of blockchain within the use of supply

832
00:58:48,304 --> 00:58:52,924
chain so they talk about the legal aspects and information handling aspects

833
00:58:52,924 --> 00:58:59,604
and stuff like that one of the things that i did was i just kind of asked a

834
00:58:59,604 --> 00:59:02,644
a little bit more like kind of a vague overall question.

835
00:59:02,764 --> 00:59:06,324
And I just wanted to see what research and what research was out there.

836
00:59:06,784 --> 00:59:11,464
And I was actually surprised of how much stuff and what they were looking for and things like that.

837
00:59:12,264 --> 00:59:16,444
And it pulled different spaces. And I was like, oh, okay.

838
00:59:16,724 --> 00:59:20,944
So it's still like, you know, they're still talking about like,

839
00:59:21,004 --> 00:59:22,784
yes, this stuff is being talked about,

840
00:59:23,524 --> 00:59:28,804
but some of the same problems that are still there after, I think,

841
00:59:28,804 --> 00:59:33,324
You know six years later of being able to write that article that went to Parliament.

842
00:59:34,164 --> 00:59:37,804
You know government takes a while to implement things just right say,

843
00:59:39,284 --> 00:59:43,424
That's a matter where the government is takes multiple election cycles to get

844
00:59:43,424 --> 00:59:48,924
something done Exactly, and then you know and another party still may block it for you. So yeah.

845
00:59:50,602 --> 00:59:53,862
So I thought this was a really good one.

846
00:59:54,922 --> 01:00:02,282
Yeah, let's check back. Okay, step four. But that Scrum article is just exhausting. The storm.

847
01:00:02,702 --> 01:00:09,562
Well, I'm using somebody else's UI, using somebody else's API keys.

848
01:00:10,422 --> 01:00:14,242
So it's going to take a while for that one.

849
01:00:14,502 --> 01:00:20,802
But at the same time, I was at least smart enough to prepare this before the

850
01:00:20,802 --> 01:00:23,502
call and show it from a local standpoint.

851
01:00:23,942 --> 01:00:30,162
And this is kind of how those who are watching are seeing the Storm one on the Stanford website.

852
01:00:30,582 --> 01:00:33,542
It's very pretty, nice, and looks really, really awesome.

853
01:00:34,342 --> 01:00:38,322
But the one that you use for yourself doesn't necessarily look as well.

854
01:00:38,702 --> 01:00:44,442
I did have some problems at first trying to install this, but if someone wanted

855
01:00:44,442 --> 01:00:47,462
to install this and wanted to go through the process or anything like that,

856
01:00:47,662 --> 01:00:49,542
feel free to hit me up on LinkedIn.

857
01:00:49,762 --> 01:00:52,662
And i'll kind of walk you through the process it is going to cost an

858
01:00:52,662 --> 01:00:55,622
api two api keys one of

859
01:00:55,622 --> 01:00:59,182
them eventually will be a hundred dollars a hundred dollars a month so you may

860
01:00:59,182 --> 01:01:03,942
not want to use it for you may want to get your uses in early but it has a free

861
01:01:03,942 --> 01:01:10,602
trial for 60 days and then the other one it does cost a gpt4 but there is a

862
01:01:10,602 --> 01:01:14,642
way you can do it for free i didn't do it for free but so So,

863
01:01:14,762 --> 01:01:17,422
but it does cost a GPT-4 API key.

864
01:01:17,922 --> 01:01:22,182
So it is, I am hitting my API credits by using this. However,

865
01:01:22,342 --> 01:01:24,422
the other one I'm using is somebody else's API key.

866
01:01:25,822 --> 01:01:29,102
Yeah, that's what we're going to appreciate, the API key sharing.

867
01:01:29,362 --> 01:01:33,322
And yeah, thank you, Marcus, for spending some API dollars with me on one of

868
01:01:33,322 --> 01:01:34,622
my favorite topics in the world.

869
01:01:34,682 --> 01:01:39,822
We'll make sure that the output of this article gets turned into a blog post

870
01:01:39,822 --> 01:01:43,222
and we'll attach it to this interview so that in the show notes,

871
01:01:43,222 --> 01:01:47,922
So you'll see a link to that blog post and the output of that article.

872
01:01:48,282 --> 01:01:52,002
We'll put the version. I'll figure out a way to get it in so that people can

873
01:01:52,002 --> 01:01:54,982
see what it generated and some of the thought process.

874
01:01:55,102 --> 01:01:58,302
We'll expand it out a little bit. So what are we looking at now, Marcus?

875
01:01:59,002 --> 01:02:04,662
So this is for somebody who maybe is struggling with prompting or anything like that.

876
01:02:05,022 --> 01:02:10,902
Claude has created this platform called, they created their own prompt generator.

877
01:02:10,902 --> 01:02:13,962
So Claude is made by Anthropic.

878
01:02:14,062 --> 01:02:20,202
Anthropic is a company that split off from OpenAI and they wanted to pursue

879
01:02:20,202 --> 01:02:24,702
AI a certain way and OpenAI wanted to pursue it in a different way.

880
01:02:24,822 --> 01:02:30,062
So they're, I would say they're cousins, but, and, and in a way that's still,

881
01:02:30,242 --> 01:02:31,582
you know, they're competitors.

882
01:02:32,922 --> 01:02:39,602
So Claude has done a thing where they, they've generated an idea.

883
01:02:39,602 --> 01:02:43,702
They said, hey, you know, since prompt generating, prompting is really hard

884
01:02:43,702 --> 01:02:47,382
for people, let's figure out a way to make, you know, prompting a lot easier for people.

885
01:02:48,721 --> 01:02:57,161
So I'm going to take that same question that was asked from Felipe and...

886
01:02:58,161 --> 01:03:03,501
How are industries using Scrum and management practices? I'm just want to go

887
01:03:03,501 --> 01:03:07,821
to this, create a prompt, or

888
01:03:07,821 --> 01:03:11,261
should I take something else in Felipe and just make a prompt from that?

889
01:03:12,021 --> 01:03:16,221
I think we can, the task will be, cause this is the question I get asked a billion times.

890
01:03:16,761 --> 01:03:20,221
I'm new to Scrum. how can i as a

891
01:03:20,221 --> 01:03:23,321
as a novice start using it in my management work

892
01:03:23,321 --> 01:03:25,961
that's the prompt so for i mean if those

893
01:03:25,961 --> 01:03:29,841
you're listening like i'm i'm beyond a decade of scrum use so this is not my

894
01:03:29,841 --> 01:03:34,581
question but it's a question i get asked all the time i just want to see like

895
01:03:34,581 --> 01:03:39,161
how good are we going to get from this yeah so this is just a basic question

896
01:03:39,161 --> 01:03:44,741
right so in like three seconds marcus Marcus now has like a fully generated prompt.

897
01:03:46,381 --> 01:03:50,701
So what I'm going to do nicely is post this in the chat.

898
01:03:52,121 --> 01:03:55,841
So not flip it. You can definitely read it out if you want to. Oops.

899
01:03:56,461 --> 01:03:59,481
Not sharing. That's the share my screen. Yeah. I'm going to check,

900
01:03:59,581 --> 01:04:01,861
uh, see what's coming up in the chat here.

901
01:04:01,981 --> 01:04:05,281
Marcus is going to share what got prompted. I'll read some of it.

902
01:04:06,021 --> 01:04:09,321
Yeah. What, uh, what came out of her? If you're watching this.

903
01:04:10,201 --> 01:04:14,341
Oh, it's actually too long. That's crazy. It's too long for Zoom chat.

904
01:04:14,921 --> 01:04:17,221
Yeah. I'm going to access my Google Drive.

905
01:04:18,161 --> 01:04:21,441
I should just ask my Google Drive not this way.

906
01:04:21,681 --> 01:04:25,401
What I can just try to do is this. I'm going to open up my notepad and then

907
01:04:25,401 --> 01:04:27,641
bring the screen over. Yeah. That'll work.

908
01:04:28,941 --> 01:04:32,061
Let me not share my screen when I'm accessing my Google Drive.

909
01:04:32,281 --> 01:04:33,401
I'll be right back, guys.

910
01:04:36,641 --> 01:04:41,261
We'll just cut that out of the show so that we don't show all your,

911
01:04:42,321 --> 01:04:47,221
yeah let's not make it easy for people to hack you get a little harder let's

912
01:04:47,221 --> 01:04:53,241
think about this for a second all right so i'm gonna go here,

913
01:04:55,581 --> 01:05:03,561
Go back to the share screen. Share. Okay, cool. Yeah, it's more than a page. Yeah. So.

914
01:05:05,901 --> 01:05:12,301
Yeah, probably. There we go. All right. So basically, hey, we're talking about

915
01:05:12,301 --> 01:05:14,821
the role, right? We're talking about that again, right?

916
01:05:15,121 --> 01:05:18,721
You're an AI-assisted task for helping a novice understand how to implement

917
01:05:18,721 --> 01:05:19,701
Scrum in their practices.

918
01:05:20,201 --> 01:05:24,021
You provide basic information about Scrum and a specific question to the user.

919
01:05:24,021 --> 01:05:27,001
Your goal is to provide clear structured and practical

920
01:05:27,001 --> 01:05:30,741
advice to help the user get started with scrum right that's

921
01:05:30,741 --> 01:05:34,521
great yeah first it gives some scrum basics next present

922
01:05:34,521 --> 01:05:37,501
them with the user's question and then it goes into analyzing

923
01:05:37,501 --> 01:05:40,581
and then some of those are code snippets so people looking at you know

924
01:05:40,581 --> 01:05:43,421
what that is that's code that the ai is going

925
01:05:43,421 --> 01:05:46,341
to understand and i'm sure that this prompt would

926
01:05:46,341 --> 01:05:49,121
work in any of the large language models so it

927
01:05:49,121 --> 01:05:53,061
doesn't have to go back to cloud necessarily but for those listening i do like

928
01:05:53,061 --> 01:05:57,661
the friendliness of claude i feel like between claude and gpt the personality

929
01:05:57,661 --> 01:06:03,921
that claude has is a little friendlier yeah and and i think that's one of the

930
01:06:03,921 --> 01:06:07,241
great things about being able to go back and forth with the large language models,

931
01:06:08,181 --> 01:06:11,701
and there's a really cool hack that i would love to talk about towards,

932
01:06:12,461 --> 01:06:17,641
felipe to kind of share with you so if you have some personal you can have some

933
01:06:17,641 --> 01:06:22,301
personal only imagination about this nice hack that I kind of found was with

934
01:06:22,301 --> 01:06:24,301
the for a workflow. Yeah.

935
01:06:26,061 --> 01:06:30,501
Okay, so we have the prompt generator that's here, right?

936
01:06:32,040 --> 01:06:34,780
You still want to generate the prompt, and that's definitely good.

937
01:06:35,820 --> 01:06:41,940
But what you can do before this is we can go to Google Gemini.

938
01:06:42,660 --> 01:06:48,140
First, what we want to do is we can screen record and talk through a process or something like that.

939
01:06:48,520 --> 01:06:53,540
So talk through a process, walk through some steps, use a screen recording to be able to do that.

940
01:06:54,260 --> 01:06:59,660
Use Google to record your screen that you're going to show whatever your flow,

941
01:06:59,840 --> 01:07:01,780
whatever your process is and things like that.

942
01:07:02,040 --> 01:07:06,960
Then what you do is you post that into Google Gemini,

943
01:07:07,040 --> 01:07:14,200
you go to cloud and cloud and you take the cloud generator prompt based off

944
01:07:14,200 --> 01:07:18,020
of kind of what you what your task is, what you wanted to do afterwards.

945
01:07:18,800 --> 01:07:23,620
So now Google Gemini has access to the video because it's the only platform

946
01:07:23,620 --> 01:07:30,240
so far out there as is recording this at August 22nd.

947
01:07:30,240 --> 01:07:35,620
Let's let's make sure that's clear as of august there's the only platform that takes video,

948
01:07:36,260 --> 01:07:39,040
so you could take that and then

949
01:07:39,040 --> 01:07:42,820
take this prompt put it into google and

950
01:07:42,820 --> 01:07:46,200
it can generate and create things for you based

951
01:07:46,200 --> 01:07:49,160
off the information that you have now provided it because they

952
01:07:49,160 --> 01:07:53,040
can now see understand what you're trying to say sure and

953
01:07:53,040 --> 01:07:55,940
then it can take this prompt that that was generated from clot

954
01:07:55,940 --> 01:07:58,840
to produce an output and or produce

955
01:07:58,840 --> 01:08:02,380
some dialogue based off of what you're asking so it's

956
01:08:02,380 --> 01:08:06,960
a nice little cool hack of taking three different or two different well three

957
01:08:06,960 --> 01:08:12,000
different things which is screen record but using two different llms in a way

958
01:08:12,000 --> 01:08:18,440
to generate what you wanted to do for your result for for your output so um

959
01:08:18,440 --> 01:08:22,180
it's a really cool cool thing that you can be able to do from that.

960
01:08:22,680 --> 01:08:27,200
And then whatever you do with the output, you can even chain it together to

961
01:08:27,200 --> 01:08:30,200
go to another LLM to do some other task for you or something like that.

962
01:08:31,437 --> 01:08:37,017
Yeah, that's super cool. That is, I have not played with that video feature on Gemini yet. No?

963
01:08:37,537 --> 01:08:40,857
No, not yet. I've done a lot of voice to text, a lot of talking.

964
01:08:41,037 --> 01:08:45,977
I've done transcriptions. I use another program as part of my editing occasionally

965
01:08:45,977 --> 01:08:49,037
to generate really rapid transcripts.

966
01:08:49,637 --> 01:08:54,417
So then I'll take that to one of these LLMs and say, here's a transcript of

967
01:08:54,417 --> 01:08:59,397
a show. And I've actually found that in the early days, it couldn't handle like

968
01:08:59,397 --> 01:09:03,637
a GPT-3-5 couldn't handle more than an hour of a show.

969
01:09:03,817 --> 01:09:08,737
And now when they've did these enhancements with GPT-4, it can handle like,

970
01:09:08,737 --> 01:09:12,517
I think like half an encyclopedia now of stuff.

971
01:09:12,697 --> 01:09:15,097
And some people don't know you want encyclopedias. It's like,

972
01:09:15,097 --> 01:09:17,937
just think of many Wikipedia pages strung together.

973
01:09:18,697 --> 01:09:24,317
Yeah. Yeah. Yeah. I think one of the things that I've, I'm going to stop sharing my screen now.

974
01:09:24,697 --> 01:09:28,437
But one of the things that I found very interesting about putting a lot of this

975
01:09:28,437 --> 01:09:37,277
stuff together and going through this exploratory process is that AI has shown

976
01:09:37,277 --> 01:09:43,817
me that there's almost a limited different ways to kind of go about this,

977
01:09:43,997 --> 01:09:46,977
to be able to kind of solve a problem or use a task.

978
01:09:46,977 --> 01:09:53,437
And you don't even need to be a programmer to be able to hack together a solution for yourself.

979
01:09:54,297 --> 01:09:57,177
You know, there's two things I didn't even go over.

980
01:09:57,297 --> 01:10:02,297
If somebody wants to research them and look at them later, please go ahead and do so.

981
01:10:02,617 --> 01:10:07,617
One is Claude's artifacts, which kind of gives you a UI that you can kind of

982
01:10:07,617 --> 01:10:09,017
create and mess with yourself.

983
01:10:09,237 --> 01:10:13,117
And then the other one is called Facebook Sam, which is segmented anything,

984
01:10:13,637 --> 01:10:18,837
which I think is going to be very useful for those who are in construction and

985
01:10:18,837 --> 01:10:21,237
they're probably going to be one of the more useful tools that are out there.

986
01:10:21,857 --> 01:10:28,777
Basically, what it allows you to do is be able to click anything in an image or in a video and.

987
01:10:30,519 --> 01:10:34,619
Cuts out whatever that image is wow so

988
01:10:34,619 --> 01:10:38,139
some crazy use cases around that for

989
01:10:38,139 --> 01:10:43,399
as far as reading diagrams construction safety i

990
01:10:43,399 --> 01:10:47,559
was thinking another one which was another one's like workplace if you want

991
01:10:47,559 --> 01:10:52,819
to track flows within places and things like that i think those are going to

992
01:10:52,819 --> 01:10:57,979
be really huge but i think for now i think the the the low hanging fruit in

993
01:10:57,979 --> 01:11:03,359
my opinion And it's going to be the diagrams and then looking at it from a safety perspective,

994
01:11:03,659 --> 01:11:07,739
you know, be able to click and identify certain things and things like that,

995
01:11:07,859 --> 01:11:09,939
at least off the top of my head.

996
01:11:10,059 --> 01:11:13,819
And when I think about, like, being able to segment and point and click and

997
01:11:13,819 --> 01:11:16,959
I can cut out, you know, you know, for a podcaster standpoint,

998
01:11:17,219 --> 01:11:20,299
it's like, oh, I want to cut myself out and put myself on a background.

999
01:11:20,399 --> 01:11:24,759
This is the super low hanging fruit for podcasting.

1000
01:11:26,279 --> 01:11:29,899
That's true. That's good. I've talked to some architects,

1001
01:11:30,139 --> 01:11:35,659
Marcus, too, that have been using AI to critique their fire life safety plans

1002
01:11:35,659 --> 01:11:37,779
to be code compliant so that they

1003
01:11:37,779 --> 01:11:41,619
can pass reviews with authorities having jurisdiction in the first round.

1004
01:11:41,979 --> 01:11:45,919
This is something that some architects have struggled with. Some of it's just

1005
01:11:45,919 --> 01:11:47,479
don't have the right experience.

1006
01:11:47,859 --> 01:11:51,599
And then building codes, they don't change every year.

1007
01:11:51,719 --> 01:11:56,099
But depending on what city you're in and what country you're in,

1008
01:11:56,199 --> 01:11:59,979
they do change. change occasionally and people got to learn.

1009
01:12:00,119 --> 01:12:03,499
So you can feed, you can feed an instance of it, the information,

1010
01:12:03,679 --> 01:12:08,659
and then you can have it review the drawing for compliance and give you recommendations

1011
01:12:08,659 --> 01:12:11,739
and pointers of things to look at that you might've missed.

1012
01:12:12,039 --> 01:12:15,619
So you can have it act as the authority having jurisdiction,

1013
01:12:15,899 --> 01:12:20,679
uh, grading your, your plan to the code, and then it can make suggestions to

1014
01:12:20,679 --> 01:12:22,199
you. And then of course you're the human in the loop.

1015
01:12:22,259 --> 01:12:25,519
You got to incorporate the change or ignore the change and they

1016
01:12:25,519 --> 01:12:28,959
found and this one individual saying like it's

1017
01:12:28,959 --> 01:12:32,139
made their first pass on review go up

1018
01:12:32,139 --> 01:12:34,979
to 100 so now they always pass first pass so like

1019
01:12:34,979 --> 01:12:38,619
if you think about that from a an architect all the architects listening like

1020
01:12:38,619 --> 01:12:41,719
depending on how big your job is like you know most of the projects i'm involved

1021
01:12:41,719 --> 01:12:46,579
in are they're north of 100 million dollars so for design of that i mean you

1022
01:12:46,579 --> 01:12:50,359
can just use the percentages that you know the effort to respond on to comments

1023
01:12:50,359 --> 01:12:54,779
on a project that large can be more than two weeks.

1024
01:12:54,839 --> 01:12:58,899
So if you think about a design team with their disciplines, I mean,

1025
01:12:58,919 --> 01:13:02,819
you easily can be on a project of that size, more than 20 people for a month.

1026
01:13:03,747 --> 01:13:10,247
So you can take 20 designers for a month and just save yourself by using AI,

1027
01:13:10,467 --> 01:13:13,107
just looking at, I know we're just picking fire, life, safety.

1028
01:13:13,267 --> 01:13:15,627
I mean, you could go, there are more use cases beyond what I'm saying,

1029
01:13:15,767 --> 01:13:20,067
but these are things that can shut you down, especially if you're doing like

1030
01:13:20,067 --> 01:13:20,807
healthcare construction.

1031
01:13:21,347 --> 01:13:26,927
So you can save. So like, think about this designers, you can save a month of

1032
01:13:26,927 --> 01:13:32,207
cost that you're just eating your profit because the what the owner contracted

1033
01:13:32,207 --> 01:13:34,107
you to do is to create a set of documents.

1034
01:13:34,887 --> 01:13:38,107
That makes it through permitting they don't

1035
01:13:38,107 --> 01:13:42,547
pay for multiple rounds of permitting they don't pay for multiple rounds of

1036
01:13:42,547 --> 01:13:47,227
comments because if you get comments and you don't pass you know by all definitions

1037
01:13:47,227 --> 01:13:51,987
and conventions like you generate your work product is subpar and it failed

1038
01:13:51,987 --> 01:13:55,347
to pass and And now all that extra work to get it on round two.

1039
01:13:55,427 --> 01:14:02,187
And Marcus, some things have failed four times and have gone back into permitting four times.

1040
01:14:02,947 --> 01:14:07,427
Projects out there in the big wide world. So that's all money lost that designers

1041
01:14:07,427 --> 01:14:11,627
cannot recoup from the client because the client bought a bill of goods,

1042
01:14:11,787 --> 01:14:13,347
which is a permittable set.

1043
01:14:13,927 --> 01:14:17,907
So how they get there after that first round is all eating their profit.

1044
01:14:18,307 --> 01:14:21,787
So this is one way. and i'm sorry

1045
01:14:21,787 --> 01:14:24,507
to the experienced designers that are that

1046
01:14:24,507 --> 01:14:27,607
qc person that checks this like this is

1047
01:14:27,607 --> 01:14:30,667
this is an example of how you better start scaling up

1048
01:14:30,667 --> 01:14:33,527
with ai because it can it

1049
01:14:33,527 --> 01:14:38,107
can and does eliminate jobs by making things faster like this is one of the

1050
01:14:38,107 --> 01:14:41,567
darker sides of you know some of these use cases like the things marcus and

1051
01:14:41,567 --> 01:14:45,347
i were showing today these are conversations in the past you'd have with other

1052
01:14:45,347 --> 01:14:49,607
people and now we're just talking to a prompt on our computer And we're getting

1053
01:14:49,607 --> 01:14:52,447
the capabilities of thousands and thousands,

1054
01:14:52,527 --> 01:14:55,387
if not millions of people. And we're getting it in seconds.

1055
01:14:56,027 --> 01:14:59,247
Yeah. No friction. Yeah. You know, one of the best use cases,

1056
01:14:59,347 --> 01:15:07,587
I think, of AI is or pro use cases is, is that you don't start off with a blank sheet of paper anymore.

1057
01:15:08,107 --> 01:15:11,867
Right. You're starting with something. And now you can create and you can iterate

1058
01:15:11,867 --> 01:15:13,147
off of those type of things.

1059
01:15:13,607 --> 01:15:17,447
But at the same time, there's a cost. There's a human capital cost of being

1060
01:15:17,447 --> 01:15:18,427
able to create those things.

1061
01:15:19,027 --> 01:15:23,707
Yes, it's paying instruction and we don't like doing those things and all that kind of stuff.

1062
01:15:24,007 --> 01:15:28,467
But I think, you know, that it will, you know, as, you know,

1063
01:15:28,467 --> 01:15:30,787
the war for talent and qualified.

1064
01:15:32,464 --> 01:15:36,404
Within the construction space, within a lot of other different industries and things like that.

1065
01:15:36,524 --> 01:15:42,444
But more in construction is that, you know, we're struggling to find qualified

1066
01:15:42,444 --> 01:15:46,224
talent and we're struggling to find people that we wanna train.

1067
01:15:46,444 --> 01:15:48,984
And sometimes we don't have the time to train those people.

1068
01:15:49,284 --> 01:15:53,744
So being able to offset those costs or being able to be able to pursue more

1069
01:15:53,744 --> 01:15:58,124
projects, sometimes that's the bottleneck is that we don't have enough people to be able to do that.

1070
01:15:58,264 --> 01:16:03,644
And we only have one person that's doing the, you know, the compliance work

1071
01:16:03,644 --> 01:16:06,044
and making sure that the quality is supposed to be where it is.

1072
01:16:06,304 --> 01:16:10,404
Well, now you have a chance to implement and put, you know, now they're able

1073
01:16:10,404 --> 01:16:14,144
to do three or four, then they don't have to, they don't have to do it as much.

1074
01:16:14,344 --> 01:16:16,744
So it gives you some opportunities to be able to get some more work.

1075
01:16:16,824 --> 01:16:20,664
And I think it gives them a potential for smaller firms to be able to compete.

1076
01:16:20,824 --> 01:16:23,724
But at the same time for those larger companies, those larger firms,

1077
01:16:23,804 --> 01:16:26,704
do we need as much staff as we need to have?

1078
01:16:27,064 --> 01:16:31,104
Right. And I think that's where the displacement is going to start happening

1079
01:16:31,104 --> 01:16:34,144
and you start questioning those type of things.

1080
01:16:34,604 --> 01:16:38,284
But for the smaller firms and things like that, yes, this is going to be amazing for you.

1081
01:16:39,184 --> 01:16:41,764
You could be able to get more work, you'd be able to be able to do more efficiently.

1082
01:16:43,044 --> 01:16:47,084
But for those larger firms and things like that, it's going to be difficult.

1083
01:16:47,524 --> 01:16:50,844
I think there's going to be some departments that were as large as they used

1084
01:16:50,844 --> 01:16:52,624
to be, won't be there anymore.

1085
01:16:54,466 --> 01:16:58,666
You know, before we used to have departments about being able to do a lot of copy paper.

1086
01:16:58,986 --> 01:17:04,546
You know, now everything's on iPads and screens, and you just pass it off on a tablet.

1087
01:17:04,846 --> 01:17:07,826
Here's the document that is sold.

1088
01:17:09,566 --> 01:17:12,646
Yeah, Marcus, some people are too young. They don't even remember that there

1089
01:17:12,646 --> 01:17:14,906
were RepoGraphics offices.

1090
01:17:15,426 --> 01:17:21,026
I remember in my high school, there was an office. It was like stenography.

1091
01:17:21,246 --> 01:17:23,906
I think it had a name. The Steno Pool. i think they

1092
01:17:23,906 --> 01:17:27,026
call it i remember walking past the door and i was like what's this

1093
01:17:27,026 --> 01:17:30,126
and you open up the door it's like these machines that

1094
01:17:30,126 --> 01:17:33,206
are like drums where they just roll paper and copy

1095
01:17:33,206 --> 01:17:38,926
with like mechanical pressure it has they used to make photocopies before photocopy

1096
01:17:38,926 --> 01:17:42,766
machines came out and it used to be like you know it could be teams of like

1097
01:17:42,766 --> 01:17:48,926
five ten people making copies like using this very manual process and now with

1098
01:17:48,926 --> 01:17:52,866
the printers and laser printers like those departments are gone gone.

1099
01:17:52,926 --> 01:17:56,006
Those jobs are gone. Nobody talks about the poor Stenal pool.

1100
01:17:56,366 --> 01:17:58,846
And what happened to all those people? Gone.

1101
01:17:59,326 --> 01:18:04,066
Yeah. I mean, you can even talk about like my wife, one of her uncles worked

1102
01:18:04,066 --> 01:18:08,386
for the UN and that was the type of department that he worked in.

1103
01:18:08,666 --> 01:18:13,586
So imagine doing a UN call and you have to translate this papers and all these

1104
01:18:13,586 --> 01:18:15,626
different papers and all these different languages, right?

1105
01:18:15,786 --> 01:18:19,606
That's a lot of paper that you're going through to, to, for, for a piece of document.

1106
01:18:19,906 --> 01:18:22,946
And so you're making sure that it's in everybody's native language.

1107
01:18:23,246 --> 01:18:27,086
Now, he said, you know, there were whole floors that are dedicated.

1108
01:18:27,206 --> 01:18:33,586
I'm not talking small, small, small places, you know, you know,

1109
01:18:33,586 --> 01:18:37,826
straight on printing stands of being able to be able to generate all that stuff.

1110
01:18:37,966 --> 01:18:44,246
But now, you know, those departments are no longer needed or a lot smaller than what they used to be.

1111
01:18:44,346 --> 01:18:46,666
If there's still some paper copies out there, right. Right.

1112
01:18:46,846 --> 01:18:51,466
But, you know, and that's translating into multiple hundreds,

1113
01:18:51,606 --> 01:18:55,986
different languages over the same document, just so people can understand what

1114
01:18:55,986 --> 01:18:57,866
it what that document meant. Right.

1115
01:18:58,246 --> 01:19:05,546
So, you know, it's, you know, it's technology is always create some type of displacement.

1116
01:19:06,206 --> 01:19:10,906
But at the same time, you know, I think as society, it's human humans.

1117
01:19:11,166 --> 01:19:14,826
We've always found different things to do. We've always found different ways to complain.

1118
01:19:14,826 --> 01:19:18,086
And that's just

1119
01:19:18,086 --> 01:19:22,746
true he's not lying people like in the ancient sumerian tablets they found when

1120
01:19:22,746 --> 01:19:26,626
they found these things like people back in the day complaining about stuff

1121
01:19:26,626 --> 01:19:30,366
that they bought in the market you know that they hated it so much they took

1122
01:19:30,366 --> 01:19:35,106
the time to etch it into clay fire the tablet and then share it with people

1123
01:19:35,106 --> 01:19:37,946
so like we'd love to complain.

1124
01:19:39,208 --> 01:19:43,268
And I think the, you know, the use cases that we come up with is that they also

1125
01:19:43,268 --> 01:19:44,428
give birth to new industries.

1126
01:19:44,608 --> 01:19:47,568
You know, somebody has to maintain that, things like that.

1127
01:19:47,708 --> 01:19:53,048
But I think the sometimes the biggest pain for some people is the reskilling.

1128
01:19:53,668 --> 01:19:58,968
But I do think with AI, a lot of that also can be alleviated with being able

1129
01:19:58,968 --> 01:20:04,788
to reskill and use an AI as a tutor or coach to be able to learn how to be able to do stuff.

1130
01:20:04,908 --> 01:20:10,028
You know, tell me like I'm five, right? Right. That's a that's a great prompt for people.

1131
01:20:10,608 --> 01:20:14,168
And then after you tell me like I'm five, OK, now I'm a novice.

1132
01:20:14,588 --> 01:20:19,988
You know, you saw the prompt that was created from Cloud Generate and that was

1133
01:20:19,988 --> 01:20:20,948
from a novice perspective.

1134
01:20:21,088 --> 01:20:24,828
But you can go another level and say, tell me like I'm five. I don't understand.

1135
01:20:25,148 --> 01:20:27,888
And then they explain it from there and then you can build up from there.

1136
01:20:28,028 --> 01:20:32,528
And I think that's one of the cool things about AI is that if you need to be

1137
01:20:32,528 --> 01:20:36,668
explained a certain way, It can take a document or take some ideas and break

1138
01:20:36,668 --> 01:20:41,548
it down and explain it to you in many different ways that you could ever once or imagine.

1139
01:20:42,088 --> 01:20:44,628
You can even have it talk like a pirate if you want it to.

1140
01:20:48,148 --> 01:20:53,388
Yeah. With the new GPTs, they've even had it. It can sing now. Yeah.

1141
01:20:54,748 --> 01:20:58,248
You're going to have it sing all the answers to you. So, Marcus,

1142
01:20:58,728 --> 01:21:02,728
thank you so much for coming on the show and sharing all your knowledge and wisdom.

1143
01:21:02,988 --> 01:21:07,548
And we're going to keep talking for sure, as we always do, and pushing the envelope.

1144
01:21:07,848 --> 01:21:11,948
Is there anything you want to leave the audience as they go off into their day?

1145
01:21:12,728 --> 01:21:15,908
Just if you have any questions or anything like that, feel free to hit me up

1146
01:21:15,908 --> 01:21:18,788
on LinkedIn or any social media. I am Marcus Turner.

1147
01:21:19,628 --> 01:21:22,428
I am Marcus Turner on everything that I could think of.

1148
01:21:22,428 --> 01:21:25,708
So you can use your preferred social media platform and

1149
01:21:25,708 --> 01:21:30,608
please feel free to message me or ask me a question please let me know that

1150
01:21:30,608 --> 01:21:34,528
you're listening to flip based podcasts so i can at least have some context

1151
01:21:34,528 --> 01:21:39,908
around it but i am marcus turner everywhere and i think that would be the best

1152
01:21:39,908 --> 01:21:44,428
way to be able to get in touch with me awesome thank you so much marcus all right,

1153
01:21:46,888 --> 01:21:52,308
very special thanks to my guest i'm felipe engineer manriquez.

1154
01:21:47,440 --> 01:21:53,680
Music.

1155
01:21:55,768 --> 01:22:02,248
The ebfc show is created by felipe and produced by a passion to build easier and better,

1156
01:22:04,848 --> 01:22:12,408
thanks for listening stay safe everybody let's go build,

1157
01:22:14,428 --> 01:22:14,728
Thank you.