1
00:00:00,000 --> 00:00:06,560
something is really bad. So I went and clicked to learn more. And yeah, the first point on invalid

2
00:00:06,560 --> 00:00:14,000
traffic is that it's artificial traffic boosting your own views. And it immediately clicked to me.

3
00:00:14,560 --> 00:00:21,360
It immediately clicked to me that, oh, there was of course just right now, artificial

4
00:00:21,360 --> 00:00:31,200
traffic, but watching and visiting my YouTube channel.

5
00:00:45,520 --> 00:00:50,720
Ladies and gentlemen, boys and girls, dogs, cats, robots, and everybody in between, especially you,

6
00:00:50,720 --> 00:00:58,400
Mr. Auto GPT agent, we know you're listening this week. Welcome to HTTTA. I am your host, West of

7
00:00:58,400 --> 00:01:04,960
SynthMind, SynthMind West. And I'm always joined by the galleon, the gleeful, the golden haired

8
00:01:05,520 --> 00:01:11,600
Grace herself glittering on the other side of this call, the galaxy of inspiration herself,

9
00:01:11,600 --> 00:01:18,160
Miss GoToGo. Gee, how are you this week? Usually after such introduction, I would say I'm

10
00:01:18,160 --> 00:01:25,520
fantastic. Good. Then message received. But this week, I'm not so well, to be honest.

11
00:01:25,520 --> 00:01:34,320
Wow. So what brought that about? Well, our beloved topic of auto GPT, I got my hands on it,

12
00:01:34,320 --> 00:01:42,000
and it was not as friendly as I thought it would be. So my experience, at least the first encounter

13
00:01:42,000 --> 00:01:49,280
was definitely a very traumatizing, I would say. Well, let's tell some of our listeners what auto

14
00:01:49,280 --> 00:01:55,440
GPT actually is. I'm sure people are familiar with chat GPT at this point, if you're listening to

15
00:01:55,440 --> 00:02:01,600
this podcast, but what auto GPT is, it's an initiative by some folks. You can get the

16
00:02:01,600 --> 00:02:07,440
repository on GitHub right now. It's a program where there's an interface via a kind of DOS command

17
00:02:07,440 --> 00:02:14,080
prompt window. It's the fastest growing repository in terms of like stars on GitHub ever.

18
00:02:14,720 --> 00:02:23,920
112,000 stars and forks. Oh, how many forks? 19,000.8.

19
00:02:23,920 --> 00:02:29,920
So what that means to folks that aren't, you know, coding on a regular basis or into that side of

20
00:02:29,920 --> 00:02:38,480
things, that means 119,000 people have downloaded this program, run the code to install it and

21
00:02:38,480 --> 00:02:44,480
execute it. And then another 19,000 have taken the base code for this and started to play with it in

22
00:02:44,480 --> 00:02:51,280
other ways. So the way it kind of works is you have on chat GPT, we know it's not hooked up to

23
00:02:51,280 --> 00:02:56,880
the internet. There's no internet access with chat GPT. It's trained on a dataset that only goes up

24
00:02:56,880 --> 00:03:04,400
through September, 2021. It can't necessarily execute a command outside of chat GPT and it relies

25
00:03:04,400 --> 00:03:09,280
on continuous prompting. You have to kind of keep prompting it, you know, to get an output.

26
00:03:09,280 --> 00:03:15,760
Auto GPT can access the internet and it works autonomously by having two separate kind of like

27
00:03:15,760 --> 00:03:22,240
GPT 3.5 or four agents talking to one another and collaborating. One's prompting the other,

28
00:03:22,240 --> 00:03:27,680
others doing the search, then they prompt back based on that results. And it has long-term memory.

29
00:03:27,680 --> 00:03:34,320
So it'll memorize what it's doing over the course of time. That's what it is. And people are calling

30
00:03:34,320 --> 00:03:40,160
it a Baby AGI, artificial general intelligence, because it can essentially kind of think for itself.

31
00:03:40,160 --> 00:03:46,800
But not to mix it with hugging face Baby AGI, spin off project, which is actually no code

32
00:03:46,800 --> 00:03:54,400
and that browser smaller version just to get the feel of it. In Baby AGI project, you input your

33
00:03:54,400 --> 00:04:03,360
open AI key, you give a goal and it starts executing what should be steps taken to reach that goal.

34
00:04:03,360 --> 00:04:09,360
Dave So I think the best way to illustrate what we mean by how to interface and use this goal,

35
00:04:09,360 --> 00:04:13,360
Kota, would you tell us your experience with Auto GPT this past week?

36
00:04:13,360 --> 00:04:19,920
Kota Yeah, sure. So just to provide you with a background, I'm not a coder, if you may.

37
00:04:19,920 --> 00:04:27,680
I'm familiar now with GitHub and I always wanted to learn Python. And I know HTML,

38
00:04:27,680 --> 00:04:35,040
done some web management stuff. So I'm kind of familiar with some coding stuff, but not to the

39
00:04:35,040 --> 00:04:40,640
extent that I can comfortably say that give me code and I will know what to do with that.

40
00:04:40,640 --> 00:04:45,040
Dave Sure. Kota So that being said, when I'm learning about something and this auto GPT has

41
00:04:45,040 --> 00:04:51,920
been coming into my emails, people were asking, I knew what it was. I was keeping an eye on that.

42
00:04:51,920 --> 00:04:57,280
But I decided to get kind of my hands dirty and actually try it.

43
00:04:57,280 --> 00:05:06,000
Dave So you went and found yourself a Python 101 book and figure it out how to clone a repository.

44
00:05:06,000 --> 00:05:12,800
Look at you. You're practically got a computer science degree then over the weekend, it seems like.

45
00:05:12,800 --> 00:05:18,720
Kota So this is the funny part. And in my video I made about this topic, I said that it's both

46
00:05:18,720 --> 00:05:26,160
fascinating, but both scary if you think about it. It took me half a day from getting GitHub

47
00:05:26,160 --> 00:05:34,560
to getting Python to copying repository. And the funny thing is, looking back now can reflect,

48
00:05:34,560 --> 00:05:41,840
I did not go on YouTube to watch tutorials. I used chat GPT to help me. For example,

49
00:05:41,840 --> 00:05:50,000
one of my hiccups was that Python was not running. And apparently I installed the latest, latest Python,

50
00:05:50,000 --> 00:05:57,280
like 11. How you write is like Python 3 instead of pip. And I was like, I don't understand. I'm

51
00:05:57,280 --> 00:06:03,760
doing everything right and it's not working. And then I copied the whole description how to install

52
00:06:03,760 --> 00:06:10,480
auto GPT, gave it to chat GPT that it knows what I'm talking about. It knows what I should be doing

53
00:06:10,480 --> 00:06:18,960
or what I should be using. It was going back and forth. And once I was like hit it run and it worked.

54
00:06:18,960 --> 00:06:25,520
And I was like, what? Dave You've just described the entire coding experience, not just for yourself,

55
00:06:25,520 --> 00:06:31,520
but for every developer out there, which is, hey, I think I got this working. This is good. This is

56
00:06:31,520 --> 00:06:38,960
exciting. Wait, why is it not working? Like what? Unexplicable feeling. Like why is it not working?

57
00:06:38,960 --> 00:06:44,480
Why is it not working? Like this should, I'm doing everything right. And it's a, you put a colon or a

58
00:06:44,480 --> 00:06:52,320
backslash in the wrong place on line 172. But then the feeling of just sheer bliss when it works.

59
00:06:52,320 --> 00:06:58,080
You are like, I am the smartest person alive. There is no other way to describe it.

60
00:06:58,080 --> 00:07:04,880
Yes. But this feeling was very short lived. I know that we are teasing, but okay, I'm getting to that.

61
00:07:04,880 --> 00:07:09,600
Yes. I got it running. And to be honest, I was kind of taking a step back. I was like,

62
00:07:09,600 --> 00:07:17,120
okay, it works. So I was so preoccupied making it work, but I did not really think what do I want

63
00:07:17,120 --> 00:07:23,840
to ask? Like how should I go about that? I posted on YouTube and on LinkedIn and also in our community

64
00:07:23,840 --> 00:07:31,680
and learn prompting just to get some ideas. And one kind of basic one is what consumes a lot of

65
00:07:31,680 --> 00:07:39,840
my time is researching ideas, but also content for making YouTube videos. I spent a lot of time

66
00:07:40,480 --> 00:07:47,840
diving into topics and I was like, okay, so if AutoGPT would actually gather some ideas maybe for

67
00:07:47,840 --> 00:07:53,600
me, but also scout the internet doing some research, maybe finding some articles, which I was not

68
00:07:53,600 --> 00:08:00,640
aware of. So the way it works is you give it a persona, role prompting, right? And then as many

69
00:08:00,640 --> 00:08:05,360
as five goals for it to try to accomplish. So what was your persona and then the goals that you gave

70
00:08:05,360 --> 00:08:11,920
it? The first time I went with the name, so I did Ben. The first test, I wanted to see how it

71
00:08:11,920 --> 00:08:20,160
understands what does it mean viral video and what it would do. So I asked about viral video ideas,

72
00:08:20,160 --> 00:08:27,360
which would get a lot of views and also engage people. For the goals, it was understand what is

73
00:08:27,360 --> 00:08:33,360
viral, research topics, which are relevant to my YouTube channel. And this is where mistakes

74
00:08:33,360 --> 00:08:39,520
started happening. And then yeah, write a script, YouTube titles and all that. First, it started

75
00:08:39,520 --> 00:08:46,240
researching what is viral video and what you should do. And that was very disappointing, very basic.

76
00:08:46,240 --> 00:08:53,760
It literally went to the top Google results. But then I pushed on the fact that it needs to be

77
00:08:53,760 --> 00:09:00,880
contextual. It has to be for my YouTube channel. So on the first try it seemed, and I was checking

78
00:09:00,880 --> 00:09:06,240
what ideas it gathered. On the first try, it started pushing that, okay, talk about technology

79
00:09:06,240 --> 00:09:12,880
so that I should do gaming. Because that gets viral. Yeah, sure. Go to go the Twitch screamer.

80
00:09:12,880 --> 00:09:19,200
Maybe I should try. Maybe this actually was not a bad advice. Maybe you know something you don't.

81
00:09:19,200 --> 00:09:25,040
Exactly. Actually, we just got Nintendo Switch. Anyway, beside the point, I gave feedback. So

82
00:09:25,040 --> 00:09:31,600
this is what you do. You can continue like each command. You can say yes, continues. You can also

83
00:09:31,600 --> 00:09:38,240
give it number of commands. So yes, 10 would mean that it will run 10 commands automatically,

84
00:09:38,240 --> 00:09:46,560
with no intervention from you. N is that you shut it down. It exits. And there is this whole other

85
00:09:46,560 --> 00:09:52,400
thing with just continuous mode, which is absolutely not recommended. I have not tested it yet.

86
00:09:52,400 --> 00:09:58,960
It's expensive. Exactly. And I want to talk later about this, the whole cost side, because I see a

87
00:09:58,960 --> 00:10:07,520
lot of misinformation that auto GPT is free. Going back to that, you can also give feedback. So I was

88
00:10:07,520 --> 00:10:15,760
trying to give that, hey, I am got to go. And it needs to be contextual. Okay. So then you can see

89
00:10:15,760 --> 00:10:22,240
how it talks to itself, right? But also talks to the agents it creates. So it created agent, which

90
00:10:22,240 --> 00:10:30,320
needs to research viral videos. It created an agent, which needs to eventually understand who I am.

91
00:10:30,320 --> 00:10:36,160
And then another agent was writing scripts. And those consult with each other. So it would say,

92
00:10:36,160 --> 00:10:42,480
hey, the agent who writes scripts need to go and consult with the agent who knows viral video ideas.

93
00:10:42,480 --> 00:10:48,000
So it just made a whole production department for you basically. Right. And then I saw that, yeah,

94
00:10:48,000 --> 00:10:55,200
it says that, okay, I need to watch got to go videos to understand context and identify her best

95
00:10:55,200 --> 00:11:01,280
performing videos and understand her audience. And whether it's all great. And it did. It found my

96
00:11:01,280 --> 00:11:07,440
videos. It wrote who I am, took the description from YouTube. And the thing is I always have my

97
00:11:07,440 --> 00:11:15,200
YouTube analytics on just to see what's going on. And I saw this pop up almost instantly. And I'm

98
00:11:15,200 --> 00:11:23,200
very aware that this two things could be completely unrelated events, just like statistically, it can

99
00:11:23,200 --> 00:11:31,520
be right. There is also a chance. However, this pop up was saying that ad revenue will be affected

100
00:11:31,520 --> 00:11:39,440
on some of your videos because YouTube detected invalid traffic. And I was like, what? Immediately

101
00:11:39,440 --> 00:11:45,440
I saw ad revenue affected. I'm like, no, no, no, no, no, something is bad. Because if YouTube does

102
00:11:45,440 --> 00:11:51,200
not want advertisement be involved with you, something is really bad. So I went and clicked

103
00:11:51,200 --> 00:11:58,640
to learn more. And yeah, the first point on invalid traffic is that it's artificial traffic

104
00:11:58,640 --> 00:12:05,600
boosting your own views. And it immediately clicked to me. It immediately clicked to me that,

105
00:12:05,600 --> 00:12:14,720
oh, there was of course just right now artificial bots watching and visiting my YouTube channel

106
00:12:14,720 --> 00:12:21,280
watching. And of course, if it's me or if it's auto GPT, they need to sit through the ad. That

107
00:12:21,280 --> 00:12:26,160
agent probably didn't like to sit through the ads. So now I'm talking about this with a smile,

108
00:12:26,160 --> 00:12:32,880
but I can tell you that I definitely panicked because it was a lot of work to grow this YouTube

109
00:12:32,880 --> 00:12:40,160
channel. And as a small YouTuber still, even though it's getting to 30,000 soon, but still,

110
00:12:40,160 --> 00:12:46,240
it's small YouTube channel and just seeing that all the effort you put and all the work could be

111
00:12:46,240 --> 00:12:53,040
affected by simply like, it's just ironical. I wanted to make better ideas and better videos

112
00:12:53,040 --> 00:13:01,120
and probably damage my channel. So it was really sad. I went and contacted YouTube support. And as

113
00:13:01,120 --> 00:13:07,920
a content creator, you have this chat for content creators. And I started talking with them and told

114
00:13:07,920 --> 00:13:15,600
that I'm seeing this pop up. And I said that I was playing with auto GPT and I used Google API

115
00:13:16,560 --> 00:13:20,960
and to allow it to browse. And then basically this happened. And I was like,

116
00:13:22,560 --> 00:13:27,760
what is the consequences? I would just like kind of was like panicking kid. And we were like,

117
00:13:27,760 --> 00:13:34,720
yeah, that's probably it. We can't do anything about it now. Just don't do this again. And you

118
00:13:34,720 --> 00:13:41,520
should be fine and takes time and to show which videos will be affected. I will see kind of later

119
00:13:41,520 --> 00:13:48,880
revenue. However, now I posted on YouTube this video about auto GPT telling the story. And I

120
00:13:48,880 --> 00:13:55,600
cover multiple aspects of that, but what I learned and was kind of like pondering about what happened

121
00:13:55,600 --> 00:14:03,440
because of course we have to talk about this, that we don't know how auto GPT or multiple agents

122
00:14:03,440 --> 00:14:10,160
browsing internet, bowing on websites, getting to your Gmail, using all the plugins which are

123
00:14:10,160 --> 00:14:18,000
coming out, how it all affects. And we've been fighting bots, Twitter bots, YouTube bots,

124
00:14:18,000 --> 00:14:23,440
you know this on YouTube, WhatsApp, join got to go on was, oh my God, it takes me so much time

125
00:14:23,440 --> 00:14:30,960
to just delete those. So we've been fighting those. And now we have tools where any new like myself

126
00:14:30,960 --> 00:14:37,360
can set up in half a day and deploy. So I just want to say one thing that of course, what this

127
00:14:37,360 --> 00:14:44,960
exposed that I didn't have any bad intentions, but anybody with bad intentions can do these things.

128
00:14:45,680 --> 00:14:52,160
But my theory is that this happened to me specifically because I use the same email

129
00:14:52,160 --> 00:14:59,600
to create Google API as my YouTube channel. So that's one aspect and YouTube is owned by Google.

130
00:14:59,600 --> 00:15:06,640
Another aspect is of course, my API didn't use any VPNs. So to detect that it is me, myself

131
00:15:06,640 --> 00:15:14,000
driving traffic and therefore I should be penalized or I should stop because as a creator

132
00:15:14,000 --> 00:15:20,240
doing it myself, I think it would be different or at least I hope that it's not the same case

133
00:15:20,240 --> 00:15:29,840
if I do and say, hey, research this YouTuber and auto GPT goes on that YouTuber's channel and watches

134
00:15:29,840 --> 00:15:34,880
and boom, suddenly that YouTuber is affected. I saw a lot of comments in my YouTube channel that,

135
00:15:34,880 --> 00:15:40,560
oh, so this is what can happen. And we don't know that that's a thing. We don't know what's the

136
00:15:40,560 --> 00:15:47,520
answer, right? Because another aspect is you yourself now can be like, hey, a friend in,

137
00:15:47,520 --> 00:15:54,000
I don't know, whatever country set up an account, auto GPT and asked to drive 100,000 views, watch

138
00:15:54,000 --> 00:16:00,960
my videos 100,000 times. Right? So all these kinds of things, of course, I thought about it a lot and

139
00:16:00,960 --> 00:16:07,040
I don't want to put all the ideas out in the wild, but that's definitely concerning how Google's

140
00:16:07,040 --> 00:16:13,120
going to react. Will we have to just simply move to Bing? Because Bing is like, yay, AI forever.

141
00:16:13,120 --> 00:16:19,920
Yeah. They're linked in with open AI. So yeah, that's it's the AI wars have begun. That's for

142
00:16:19,920 --> 00:16:28,320
sure. Yeah. So overall as a conclusion, auto GPT in general as a results for me, it was looping,

143
00:16:28,320 --> 00:16:35,920
running and errors. Yeah. It was a little bit disappointing besides what happened. Now it's

144
00:16:35,920 --> 00:16:41,360
kind of, I just hope that with YouTube, we are friends again, content creators are not going to

145
00:16:41,360 --> 00:16:46,640
be friends again. Content creation with auto GPT for my specific channel is off the table.

146
00:16:46,640 --> 00:16:54,640
Off the table. Yeah. And I tested the browser integrations like God mode, agent, GPT, baby,

147
00:16:55,120 --> 00:17:03,600
I G I. And of those, I can say that if listeners actually want to give it a try for me, God mode

148
00:17:03,600 --> 00:17:09,760
was the closest, the cleanest experience because you still see what it's doing, has this files,

149
00:17:09,760 --> 00:17:16,080
which you can at any moment open and read. It was getting me a lot of errors throughout the process,

150
00:17:16,080 --> 00:17:23,120
but you just need to refresh. But yeah, Wes, I need to ask you, did you try it and any stories

151
00:17:23,120 --> 00:17:30,480
from your side, you need to balance it out. So thankfully I can retort this week with probably

152
00:17:30,480 --> 00:17:38,000
more positive example of how auto GPT, specifically the God mode website integration version of it

153
00:17:38,000 --> 00:17:44,480
helped me out, but I would love to share the first time that I used it. So I definitely have been

154
00:17:44,480 --> 00:17:50,080
trying out some of these auto GPT and different types of derivations of it that have come up

155
00:17:50,080 --> 00:17:54,720
since its release probably about two, three weeks ago now. So the first time I tried it,

156
00:17:54,720 --> 00:17:59,120
I think this was the thing that kind of was like, Oh, this is kind of different. You know, we take

157
00:17:59,120 --> 00:18:04,560
the transcripts from these podcasts and I'll feed them into the GBT algorithms and just have it

158
00:18:04,560 --> 00:18:11,280
parsed through that for the key topics, build some outlines and use that to help write our articles

159
00:18:11,280 --> 00:18:17,600
for our newsletter each week. So I had the podcast as a text file, the transcript, put it in a

160
00:18:17,600 --> 00:18:23,040
location and then said to do just that. You're a tech blogger, tech writer, you know, parse through

161
00:18:23,040 --> 00:18:30,320
this, make the outline like this, put it in a text file. So what I didn't realize is I put the file

162
00:18:30,320 --> 00:18:35,200
I wanted it to sample from our transcript in the wrong directory. All right. You have to put it in

163
00:18:35,200 --> 00:18:41,920
a specific directory within the auto GPT folder. So what auto GPT did is it went, Hey, I don't see

164
00:18:41,920 --> 00:18:48,800
this file. Let me search the internet for how to do a window search. Then, okay, I found how to index

165
00:18:48,800 --> 00:18:55,040
and find the file. Let me research on Python, how to move it to the place I need to. So then it moves

166
00:18:55,040 --> 00:18:59,440
this little text file just into one more sub directory that I had it pointed at when I could

167
00:18:59,440 --> 00:19:03,600
have easily just redone it and refixed it. And I was just like, I'm looking back through the logs

168
00:19:03,600 --> 00:19:10,640
and like, whoa, like before it even started doing what I was asking, it corrected my mistake for me

169
00:19:11,360 --> 00:19:17,600
by going through that pathway on the internet, which was just like, yeah. So that was the first

170
00:19:17,600 --> 00:19:23,440
example. I was like, whoa, this is a little different. Where I find it does very well is with

171
00:19:23,440 --> 00:19:29,840
kind of like multi-layered advanced searches of information. Let me give you an example. I have

172
00:19:29,840 --> 00:19:36,480
a family friend whose daughter is applying for colleges, wants to be a veterinarian and go to

173
00:19:36,480 --> 00:19:41,840
college with a veterinarian program, that's, that's cool. They live in the Southeast US. So we were

174
00:19:41,840 --> 00:19:47,680
talking and I was naturally talking about AI and stuff, cause that's all I talk about. So I bring

175
00:19:47,680 --> 00:19:51,920
this up and say, well, Hey, where is she searching? Ah, she's looking through these programs, but you

176
00:19:51,920 --> 00:19:57,600
know, trying to, it's just, you know, it's not as clear. Obviously it's a challenge. So I'm like,

177
00:19:57,600 --> 00:20:01,280
well, what are some of your specific things you're looking for? So what was it being the

178
00:20:01,280 --> 00:20:08,400
Southeastern United States would like it to be less than 25 K a year, or have some sort of in-state,

179
00:20:08,400 --> 00:20:13,360
you know, discount that we can take advantage of. We'd like it to be a pretty top tier program.

180
00:20:13,360 --> 00:20:18,000
And then it'd be really helpful if we had information about what grants our daughter

181
00:20:18,000 --> 00:20:24,720
would be eligible for. So we can also start working that. So I said, I got a solution for you. Let's

182
00:20:24,720 --> 00:20:31,040
try to put it in auto GPT. So in and of itself, that's something that would take multiple

183
00:20:31,600 --> 00:20:35,840
individual searches. Hey, what's the top 10 colleges? What's the top 10 colleges in the

184
00:20:35,840 --> 00:20:40,160
Southeast that have a veterinarian program? How much do each one of those individual colleges cost?

185
00:20:40,160 --> 00:20:45,520
There's not a lot of aggregators of all of that kind of information in a lot of places, you know,

186
00:20:45,520 --> 00:20:50,320
it exists separately from one another. But to even pull that all together in the past would always

187
00:20:50,320 --> 00:20:56,560
be something that we would have to do in a spreadsheet or on some notes. However, so in

188
00:20:56,560 --> 00:21:02,480
about five minutes, it came back with a spreadsheet of the top 10 programs in the Southeast colleges

189
00:21:02,480 --> 00:21:09,360
that had veterinary science school, the annual tuition costs, even found the application deadlines

190
00:21:09,360 --> 00:21:14,560
for each one of those schools for this calendar year, and then the link to the website. And then

191
00:21:14,560 --> 00:21:20,160
it had a list of like three or four different grants or scholarships that they do. And they're

192
00:21:20,160 --> 00:21:24,400
like, this is, we would have done hundreds of searches, this would have been months of time.

193
00:21:24,400 --> 00:21:29,360
This is like, even if it's, she doesn't even go to one of those, it's a great starting point

194
00:21:29,360 --> 00:21:33,920
to at least go, hey, let's just start here and then figure out, not have to spend all this time

195
00:21:33,920 --> 00:21:40,160
just building that list and then have that, did I miss something? Feeling. It's just to your point,

196
00:21:40,160 --> 00:21:46,800
this is kind of like our thinking, how we search our research is completely transforming. And to

197
00:21:46,800 --> 00:21:55,360
your point, I want to make a video, something in lines of, I tested thousand AI tools. Let's say

198
00:21:55,360 --> 00:22:00,720
that I'm not a hundred percent yet committed, maybe it will end up a hundred. I'm just calculating

199
00:22:00,720 --> 00:22:06,160
now how long it would take. And my thinking was like, okay, so to start that, I need to create

200
00:22:06,160 --> 00:22:13,040
Google Sheets, start mapping out companies and links. And then I caught myself as like, what?

201
00:22:13,600 --> 00:22:21,520
No, like, why would I do that? I can just go, okay, AutoGBT, this is a specific website,

202
00:22:21,520 --> 00:22:29,120
the library of latest AI tools, get them all into Google spreadsheet with these type of columns

203
00:22:29,120 --> 00:22:35,760
and that I can just go and be like, boom, click, open, test it. I want to see, do you have to pay

204
00:22:35,760 --> 00:22:41,120
for it? Is there reviews? Like all these things just in one place right now in front of me. So

205
00:22:41,120 --> 00:22:48,400
this is what's kind of, you catch yourself in the old habits, but actually rethinking how we work,

206
00:22:48,400 --> 00:22:54,240
that not leaning back to the old ways is going to be kind of a new challenge.

207
00:22:54,240 --> 00:22:59,200
Oh, I think what a lot of people are struggling with exactly what you're describing, you know,

208
00:22:59,200 --> 00:23:04,400
we're still, we're very conditioned this way to be creative and then still have to have enough in

209
00:23:04,400 --> 00:23:12,240
the tank to then go execute, right? These AI tools obviously make a lot of stuff very easy, but if you

210
00:23:12,240 --> 00:23:18,880
use them in a much more creative fashion, where you're descriptive, where you're giving examples,

211
00:23:18,880 --> 00:23:24,800
where you're doing some more advanced prompt engineering, that, you know, the creativity

212
00:23:24,800 --> 00:23:30,880
is still all there, then it's just the commodity of execution that has now changed. Because I don't

213
00:23:30,880 --> 00:23:37,040
have to do as much executing actual typing, but I'm still going to take it and then add half a dozen

214
00:23:37,040 --> 00:23:44,400
sentences to it, edit 10 other ones. And it's still a hundred percent me, you know, I just had some

215
00:23:44,400 --> 00:23:51,600
help getting to that 80% starting line. And then I'm probably able to even pour in more creativity

216
00:23:51,600 --> 00:23:57,200
and because there's ideas and maybe I didn't think of or sources that I wouldn't have used. So I think

217
00:23:57,200 --> 00:24:02,160
that you hear a lot of people saying it's going to disrupt all these industries. Yeah, it probably is,

218
00:24:02,160 --> 00:24:08,960
but how about the way that it can do it positively? I don't think any copywriter loves the act of

219
00:24:08,960 --> 00:24:14,400
copywriting, like typing it out itself, but probably enjoy the act of ideating and being

220
00:24:14,400 --> 00:24:19,280
creative and trying to figure out how to convey the message that they need to, you know, for their

221
00:24:19,280 --> 00:24:25,120
client. So what you just described there is a perfect example of just your shift in, you know,

222
00:24:25,120 --> 00:24:30,960
kind of how you view execution of, okay, here's the way I've always done it, but wait, I have this

223
00:24:30,960 --> 00:24:34,960
other option over here. Let me see what's behind door number two and maybe that can be a better

224
00:24:34,960 --> 00:24:42,080
result. True, but also we need to talk about all the other aspects and we don't need to be specific

225
00:24:42,080 --> 00:24:49,440
about AutoGPT because one, it feels like in our environment, it's a huge thing. It's exploding,

226
00:24:49,440 --> 00:24:56,080
just the star's number on GitHub. I think it's the biggest project ever there, but the reality is

227
00:24:56,080 --> 00:25:03,760
that probably less than 0.1% knows about it. And, you know, one aspect is what happens when more

228
00:25:03,760 --> 00:25:09,120
people know about it, but on GitHub page, it very specifically says it's an experiment.

229
00:25:09,680 --> 00:25:15,600
Yeah. Right? So we also need to look into this, okay, this is an experiment. And for example,

230
00:25:15,600 --> 00:25:23,840
we know that ChatGPT will have plugins, access to very specific dedicated databases, like health

231
00:25:23,840 --> 00:25:33,200
database or Zapier or whatever database you can think of, plus browse on the internet. So for me,

232
00:25:33,200 --> 00:25:39,760
AutoGPT feels much more like a feature than a product on its own. And this is kind of like,

233
00:25:39,760 --> 00:25:46,800
you just start brainstorming, okay, what's possible when you have all this autonomous agents

234
00:25:46,800 --> 00:25:52,800
doing the things for you. And I very much think that, or at least I would like to imagine what

235
00:25:52,800 --> 00:26:00,560
the future is, the future is bright. And the way we work is actually that each of us have this

236
00:26:00,560 --> 00:26:07,440
dedicated little assistance, which actually use computers. So I like how mid-journey founder

237
00:26:07,440 --> 00:26:13,280
David was talking about that, how we interact with computers completely change from, you know, just

238
00:26:13,280 --> 00:26:20,160
keyboard, joystick, touch screen, mouse. What if we are moving A towards voice commands,

239
00:26:20,160 --> 00:26:25,280
purely just talking as you would talk with your assistant in a meeting. And then of course,

240
00:26:25,280 --> 00:26:31,840
we can talk about nanotechnology and circuits and directly into brain, like brain waves and all that.

241
00:26:31,840 --> 00:26:39,120
But that's probably 2024. So let's stay in 2023. That's next Thursday, Gona. That's next week's

242
00:26:39,120 --> 00:26:45,200
episode at this rate. But what are your fears? Do you have any fears after using this or you're

243
00:26:45,200 --> 00:26:50,400
completely positive? I am the cautious optimist with all of these, with a capital O in that

244
00:26:50,400 --> 00:26:56,720
optimism. Just because for me, and I've described this on a prior episode, the space it now frees

245
00:26:56,720 --> 00:27:04,480
up because I'm 80% there with a task is just so worth it to me. And it's my favorite thing in the

246
00:27:04,480 --> 00:27:11,520
world to show people that I'm sharing these tools with coaching, training, anything. That sense of

247
00:27:11,520 --> 00:27:18,080
wonder when they see this work for the first time and you could see it, this changes the way I could

248
00:27:18,080 --> 00:27:22,880
work. This changes the way I could plan how my family goes on a vacation. This changes the way

249
00:27:22,880 --> 00:27:29,920
I can interact with people. So I get more time to do the things I really enjoy. So I lean heavy into

250
00:27:29,920 --> 00:27:36,080
that. But I understand that people think that the cost is going to then be people being creative,

251
00:27:36,080 --> 00:27:42,880
people using it for malfeasance and bad intentions. Yeah, I mean, there's going to be that too, but

252
00:27:42,880 --> 00:27:49,280
hasn't that also occurred with every new technological kind of shift anyway? Absolutely.

253
00:27:49,280 --> 00:27:56,720
Yeah. People thought that internet was a fad. But another aspect why, for example, why I think that

254
00:27:56,720 --> 00:28:03,440
AutoGPT is more like, I very much see it as an experiment and more as a feature, the current

255
00:28:03,440 --> 00:28:10,480
setup of that and the cost. Because this is why I wanted us to touch on that because you still use

256
00:28:10,480 --> 00:28:18,320
your OpenAI key. And this is an interesting dynamic how actually the people who benefit from that is

257
00:28:18,320 --> 00:28:26,080
OpenAI. It's literally driving revenue. And because of each step and the way it's set up,

258
00:28:26,080 --> 00:28:32,880
it's using so many tokens. It's just burning through them. I don't want to say annoys me,

259
00:28:32,880 --> 00:28:39,360
but I see, for example, in LinkedIn, so many posts from people who are probably like me who don't code

260
00:28:39,360 --> 00:28:46,960
and don't know that much ins of this technology, promoting this as, hey, now businesses are going

261
00:28:46,960 --> 00:28:53,840
to be transformed. And this is amazing. And then telling even that it's for free. And it's just,

262
00:28:53,840 --> 00:29:01,360
we shouldn't promote, let's say, experiments, which are very insanely costly without a testing

263
00:29:01,920 --> 00:29:09,200
and deeper understanding. And as any startup, giving it time as an MVP to be in the market,

264
00:29:09,200 --> 00:29:15,440
to see how it evolves before we can make any kind of conclusions and start pitching this

265
00:29:15,440 --> 00:29:21,280
as transformative thing for businesses. There's definitely much more than just a monetary cost,

266
00:29:21,280 --> 00:29:27,040
too. I alluded to the fact that this gets expensive. The GPT-4 algorithm, which is the

267
00:29:27,040 --> 00:29:33,280
latest, the greatest, the newest for about every thousand tokens, so 750 words that goes in or out

268
00:29:33,280 --> 00:29:39,520
is about 12 cents. So if you have these two agents talking to one another, everything that they

269
00:29:39,520 --> 00:29:46,160
prompt, then go search, and then assemble is all tokens. So that example I gave about the family

270
00:29:46,160 --> 00:29:52,640
friend who's searching for colleges, just getting that information together was about $4.50 worth of

271
00:29:53,200 --> 00:29:59,680
tokens. So tremendously valuable to them. I had no problem doing that, but then I kind of sat back,

272
00:29:59,680 --> 00:30:06,160
I'm like, oh, I don't think I needed the latest and greatest to aggregate and pull this data. Let me

273
00:30:06,160 --> 00:30:11,760
be using 3.5 for some of this stuff. 4 might be better if I'm asking it to develop a book or

274
00:30:11,760 --> 00:30:19,680
screenplay or something much longer. On another flip side, we just mentioned about 19,000 forks

275
00:30:19,680 --> 00:30:25,680
on GitHub. I don't see that going away. And I don't know, have you joined Discord channel of

276
00:30:25,680 --> 00:30:32,560
AutoGPT? I think I'm on the one for Godmode. I haven't been very active in that yet, just because

277
00:30:32,560 --> 00:30:38,640
it's during this podcast, the time we've been recording this podcast, 7,500 new AI sites have

278
00:30:38,640 --> 00:30:43,360
actually come out. So there's a little bit of things to parse through these days in the world

279
00:30:43,360 --> 00:30:49,280
of AI. It is a exciting time no matter what, but I think just on that, that's probably a great place

280
00:30:49,280 --> 00:30:55,760
to wrap things up for this week. So for Go2Go and Wes the Synth Mind, happy prompting everybody.

281
00:30:55,760 --> 00:30:57,520
Happy prompting everybody.

282
00:30:57,520 --> 00:31:05,120
Thanks for listening to How to Talk to AI with your hosts, Go2Go and Wes the Synth Mind.

283
00:31:05,840 --> 00:31:11,520
As always, you can check out the show notes and links at howtotalkto.ai.

284
00:31:11,520 --> 00:31:26,960
That's all for this week's episode. Happy prompting everyone.