EP 548 Testing 5 'Banned' AI Side Hustles | Online Opportunities Exposed

1
00:00:00,046 --> 00:00:05,064
API arbitrage using OpenAI and
Anthropic's APIs is making lots of

2
00:00:05,064 --> 00:00:07,290
people $15,000 a month right now.

3
00:00:07,386 --> 00:00:11,499
And five of these AI side hustles
from things like voice cloning with

4
00:00:11,499 --> 00:00:15,438
ElevenLabs, they violate the terms
of service from the very platforms

5
00:00:15,438 --> 00:00:17,273
that are powering these side hustles.

6
00:00:17,377 --> 00:00:21,255
Not because they're actually illegal,
but because once you see how much

7
00:00:21,255 --> 00:00:24,960
money moves through these gray zones,
you'll never be able to unsee it.

8
00:00:25,020 --> 00:00:26,255
Let's be s- specific here.

9
00:00:26,342 --> 00:00:29,446
The top earners in the AI
monetization space aren't the

10
00:00:29,446 --> 00:00:30,864
ones building chatbots out there.

11
00:00:30,864 --> 00:00:31,742
Mm, no, no.

12
00:00:31,838 --> 00:00:35,933
They're not the ones creating ethical
frameworks or even the ones waiting for

13
00:00:36,012 --> 00:00:39,838
permission for platforms that change
their terms of service every quarter.

14
00:00:40,046 --> 00:00:44,673
They're the ones operating in the spaces
between the rules where the money pools,

15
00:00:44,916 --> 00:00:48,142
where platforms haven't quite caught
up yet, and where your nine-to-five

16
00:00:48,212 --> 00:00:50,699
paycheck starts to look like a warmup lap.

17
00:00:50,933 --> 00:00:54,612
Unlike the videos that say, "Make
money with ChatGPT," the five methods

18
00:00:54,612 --> 00:00:57,620
I'm about to cover, starting with
API arbitrage, are the ones that

19
00:00:57,646 --> 00:00:59,725
nobody wants to name out loud.

20
00:00:59,899 --> 00:01:03,568
But by the end of this, you'll
understand why major creators delete

21
00:01:03,620 --> 00:01:07,394
videos about this very topic, why
some people have gotten cease and

22
00:01:07,394 --> 00:01:11,090
desist letters, and more importantly,
you'll know which ones actually work,

23
00:01:11,412 --> 00:01:15,455
which ones have expiration dates, and
which ones will flag your account in

24
00:01:15,455 --> 00:01:17,855
ways that take weeks to recover from.

25
00:01:17,933 --> 00:01:20,855
Not because I'm trying to scare you,
no, no, no, but because you deserve

26
00:01:20,881 --> 00:01:24,890
the unfiltered version before you
invest any of your time or any of

27
00:01:24,890 --> 00:01:28,394
your money into something that could
literally blow up in your face.

28
00:01:32,921 --> 00:01:33,382
What is up?

29
00:01:33,408 --> 00:01:33,826
What is up?

30
00:01:33,860 --> 00:01:35,782
What the hell is up, my dark
horse friends and family?

31
00:01:36,086 --> 00:01:39,060
Welcome back to another episode of
The Dark Horse Entrepreneur AI Escape

32
00:01:39,060 --> 00:01:42,226
Plan, where we break free from the
grind, share some AI-powered tips and

33
00:01:42,443 --> 00:01:45,608
side hustle strategies, and support
my amazing parents out there chasing

34
00:01:45,608 --> 00:01:47,356
freedom like modern-day trailblazers.

35
00:01:47,739 --> 00:01:50,234
And if you know someone who dreams
of ditching the grind but still

36
00:01:50,234 --> 00:01:53,747
makes time for those little lovely
chillins out there, you know, bedtime

37
00:01:53,747 --> 00:01:57,173
stories, snacks, whatever, go ahead
and forward this episode to them.

38
00:01:57,452 --> 00:02:00,321
Now, let's get stuck right in here,
'cause w- I got a little bit to cover

39
00:02:00,321 --> 00:02:03,060
and I wanna make sure I get, get it all
to you, and I wanna start with the one

40
00:02:03,217 --> 00:02:07,573
that's making the most money right now,
but it also has the shortest runway.

41
00:02:07,573 --> 00:02:11,417
And the first one is what I
call the API arbitrage play.

42
00:02:11,695 --> 00:02:12,904
And here's how it actually works.

43
00:02:13,052 --> 00:02:19,121
You take an API from OpenAI or Claude or
Gemini, GPT-4, GPT-4o, whatever it is,

44
00:02:19,391 --> 00:02:22,165
uh, specifically being resold, right?

45
00:02:22,313 --> 00:02:27,817
Something that costs you from $0 to one
cent per 1,000 tokens to run, right?

46
00:02:27,913 --> 00:02:28,591
Low cost.

47
00:02:28,782 --> 00:02:31,721
And then you resell access to
it through your own interface.

48
00:02:31,808 --> 00:02:35,852
Maybe you use a Discord bot or you create
your own platform, and you end up charging

49
00:02:35,852 --> 00:02:41,313
users 10 to 15 bucks per month for
unlimited access, he says with air quotes.

50
00:02:41,356 --> 00:02:42,826
The spread here is insane.

51
00:02:42,913 --> 00:02:44,739
Someone pays you, let's say, $29.

52
00:02:45,026 --> 00:02:49,347
You spend nothing to 50 cents in
actual tokens from their usage.

53
00:02:49,460 --> 00:02:49,939
What do you do?

54
00:02:50,165 --> 00:02:53,008
You pocket 28 bucks and
50 cents for one person.

55
00:02:53,165 --> 00:02:54,347
The catch is not hidden.

56
00:02:54,391 --> 00:02:56,913
It's just that no one
talks about it out loud.

57
00:02:57,139 --> 00:03:01,869
These platforms out there
explicitly forbid reselling API

58
00:03:01,869 --> 00:03:03,652
access in their terms of service.

59
00:03:03,791 --> 00:03:08,486
OpenAI's terms say you can't provide the
services to third parties as a service

60
00:03:08,513 --> 00:03:10,652
bureau or time-sharing arrangement.

61
00:03:10,895 --> 00:03:12,852
Claude's terms are very much similar.

62
00:03:13,104 --> 00:03:15,452
But here's where, uh,
the gray area comes in.

63
00:03:15,782 --> 00:03:19,217
Thousands of Discord communities
are doing this right this very

64
00:03:19,217 --> 00:03:20,582
minute as you're listening to me.

65
00:03:20,730 --> 00:03:21,200
Thousands.

66
00:03:21,304 --> 00:03:22,747
Not dozens, thousands.

67
00:03:22,826 --> 00:03:27,469
They're charging for exclusive access,
calling it a premium community,

68
00:03:27,539 --> 00:03:30,886
wrapping it all up in membership
language, and they're making anywhere

69
00:03:30,886 --> 00:03:35,678
from, I don't know, 3,000 to 15,000
a month before they get flagged.

70
00:03:36,069 --> 00:03:36,234
Yeah.

71
00:03:36,365 --> 00:03:40,008
And the expiration date here on this
one is getting shorter and shorter.

72
00:03:40,147 --> 00:03:42,791
OpenAI is starting to
actively crack down on this.

73
00:03:43,095 --> 00:03:47,060
They're looking for patterns, sudden
spikes in token usage, usage patterns that

74
00:03:47,060 --> 00:03:49,260
don't match individual behavior, right?

75
00:03:49,295 --> 00:03:52,200
Accounts that route traffic
through proxies, et cetera.

76
00:03:52,547 --> 00:03:54,991
And when they catch you,
they don't bother suing you.

77
00:03:54,991 --> 00:03:56,852
They just turn off your API key.

78
00:03:57,078 --> 00:03:59,486
No warning, no appeal
process worth mentioning.

79
00:03:59,617 --> 00:03:59,921
That's it.

80
00:04:00,182 --> 00:04:00,660
You're done.

81
00:04:00,730 --> 00:04:02,808
And the customers you've
built, well, guess what?

82
00:04:03,078 --> 00:04:04,547
They are gone overnight.

83
00:04:04,773 --> 00:04:08,165
But the people doing it right now before
the walls tighten, they're doing what

84
00:04:08,165 --> 00:04:10,417
we're gonna call extraction mining.

85
00:04:10,678 --> 00:04:14,234
They know that they only have about three
to six months before the enforcement gets

86
00:04:14,269 --> 00:04:16,060
tight and maybe even catches up with them.

87
00:04:16,321 --> 00:04:18,999
So they're not building
sustainable businesses, they're

88
00:04:18,999 --> 00:04:20,826
just harvesting revenue, okay?

89
00:04:21,043 --> 00:04:22,434
That brings us to n- the second one.

90
00:04:22,434 --> 00:04:24,765
The second one is a little
bit more insidious because

91
00:04:24,765 --> 00:04:26,704
it actually looks legitimate.

92
00:04:26,947 --> 00:04:30,669
Content generation at scale
with attribution loopholes.

93
00:04:30,852 --> 00:04:34,704
You build a system that generates
articles and social media posts or

94
00:04:34,704 --> 00:04:39,234
even video scripts using AI, and you
use tools like Midjourney for image

95
00:04:39,234 --> 00:04:43,400
assets, Stable Diffusion, uh, for open
source image generation, or Claude or

96
00:04:43,400 --> 00:04:48,408
GPT-4 for text, and then you license
this out to content creators and small

97
00:04:48,443 --> 00:04:50,669
businesses or even educational platforms.

98
00:04:50,808 --> 00:04:52,591
You're not claiming that AI wrote them.

99
00:04:52,660 --> 00:04:57,260
Technically, you're providing templates
that happen to be almost entirely

100
00:04:57,260 --> 00:05:02,999
generated by Claude or GPT-4 with
maybe five to ten percent human touches

101
00:05:02,999 --> 00:05:04,704
and adjustments made in the mix.

102
00:05:04,843 --> 00:05:05,773
Here's something worth noting.

103
00:05:05,895 --> 00:05:10,417
The EU AI Act Regulation EU, uh,
twenty twenty-four slash sixteen

104
00:05:10,417 --> 00:05:15,156
eighty-nine is now enforceable, with
Article 50 transparency obligations

105
00:05:15,399 --> 00:05:18,400
coming in full effect by August
of twenty twenty-six, August of

106
00:05:18,426 --> 00:05:20,243
this year that I'm recording this.

107
00:05:20,243 --> 00:05:21,991
It's about two and a half months away.

108
00:05:22,295 --> 00:05:27,686
It actually requires disclosure of
when AI generates content for certain

109
00:05:27,686 --> 00:05:30,147
regulated sectors, including education.

110
00:05:30,339 --> 00:05:34,257
So the legal exposure here is
beyond just the United States The

111
00:05:34,257 --> 00:05:38,326
educational niche, uh, is where
this seems to get the most money.

112
00:05:38,483 --> 00:05:43,553
A single well-built AI prompt system that
generates, say, study guides can pull

113
00:05:43,553 --> 00:05:48,674
in $5,000 to $20,000 per month because
schools and tutoring platforms, they're

114
00:05:48,883 --> 00:05:53,066
desperate for content, and they just do
not have enough human writers out there.

115
00:05:53,179 --> 00:05:56,196
So what you're really doing is
you're selling efficiency, not

116
00:05:56,344 --> 00:05:58,596
deception, and that's the frame, okay?

117
00:05:58,883 --> 00:06:02,344
But here's where the poison gets
leaked into this particular business.

118
00:06:02,387 --> 00:06:06,944
Most of these educational platforms are
explicitly telling their users that this--

119
00:06:07,013 --> 00:06:08,857
that their content is human-created.

120
00:06:09,153 --> 00:06:13,187
And if you know that and you're
feeding them generated content anyway,

121
00:06:13,387 --> 00:06:16,692
you're now a part of a chain that
deceives students and their parents.

122
00:06:16,770 --> 00:06:18,883
The platform gets caught, not you.

123
00:06:19,344 --> 00:06:21,587
Okay, there something a
little devious about that.

124
00:06:21,770 --> 00:06:25,274
But here's the thing, once one of
these platforms gets exposed, you

125
00:06:25,405 --> 00:06:29,631
are done, and suddenly there's an
investigation, and investigators ask

126
00:06:29,726 --> 00:06:31,648
suppliers about content sourcing.

127
00:06:31,883 --> 00:06:35,639
Stripe, PayPal, and other payment
processors have started quietly

128
00:06:35,639 --> 00:06:39,805
rejecting accounts, uh, that are
tied to educational AI content the

129
00:06:39,805 --> 00:06:41,492
moment they flag the deception.

130
00:06:41,709 --> 00:06:45,587
It's not because it's illegal,
but because it's a liability they

131
00:06:45,587 --> 00:06:47,179
don't wanna carry on their books.

132
00:06:47,318 --> 00:06:48,170
So what happens?

133
00:06:48,222 --> 00:06:51,867
Yeah, the money dries up
Without even an explanation.

134
00:06:51,954 --> 00:06:52,189
Okay?

135
00:06:52,493 --> 00:06:53,893
So let's move on to number three.

136
00:06:53,893 --> 00:06:57,702
The third one is where things
get genuinely unsettling,

137
00:06:57,702 --> 00:06:58,754
is the word I'm gonna use.

138
00:06:59,041 --> 00:07:02,398
AI voice cloning for
commercial use without consent.

139
00:07:02,763 --> 00:07:03,111
All right?

140
00:07:03,154 --> 00:07:04,798
Some of you are going, "Well, duh," right?

141
00:07:04,946 --> 00:07:07,302
But you've probably seen these
TikToks where someone else takes

142
00:07:07,302 --> 00:07:10,476
a celebrity's voice, usually a, a
deceased one like James Earl Jones,

143
00:07:10,476 --> 00:07:13,798
for example, or even someone that's
real and alive like Morgan Freeman.

144
00:07:14,033 --> 00:07:18,198
They clone it on a platform like
ElevenLabs or Descript, and they create

145
00:07:18,198 --> 00:07:22,789
a narration service for an audiobook, and
they sell it, and they put it on Audible,

146
00:07:23,024 --> 00:07:26,146
and they make $2,000 to $10,000 per title.

147
00:07:26,276 --> 00:07:28,172
Yeah, for just one, one book done.

148
00:07:28,285 --> 00:07:32,033
The legal framework here is just
starting to s- solidify, and it's

149
00:07:32,033 --> 00:07:33,867
not solidifying in your favor.

150
00:07:33,972 --> 00:07:38,441
Tennessee passed the Elvis Act,
E-L-V-I-S Act, and it's the Ensuring

151
00:07:38,441 --> 00:07:43,111
Likeness, Voice, and Image Security
Act, which explicitly protects voice as

152
00:07:43,233 --> 00:07:46,911
part of a person's protected identity,
including AI-generated replicas.

153
00:07:47,024 --> 00:07:51,728
California Civil Code 3344 covers the
right of publicity for living persons,

154
00:07:51,946 --> 00:07:57,624
and California AB 2602 extends that
specifically to AI-generated performances.

155
00:07:57,798 --> 00:08:03,598
At a federal level, the No Fakes Act i- is
pending, and the Nurture Originals Foster

156
00:08:03,598 --> 00:08:08,041
Act and the Keep Entertainment Safe Act,
which would create a, a nationwide right

157
00:08:08,102 --> 00:08:10,806
against unauthorized digital replicas.

158
00:08:10,946 --> 00:08:12,702
Man, all that was a mouthful to say.

159
00:08:12,754 --> 00:08:17,078
On top of that The C2PA, right, the
Coalition for Content Preservation

160
00:08:17,078 --> 00:08:21,052
and Authenticity, is establishing
watermarking standards that will make

161
00:08:21,052 --> 00:08:24,808
it increasingly difficult to pass
off cloned au- audio as legitimate.

162
00:08:25,113 --> 00:08:28,547
So as you can see, the law is
catching up faster and faster than

163
00:08:28,547 --> 00:08:30,756
most people doing this even realize.

164
00:08:31,000 --> 00:08:33,852
Right now, at this moment, at the
time I'm recording this, there's

165
00:08:33,852 --> 00:08:35,904
still no clear federal precedent.

166
00:08:36,156 --> 00:08:39,504
The estate of James Earl Jones,
for example, would argue it's

167
00:08:39,556 --> 00:08:40,852
their intellectual property.

168
00:08:40,904 --> 00:08:44,600
A real person like Morgan Freeman
would argue that his right of

169
00:08:44,600 --> 00:08:46,313
publicity under California law.

170
00:08:46,599 --> 00:08:49,973
But the patchwork of state laws
that are out there are quickly

171
00:08:49,973 --> 00:08:53,408
starting to fill in, and the federal
legislation is moving forward.

172
00:08:53,504 --> 00:08:57,147
This is one where I think most
creators go silent because the

173
00:08:57,147 --> 00:08:59,956
moral answer is obvious, right?

174
00:09:00,165 --> 00:09:01,408
For those of us who have morals.

175
00:09:01,521 --> 00:09:05,173
The legal answer is becoming less
and less certain by the month,

176
00:09:05,295 --> 00:09:08,747
and this shrinking uncertainty is
where the money starts leaving.

177
00:09:08,913 --> 00:09:11,365
Some people that are doing this
right now are making some real money.

178
00:09:11,556 --> 00:09:13,339
Others are getting cease
and desist letters.

179
00:09:13,426 --> 00:09:16,686
The difference isn't always about
how well-known the person is.

180
00:09:16,686 --> 00:09:19,626
Sometimes it's just about whether
their legal team is even paying

181
00:09:19,626 --> 00:09:21,121
attention to what's going on out there.

182
00:09:21,165 --> 00:09:23,947
Sometimes it's about whether
they have hired lawyers, and

183
00:09:23,947 --> 00:09:25,452
sometimes it's just luck.

184
00:09:25,756 --> 00:09:29,408
If you do this right now or are thinking
of doing it and start it, I want you

185
00:09:29,408 --> 00:09:31,513
to understand this is not a gray area.

186
00:09:31,782 --> 00:09:36,452
You're in a space where a letter from a
lawyer could end everything in one day.

187
00:09:36,817 --> 00:09:41,069
And unlike the A, uh, API arbitrage play,
this one doesn't have a warning system.

188
00:09:41,321 --> 00:09:44,982
You'll wake up to a cease and desist,
and then you'll have to decide whether

189
00:09:44,982 --> 00:09:48,304
you want to fight it or just fold
up shop and go somewhere else, okay?

190
00:09:48,443 --> 00:09:51,391
Look, if you're a parent out there
trying to build something real,

191
00:09:51,556 --> 00:09:54,956
something that doesn't put your family
at risk, you need strategies that

192
00:09:54,956 --> 00:09:57,121
work with the system, not against it.

193
00:09:57,286 --> 00:10:01,234
And that's exactly the type of thing we
cover in the AI Escape Plan newsletter,

194
00:10:01,495 --> 00:10:05,173
practical AI-powered strategies to start
growing, streamline your side hustles,

195
00:10:05,234 --> 00:10:08,086
all designed to protect your family
time while still boosting your income.

196
00:10:08,113 --> 00:10:10,573
You could call it your roadmap
to more money, more freedom,

197
00:10:10,730 --> 00:10:12,017
and more of what truly matters.

198
00:10:12,052 --> 00:10:15,939
No gray zones, no legal landmines,
just proven paths that respect

199
00:10:15,939 --> 00:10:17,234
your time and your values.

200
00:10:17,234 --> 00:10:19,895
Go ahead and go over to
darkhorseinsider.com and

201
00:10:19,895 --> 00:10:20,765
get signed up right away.

202
00:10:21,294 --> 00:10:22,884
Okay, quick break before
we dive into this.

203
00:10:22,930 --> 00:10:24,939
I have to share a song I've
been listening to lately.

204
00:10:25,248 --> 00:10:27,621
It's called "Calling All Angels" by Mero.

205
00:10:28,184 --> 00:10:32,275
It's a really emotional, kind
of haunting track, but in a

206
00:10:32,275 --> 00:10:33,939
really subtle, beautiful way.

207
00:10:34,316 --> 00:10:36,962
It actually came out of a difficult
time in their life, and you can

208
00:10:36,962 --> 00:10:38,107
feel that throughout the song.

209
00:10:38,107 --> 00:10:42,144
It has that feeling of reaching out
or searching for something and not

210
00:10:42,144 --> 00:10:43,753
really knowing if anyone's listening.

211
00:10:44,080 --> 00:10:46,707
The sound is super immersive too,
almost cinematic, but still really

212
00:10:46,707 --> 00:10:48,189
stripped back, which I love.

213
00:10:48,634 --> 00:10:51,453
Mero's just one of those artists where
the voice and writing feel really raw

214
00:10:51,453 --> 00:10:54,416
and original, and it's cool to see it
already getting attention from places

215
00:10:54,416 --> 00:10:57,007
like Flaunt, V Magazine, and Grammy.com.

216
00:10:57,407 --> 00:11:00,125
If you're into more emotional,
low-key music, I think

217
00:11:00,125 --> 00:11:01,134
you'll really like this one.

218
00:11:01,398 --> 00:11:03,016
I'll link it in the show
notes so you can listen.

219
00:11:03,353 --> 00:11:06,553
You can stream "Calling All Angels"
by Mero on Spotify, Apple Music,

220
00:11:06,816 --> 00:11:08,544
or wherever you listen to music.

221
00:11:08,898 --> 00:11:10,016
Okay, now let's get back into our

222
00:11:10,016 --> 00:11:12,208
Okay, so that takes us to number four.

223
00:11:12,277 --> 00:11:15,782
The fourth one has the longest
runway, and it's the one that actual

224
00:11:15,782 --> 00:11:17,834
sustainable income is coming from.

225
00:11:18,068 --> 00:11:22,408
And this one is the AI training
data curation and quality assurance,

226
00:11:22,590 --> 00:11:23,616
and you can do this at scale.

227
00:11:23,877 --> 00:11:24,677
Here's what's happening.

228
00:11:24,903 --> 00:11:29,008
Companies like Anthropic, Scale
AI, and OpenAI, they need human

229
00:11:29,008 --> 00:11:34,077
trainers to evaluate, to rank, and
to improve AI outputs through a

230
00:11:34,077 --> 00:11:38,816
process called RLHF, reinforcement
learning from human feedback.

231
00:11:38,912 --> 00:11:43,225
The platforms facilitating this
work include Scale AI's contractor

232
00:11:43,225 --> 00:11:47,512
network, Remote Tasks, as well as
Outlier, which runs similar projects

233
00:11:47,512 --> 00:11:49,121
for multiple AI labs out there.

234
00:11:49,338 --> 00:11:53,442
You can find these gigs out there right
now on Upwork and Fiverr or, or directly

235
00:11:53,442 --> 00:11:55,182
through these contractor networks, okay?

236
00:11:55,303 --> 00:11:57,703
They pay, I don't know, fifteen
to twenty-five bucks an hour.

237
00:11:57,868 --> 00:12:00,995
It sounds like pennies But here's
where it really gets interesting.

238
00:12:01,213 --> 00:12:04,908
If you build a system to manage
multiple trainers-- So what I'm saying

239
00:12:04,908 --> 00:12:08,395
is if you go out and recruit a team
of five to ten people who are out

240
00:12:08,395 --> 00:12:12,300
there doing all this work, and you
take a cut of their earnings, you're

241
00:12:12,326 --> 00:12:14,778
suddenly running an agency, right?

242
00:12:14,778 --> 00:12:18,378
And you could make ten to thirty
thousand dollars per month

243
00:12:18,778 --> 00:12:19,908
by just being that middleman.

244
00:12:19,908 --> 00:12:21,917
You don't even have to
do the work yourself.

245
00:12:22,108 --> 00:12:25,326
Your income from this gets reported
on a ten ninety-nine as a legitimate

246
00:12:25,326 --> 00:12:29,143
business arrangement, Schedule C,
self-employment income, which is

247
00:12:29,143 --> 00:12:32,656
actually one of the cleanest structures
from a payment processor standpoint.

248
00:12:32,717 --> 00:12:36,178
The legal and ethical question
is minimal here because you're

249
00:12:36,178 --> 00:12:37,604
working directly with the companies.

250
00:12:37,769 --> 00:12:38,761
They know what's happening.

251
00:12:38,917 --> 00:12:43,422
They don't care as long as the work gets
done and the trainers are real humans in

252
00:12:43,422 --> 00:12:45,074
countries who are actually doing the work.

253
00:12:45,204 --> 00:12:48,448
Some countries have labor laws that
would complicate this a little bit,

254
00:12:48,735 --> 00:12:51,395
but if you're operating in the US
and in Europe, you're basically fine.

255
00:12:51,482 --> 00:12:55,908
The catch is that this requires real
people, real time, and real management.

256
00:12:55,987 --> 00:12:59,022
So it's not passive, but
it's also not going away.

257
00:12:59,022 --> 00:13:03,082
The industry, I think, is going to
expand for a few years here until they

258
00:13:03,082 --> 00:13:05,108
find some better way to train the AI.

259
00:13:05,378 --> 00:13:07,613
I'm not sure when that's
gonna happen or if.

260
00:13:07,795 --> 00:13:11,404
So that brings us to number five, and
the fifth one here I need to be most

261
00:13:11,404 --> 00:13:13,517
careful with the words, so stick with me.

262
00:13:14,004 --> 00:13:17,482
Data licensing agreements
that border on violation.

263
00:13:17,752 --> 00:13:19,178
And here's, here, here's
what I mean by this.

264
00:13:19,343 --> 00:13:21,030
You scrape publicly available data.

265
00:13:21,074 --> 00:13:23,961
We're talking things like tweets
and blog posts and Reddit comments.

266
00:13:24,204 --> 00:13:29,074
And then you license this scraped data
to AI companies as training data, right?

267
00:13:29,074 --> 00:13:32,465
You're technically not breaking
any terms of service if the data is

268
00:13:32,561 --> 00:13:34,482
publicly available, public-facing.

269
00:13:34,813 --> 00:13:39,082
But you might be violating copyright
laws if the data is copyrighted material,

270
00:13:39,152 --> 00:13:41,795
and the DMCA creates real exposure here.

271
00:13:42,074 --> 00:13:44,526
That's the Digital
Millennium Copyright Acts.

272
00:13:44,552 --> 00:13:49,430
Its provisions on reproduction apply
even to public-facing content when

273
00:13:49,430 --> 00:13:51,439
it gets used commercially, okay?

274
00:13:51,604 --> 00:13:55,143
The United States Copyright Office
has also signaled that AI training

275
00:13:55,143 --> 00:13:57,126
data sourcing is under active review.

276
00:13:57,378 --> 00:13:58,152
Here's what this means.

277
00:13:58,335 --> 00:14:01,482
It means that the regulatory
ceiling here is getting lower.

278
00:14:01,552 --> 00:14:02,369
And here's the weird part.

279
00:14:02,717 --> 00:14:07,164
Some AI companies will buy
this data Knowing it's a legal

280
00:14:07,434 --> 00:14:09,182
gray area, they do it anyway.

281
00:14:09,399 --> 00:14:12,060
They know that if something
goes wrong, the supplier is

282
00:14:12,060 --> 00:14:13,547
gonna take the heat, not them.

283
00:14:13,782 --> 00:14:16,860
Now, someone could make three thousand
to eight thousand a month doing this,

284
00:14:17,260 --> 00:14:21,155
but they're essentially betting that
no one with a legal team is gonna trace

285
00:14:21,155 --> 00:14:23,260
that data all the way back to them.

286
00:14:23,416 --> 00:14:25,286
That's a bet that could
easily go sideways.

287
00:14:25,286 --> 00:14:25,877
You, you feel me?

288
00:14:25,929 --> 00:14:26,512
Yeah, okay.

289
00:14:26,790 --> 00:14:28,651
But here's the floor beneath all this.

290
00:14:29,077 --> 00:14:32,790
Every single one of these methods
work because of a timing gap.

291
00:14:32,877 --> 00:14:36,521
What I mean by that is the technology
is moving faster than any of the

292
00:14:36,521 --> 00:14:38,425
regulations and the enforcement.

293
00:14:38,425 --> 00:14:40,868
The enforcement moves even
slower than the adoption.

294
00:14:40,955 --> 00:14:44,242
And given that gap, money flows
to the people who understand

295
00:14:44,242 --> 00:14:47,468
that gap even exists in the first
place and are willing to use it.

296
00:14:47,738 --> 00:14:49,025
Here's what no one is gonna tell you.

297
00:14:49,251 --> 00:14:52,973
The moment you make enough money from any
one of these methods, you become visible.

298
00:14:53,268 --> 00:14:55,590
Platforms, they're great
at noticing patterns.

299
00:14:55,790 --> 00:14:58,555
Payment processors, yeah,
they're gonna flag volume.

300
00:14:58,634 --> 00:15:02,164
And let's be honest, the tax
authorities, they're gonna ask questions.

301
00:15:02,373 --> 00:15:05,677
And the thing that made you fifteen
thousand dollars last month could

302
00:15:05,677 --> 00:15:07,521
become a liability next month.

303
00:15:07,616 --> 00:15:11,277
The sustainability play isn't to pick
one of these that I've talked about here.

304
00:15:11,347 --> 00:15:16,008
It's to understand which one has
the longest runway wherever you are,

305
00:15:16,034 --> 00:15:17,355
wherever you live in your region.

306
00:15:17,555 --> 00:15:21,277
And then which payment processors
are still gonna be-- accept them,

307
00:15:21,468 --> 00:15:25,425
and which ones are the ones that
are already in the descent phase.

308
00:15:25,547 --> 00:15:25,912
Okay?

309
00:15:25,999 --> 00:15:27,460
So let me give you a, a quick breakdown.

310
00:15:27,460 --> 00:15:29,982
The, the API arbitrage
play, already descending.

311
00:15:30,190 --> 00:15:32,182
Voice cloning, I think
that one's uncertain.

312
00:15:32,182 --> 00:15:36,442
Uh, I believe it's heading down, and the
Elvis Act is accelerating that descent.

313
00:15:36,503 --> 00:15:39,303
Data curation through remote
tasks in Outlier, like I

314
00:15:39,303 --> 00:15:40,999
mentioned earlier, is ascending.

315
00:15:40,999 --> 00:15:44,573
It's going up, and I think it's gonna stay
that way for at least two or three years.

316
00:15:44,773 --> 00:15:48,660
Data licensing, kinda neutral, but
it's getting tighter, you know, with

317
00:15:48,660 --> 00:15:52,938
the DMCA enforcement and the Copyright
Office scrutiny catching up to it.

318
00:15:53,324 --> 00:15:56,046
The real income for the
nine-to-five workers isn't in

319
00:15:56,046 --> 00:15:57,350
doing one of these things alone.

320
00:15:57,620 --> 00:16:01,289
It's in understanding that each
one of them have a lifespan.

321
00:16:01,333 --> 00:16:04,698
Some of them, it's a few months, maybe
it's a year if you're strategic, and

322
00:16:04,698 --> 00:16:05,872
then you move on to the next one.

323
00:16:06,463 --> 00:16:09,602
A-and remember, if you pick one of
these, which is why I'm sharing this

324
00:16:09,602 --> 00:16:14,176
with you, you're not building a business,
you're just building harvesting windows.

325
00:16:14,228 --> 00:16:17,142
You're not even planting seeds, you're
just going out and harvesting someone

326
00:16:17,142 --> 00:16:19,594
else's seeds and then walking away.

327
00:16:19,663 --> 00:16:23,794
And yes, there's a moral weight here that
you're gonna need to carry consciously.

328
00:16:24,011 --> 00:16:25,185
Are you deceiving users?

329
00:16:25,342 --> 00:16:26,698
Are you stealing voices?

330
00:16:26,811 --> 00:16:28,333
Are you exploiting loopholes?

331
00:16:28,481 --> 00:16:32,437
Some of these, the answer is yes, and
knowing that answer and doing it anyway,

332
00:16:32,698 --> 00:16:35,811
that's the choice that you're gonna have
to make when the money is good enough.

333
00:16:35,994 --> 00:16:40,698
So whether you're looking at remote
task or Upwork for RLHF work, weighing

334
00:16:40,698 --> 00:16:44,089
the Elvis act before touching voice
cloning with Eleven Labs or Descript,

335
00:16:44,402 --> 00:16:48,802
navigating OpenAI's, uh, terms of
service in an arbitrage play, or figuring

336
00:16:48,802 --> 00:16:53,898
out where the EU AI Act's Article
50 draws the line on your content

337
00:16:53,898 --> 00:16:57,246
licensing, every single one of these
businesses I've mentioned here, all

338
00:16:57,246 --> 00:16:58,975
five of them They have a clock on them.

339
00:16:59,314 --> 00:17:03,114
The ones that work longest are the
ones where you're not deceiving anyone.

340
00:17:03,227 --> 00:17:06,088
The ones that blow up fastest
obviously are the ones where you are.

341
00:17:06,305 --> 00:17:10,662
The difference between five figures
and 5,000 is usually how willing you

342
00:17:10,662 --> 00:17:12,992
are to sit inside that discomfort.

343
00:17:13,235 --> 00:17:15,157
And some of you are not,
while some of you are.

344
00:17:15,383 --> 00:17:16,357
And why does this even matter?

345
00:17:16,601 --> 00:17:19,383
Well, because we're living through
let, let's call this the Wild

346
00:17:19,383 --> 00:17:21,331
Wild West of AI monetization.

347
00:17:21,383 --> 00:17:25,088
It's like the early internet, and just
like w- the early days of social media in

348
00:17:25,088 --> 00:17:28,609
its infancy, there's a window where the
rules haven't caught up to reality yet.

349
00:17:28,792 --> 00:17:30,522
And it's not just about making money.

350
00:17:30,705 --> 00:17:34,157
It's about understanding how
technology adoption cycles work.

351
00:17:34,296 --> 00:17:38,627
Every major technological shift
creates these gray areas where early

352
00:17:38,627 --> 00:17:42,922
adopters can extract some value before
the regulators catch up to them.

353
00:17:43,192 --> 00:17:46,357
The bigger paradigm here is we're
watching the birth of an entirely

354
00:17:46,357 --> 00:17:48,314
new economy, the AI economy.

355
00:17:48,375 --> 00:17:51,305
And like every economic shift,
there's gonna be winners and

356
00:17:51,305 --> 00:17:52,288
there are gonna be losers.

357
00:17:52,470 --> 00:17:54,418
And the winners understand timing.

358
00:17:54,679 --> 00:17:58,566
The losers chase tactics
without understanding the

359
00:17:58,644 --> 00:18:00,079
underlying forces involved.

360
00:18:00,340 --> 00:18:03,348
So all that said, here's your
Whisker wisdom for this week.

361
00:18:03,479 --> 00:18:07,279
The most dangerous money is the
money that feels too damn easy.

362
00:18:07,392 --> 00:18:08,914
Yes, I've been there, trust me.

363
00:18:09,209 --> 00:18:11,948
When something is making you thousands
of dollars a month and you're not

364
00:18:11,948 --> 00:18:15,983
sure why or if it's legal, that, my
friends, is not a business model.

365
00:18:16,253 --> 00:18:17,618
That is a countdown timer.

366
00:18:17,775 --> 00:18:20,244
So here's your specific
action for this week.

367
00:18:20,470 --> 00:18:24,435
Pick one legitimate AI skill and
spend the next 30 days getting

368
00:18:24,435 --> 00:18:26,244
genuinely good at that skill.

369
00:18:26,401 --> 00:18:31,505
Not gray zone good, not loophole good,
actually good, whether it's prompt

370
00:18:31,505 --> 00:18:35,201
engineering, whether it's getting AI
training data work through legitimate

371
00:18:35,201 --> 00:18:39,270
platforms like Remote Task, or maybe it's
building AI assistant content systems

372
00:18:39,496 --> 00:18:41,227
that you own and that you control.

373
00:18:41,462 --> 00:18:44,914
Look, uh, what I'm telling you here is I
think the goldmine is real, but the front

374
00:18:44,940 --> 00:18:47,592
door is not through some dark back alley.

375
00:18:47,783 --> 00:18:50,975
You gotta go through the front door, even
if you have to pay the fee to get there,

376
00:18:50,975 --> 00:18:53,288
and sometimes paying that fee is in time.

377
00:18:53,557 --> 00:18:57,401
Look, I'm not here to judge anyone
for the choices that they make when

378
00:18:57,401 --> 00:18:58,714
they're trying to feed their families.

379
00:18:58,870 --> 00:19:01,688
No, I've made some questionable choices
over the course of my years as well.

380
00:19:01,948 --> 00:19:05,644
But I am here to make sure you
understand what those choices

381
00:19:05,835 --> 00:19:06,879
are really gonna cost you.

382
00:19:07,183 --> 00:19:10,479
The sustainable path to AI
income isn't in those gray zones.

383
00:19:10,844 --> 00:19:14,427
It's in understanding the technology
well enough to build something that

384
00:19:14,427 --> 00:19:18,262
doesn't require you to keep looking over
your shoulder or jump every time a blue

385
00:19:18,262 --> 00:19:20,053
light flashes in your rear view mirror.

386
00:19:20,340 --> 00:19:25,148
Guard your time, grow your freedom,
and remember, the best side hustles are

387
00:19:25,148 --> 00:19:27,209
the ones that y- you sleep at night.

388
00:19:27,522 --> 00:19:29,122
Think successfully and take action.