0:00

Welcome to the prompt engineering podcast,

0:03

where we teach you the art of writing effective

0:05

prompts for AI systems like chat,

0:07

GPT, mid journey, Dolly,

0:10

and more. Here's your host, Greg

0:12

Schwartz.

0:15

Hello, I'm currently sick with a cold, so

0:17

I'll be "speaking" via AI. This time

0:19

I'm doing it via the Descript app. This is

0:21

a quick episode because I wanted to discuss

0:23

Open A I Developer Day, and their mindblowing

0:25

accomplishment which you may not have realized

0:27

yet! First, a few highlights of the

0:29

opening keynote:

0:30

GPTs are tailored versions of ChatGPT

0:34

for a specific purpose. You

0:36

can build a GPT, a customized

0:38

version of ChatGPT, for almost anything.

0:41

With instructions, expanded knowledge,

0:43

and actions. And then you can publish it for others to

0:45

use. You can program a GPT just

0:48

by talking to it. And Later this

0:50

month, we're going to launch the GPT

0:52

store. You can list a GPT there and

0:55

we'll be able to feature the best and the most popular

0:57

GPTs. Revenue sharing is important to us.

1:00

We're going to pay people who build the most useful

1:02

and the most used GPTs a portion

1:04

of our revenue You want these models to be

1:07

able to access better knowledge about the world, so do

1:09

we. So we're launching retrieval

1:11

in the platform. You can bring knowledge

1:13

from outside documents or databases into

1:15

whatever you're building.

1:17

A special announcement.

1:19

I've just released my course, master

1:21

ChatGPT in two hours,

1:24

AKA how to not lose your job

1:26

to AI. I go over the basic

1:28

uses and techniques for prompt engineering,

1:30

the intermediate slide shot, prompting

1:32

and chain of thought. And then some advanced

1:35

ones. Plus there's guided practice

1:37

sessions to help you see exactly

1:39

how to use them. And by the way,

1:41

I have a 60 day money back guarantee.

1:43

So if you don't like it, let me know, and I will

1:45

give you a full refund. Normally, this is

1:47

$145. But I've discounted

1:50

it to $29. And

1:52

for listeners of the podcast use

1:54

coupon podcast. And

1:56

you'll get $20 off. So for

1:58

only nine bucks. You can learn

2:00

how to use Chatsy PT. Take

2:03

a look at the testimonials. Go get

2:05

the course at: promptengineeringpodcast.com/course1

2:09

that's numeral one, by the way. And the link

2:11

is also in the description. All right back

2:14

to the episode.

2:16

And three more things from Developer Day...

2:18

with our new text to speech model, you'll

2:20

be able to generate incredibly natural

2:22

sounding audio from text in the API

2:25

with six preset voices to choose from. I'll

2:27

play an example. Did you know that Alexander

2:30

Graham Bell, the eminent inventor, was

2:32

enchanted by the world of sounds? So

2:35

we're introducing Copyright Shield. Copyright

2:37

Shield means that we will step in and defend our customers

2:41

and pay the costs incurred if you face

2:43

legal claims around copyright infringement. And

2:45

this applies both to ChatGPT Enterprise

2:48

and the API. the assistance

2:50

API includes persistent threads.

2:52

So they don't have to figure out how to deal with long conversation

2:54

history. Built in retrieval,

2:57

code interpreter, a working Python interpreter

3:00

in a sandbox environment, and

3:02

of course, the improved function calling that

3:04

we talked about earlier

3:05

These are momentous moves. And there's more:

3:07

pricing, longer context window and

3:09

J son mode. However, did you get the mindblowing

3:11

news? Open AI created a moat, by

3:13

dramatically increasing switching costs.

3:16

Here's an explanation via a voice from eleven

3:18

labs.

3:19

Wikipedia defines switching costs.

3:22

Customers might be less likely

3:24

to change providers if the move

3:26

will incur costs, delays, or

3:28

effort. Until now,

3:31

switching from OpenAI was one

3:33

line of code. But no one else

3:36

provides uploading files, threading,

3:38

nor the GPT store. Before

3:41

developers built that, now OpenAI

3:43

does it and locks you in.

3:45

If you're still unsure why this is so momentous,

3:48

here is Open AI's new text to speech to explain.

3:50

Open AI makes it easier for developers

3:52

to build by providing those features.

3:55

Threads is valuable because although many

3:57

developers will still need an external

3:59

database to store other data for

4:02

many uses, threads is enough.

4:04

File uploading is more valuable because

4:06

it replaces very expensive alternatives

4:08

like Pinecone and GPT trainer, albeit

4:11

only if you have 25 files or

4:13

less, you can be sure that limit

4:15

will increase. The biggest

4:17

value is GPT store. It not

4:19

only makes it easier for people to find

4:21

your app because it's in the store,

4:24

just like the Apple app store and Google play

4:26

store. It allows you to create

4:28

simple applications without doing any

4:30

other work. Many startups

4:33

have been trying to build that kind of experience,

4:35

but none have even moderately succeeded.

4:38

Open AI, however, likely will.

4:41

Altogether, they changed your choice

4:43

of an AI provider from an easy

4:45

to replace commodity to something

4:47

with potentially huge switching costs.

4:50

Even if OpenAI raises their prices,

4:53

having to rebuild your application to store

4:55

conversations, threading, or integrate

4:57

a vector store, file uploading would

5:00

be very time consuming. And

5:02

developers using GPTstore will mostly

5:04

be people who do not want to build and maintain

5:07

a full application. So their

5:09

switching costs would be astronomical. All

5:12

in all, Dev Day created quite a moat.

5:15

By the way, if you want more ideas about moats,

5:17

read article "The New New Moats" by Jerry

5:19

Chen, linked in the description. Thanks

5:22

for coming to the prompt engineering podcasts podcast

5:25

dedicated helping you be a better

5:27

prompt engineer I also host

5:29

masterminds where you can collaborate with

5:31

me and 50 other people live on

5:34

zoom to improve your prompts Join

5:37

us at promptengineeringmastermind.

5:39

com for the schedule of the upcoming masterminds.

5:42

Finally, please remember to like

5:44

and subscribe. If you're listening to the

5:47

audio podcast, rate us five stars.

5:49

That helps us teach more people. And

5:51

if you're listening to the podcast, you might

5:53

want to join us on YouTube so you can actually

5:56

see the prompts. You can do

5:58

that by going to youtube. com

6:00

slash at prompt engineering

6:03

podcast. See

6:05

you next week.