220: The Dark Truth of AI as a Data Analyst

Speaker: 00:00:00

Avery Smith-1: So there's a pretty dark

future of AI that I don't hear a lot of

2

: 00:00:02

people talking about, and no, I'm not

even talking about all of the ethics

3

: 00:00:06

of how AI is ruining the environment

or how, you know, it's dehumanizing

4

: 00:00:11

the world or that people are just

trusting in all these all-powerful

5

: 00:00:14

AI chatbots and making dumb choices.

6

: 00:00:16

Although, we can argue those

in the comments if you'd like.

7

: 00:00:19

But for now, I'd really like to

focus on these dark costs of using

8

: 00:00:23

AI as a data analyst right now,

or a data professional really.

9

: 00:00:26

And for your career at this very

second, I think, like, ChatGPT

10

: 00:00:30

and Claude are fooling you.

11

: 00:00:32

I think that if you're using AI heavily

right now, it's either going to make or

12

: 00:00:36

break your career and, uh, I'll give you

all my arguments for that in this episode.

13

: 00:00:40

But if you're new here,

my name's Avery Smith.

14

: 00:00:43

I'm a senior data analyst, and I

try to make helpful content about

15

: 00:00:45

the future of data careers and

how AI and data play together.

16

: 00:00:49

And if you're interested in that,

which, let's be honest, you probably

17

: 00:00:51

are 'cause you're here, right?

18

: 00:00:52

Then go ahead and hit

the subscribe button.

19

: 00:00:54

But obviously, AI is here,

and it's really revolutionary.

20

: 00:00:57

Admittedly, I haven't written a line

of code unassisted in probably, like,

21

: 00:01:01

two-plus years now, and that's amazing.

22

: 00:01:03

It's opened up windows I

never even thought possible.

23

: 00:01:06

It's made me a better, a faster data

analyst and a developer, and it's

24

: 00:01:09

allowed me to build helpful products for

the data community like My Datafolio,

25

: 00:01:13

which is an online data portfolio

hosting platform, and findadatajob.com,

26

: 00:01:17

which is my job site that I

post lots of data jobs there.

27

: 00:01:20

And you probably feel the same.

28

: 00:01:21

Like, AI has really helped you

in a lot of ways in your life.

29

: 00:01:24

Like, maybe it's helped you look things

up faster, like no more Google, right?

30

: 00:01:27

Woo-hoo.

31

: 00:01:27

Maybe it's helped you

write SQL queries faster.

32

: 00:01:29

Maybe it's helped you with your

taxes, booked flights, whatever.

33

: 00:01:31

I don't know, all sorts

of different things.

34

: 00:01:33

But for all those good things, I

think there is a dark side that

35

: 00:01:36

both you and I are really ignoring

right now, and to our own detriment.

36

: 00:01:40

And no, it's not the obvious one of

you're actually learning the skills.

37

: 00:01:44

I think that's important,

but it's one that's much more

38

: 00:01:46

hidden and much more diabolical.

39

: 00:01:48

And it's important to know the skills.

40

: 00:01:50

Like, if AI's doing something,

you should be able to correct it.

41

: 00:01:53

You should be able to know if it's

actually giving you the right answers

42

: 00:01:56

to confirm what's actually going on.

43

: 00:01:57

And if you can't do that and you give that

analysis to a stakeholder, then there's

44

: 00:02:01

potential for a lot of big problems.

45

: 00:02:03

You know, maybe that's a big loss

in dollars, but it could be loss

46

: 00:02:05

in time, productivity, or even

lives at the worst case, right?

47

: 00:02:08

So if you're relying heavily on AI right

now, you might not be learning the core

48

: 00:02:12

skills necessary to actually become

a truly powerful data professional.

49

: 00:02:15

And that is serious, and

that is a problem, but you

50

: 00:02:18

probably already know that.

51

: 00:02:20

So what if I told you instead that

there's even a bigger rug pull at play,

52

: 00:02:24

one that you maybe haven't heard of yet,

that there's something slowly spreading

53

: 00:02:28

quietly in the AI and data industry that

people are quietly whispering about, not

54

: 00:02:32

really talking it out loud, and it's this:

that the biggest players in the space,

55

: 00:02:35

Claude and ChatGPT, are scamming us.

56

: 00:02:38

They're setting us up for an epic

realization that's going to leave us

57

: 00:02:41

all shell-shocked and kind of screwed.

58

: 00:02:43

And it's this: the current

AI prices are fake, fugazi.

59

: 00:02:46

They're literally artificial, and

I totally mean that to be a pun.

60

: 00:02:50

"But Avery, what do you mean?"

61

: 00:02:51

Guys, these prices,

they're subsidized heavily.

62

: 00:02:54

Every time you send an AI, they

are charging you one cent or

63

: 00:02:57

something like that, but it's

really costing them two to reply.

64

: 00:03:00

These companies are literally losing

money, literally burning it, and not just

65

: 00:03:04

a little bit of it, a lot of it, millions

and millions of dollars, hundreds of

66

: 00:03:08

millions of dollars they're just literally

throwing out the window, all in an effort

67

: 00:03:11

to get us addicted, to get us hooked.

68

: 00:03:13

And when we become reliant on it and we

can't live without it, that's when they'll

69

: 00:03:17

finally raise prices to what they actually

should be And if you don't believe me, no

70

: 00:03:20

problem, you know, don't trust me, right?

71

: 00:03:23

A recent study showed that

current AI prices from ChatGPT

72

: 00:03:25

and Claude are subsidized by up

to 70 times their actual prices.

73

: 00:03:30

Meaning if you have a Claude Max plan or

a ChatGPT Max plan like me, you're really

74

: 00:03:34

getting like 20 times, 70 times the usage

for what you're actually paying for.

75

: 00:03:38

Meaning if you do the same amount

of usage at actual break-even prices

76

: 00:03:42

for these companies, you're literally

going to be paying hundreds of dollars

77

: 00:03:45

more than you are right now to do

the same type of analysis, to do the

78

: 00:03:48

same type of queries, to do the same

types of chats that you're doing.

79

: 00:03:51

So just come with me for a second

and imagine that you don't learn the

80

: 00:03:54

fundamentals of SQL and Python, and right

now you're able to do some pretty decent

81

: 00:03:58

analysis just with ChatGPT or Claude.

82

: 00:04:00

You can do some pretty cool graphs.

83

: 00:04:01

You can make some pretty cool scripts.

84

: 00:04:02

You can do some pretty cool queries,

and maybe it costs you $10 to

85

: 00:04:05

generate all of these different

reports that, what you want.

86

: 00:04:07

Well, what happens down the road

when you want to generate a similar

87

: 00:04:10

report, but instead of $10 it costs

you $700 to do something similar?

88

: 00:04:15

Are you personally willing

to pay that amount?

89

: 00:04:17

Do the economics actually make

sense for you in that case, and

90

: 00:04:20

do they make sense for the company

that you'll be working for?

91

: 00:04:22

What you think you can do with AI

right now for $100 will not be what

92

: 00:04:27

you can do in the future for $100.

93

: 00:04:29

You'll be able to do a

lot less in the future.

94

: 00:04:31

Now, how far into the future?

95

: 00:04:33

I'm not going to just guess and

give you the timeline, but there's

96

: 00:04:36

a lot of hints being left right now.

97

: 00:04:38

Both Anthropic and OpenAI have had

discussions recently of IPO-ing and

98

: 00:04:41

becoming a publicly traded company,

and when that happens, their financials

99

: 00:04:45

will get scrutinized a lot more.

100

: 00:04:47

They have to actually publish all the

different, you know, financials that

101

: 00:04:49

public companies have to do, and when

that happens, they will feel the public

102

: 00:04:53

and shareholder pressure of this insane

cash burning that they're doing right now.

103

: 00:04:57

They will eventually need to have

profit or at least way less loss

104

: 00:05:01

than they're currently happening.

105

: 00:05:02

They can't just continue

to burn money forever.

106

: 00:05:05

So this may be in just a few

weeks, it may be in a few

107

: 00:05:07

months, it may be even in years.

108

: 00:05:09

I don't know.

109

: 00:05:09

But I think it's going to happen,

and there's a chance that these

110

: 00:05:12

prices not only go up, but go up

from what they are today, and it's

111

: 00:05:15

important for you to know that.

112

: 00:05:17

Now, why is that the case?

113

: 00:05:18

Why are prices so high and why

might they go higher in the future?

114

: 00:05:21

Well, I think there's two key factors,

basically, in all of AI: the hardware

115

: 00:05:24

and the power for the hardware.

116

: 00:05:25

Now, historically, hardware and computer

chips have gotten cheaper over time

117

: 00:05:29

and better over time, but that doesn't

actually really factor in the idea of

118

: 00:05:33

the demand that we're in right now.

119

: 00:05:35

We've seen great demand for

GPUs with cryptocurrency,

120

: 00:05:38

with artificial intelligence.

121

: 00:05:39

Demand is way up, and if supply

can't actually meet the demand,

122

: 00:05:43

we, you know, have an imbalance

and we may see prices go up.

123

: 00:05:46

Similarly, the power needed to power

AI and all these, you know, data

124

: 00:05:50

farms and all these compute farms,

all these data centers around the

125

: 00:05:53

world is going up, and the electricity

prices are going up as well.

126

: 00:05:58

Electricity prices in the US have nearly

increased 50% in the last five years.

127

: 00:06:02

As you can see from this graph right

here, it is a crazy big increase,

128

: 00:06:06

and I don't really think there's

any sign of this increase stopping.

129

: 00:06:09

Like electricity is just going to get

more and more expensive, and as it does,

130

: 00:06:13

so will the price of AI, unfortunately.

131

: 00:06:15

So what's actually my point here?

132

: 00:06:17

Well, if you think that you can

use AI to do data a- analytics for

133

: 00:06:20

the rest of your life and the rest

of your career, you may be right.

134

: 00:06:22

There's also a chance that you'll

actually be using it less than you think.

135

: 00:06:26

If the prices make it uneconomical

for companies, they will have to

136

: 00:06:29

take action and take quick action.

137

: 00:06:31

We've actually already started

to see it, whether you actually

138

: 00:06:34

have heard about it or not.

139

: 00:06:35

Microsoft, obviously a huge tech company,

literally cut their engineers from

140

: 00:06:40

AI because of how big the bill was.

141

: 00:06:41

Uber is now capping their employees

after they blew the entire AI budget

142

: 00:06:46

for the entire year in one quarter.

143

: 00:06:48

And even tech giants like Meta and Amazon

are feeling the costs, and if they're

144

: 00:06:51

feeling it, we're all probably feeling it

as, you know, smaller fish in the pond.

145

: 00:06:56

Now, that's not to say that you should

just start ignoring AI and just do all

146

: 00:07:00

your SQL queries and Python code by

hand, because you really shouldn't.

147

: 00:07:03

I think that's a big mistake as well.

148

: 00:07:05

AI is going to be used in the future,

and you as a data professional

149

: 00:07:09

need to be paying close attention

and getting practice on it.

150

: 00:07:12

But at the same time, maybe you

shouldn't be token maxing and just,

151

: 00:07:15

like, running up your AI bill like crazy

without actually having foundational

152

: 00:07:19

knowledge of what you're actually doing

and what you're actually building.

153

: 00:07:22

It's really a delicate balance to use

AI enough to stay dangerous, but not use

154

: 00:07:26

it enough that you become dependent, and

it's a balance that I'm exploring on my

155

: 00:07:30

own as well, and I'm actually gonna be

talking about it a lot in my newsletter.

156

: 00:07:34

So you should consider subscribing.

157

: 00:07:36

It's datacareernewsletter.com,

158

: 00:07:37

and I send literally meaningful

advice every single week on the

159

: 00:07:40

future of data analytics and AI.

160

: 00:07:42

There will also be a link in

the description down below.

Share Episode

Shownotes

Transcripts

Follow

Links

Chapters

Video

More from YouTube