1:55

I

2:00

wrong it is a lot. It's a daily right?

2:02

Yeah, we haven't missed a day The goal was every

2:05

weekday morning have a new one ready to go and

2:07

four years coming up right on four years. That's

2:09

crazy That's fun. Yeah, that's

2:12

nuts And that's like navigating

2:14

around things that are outside of your control

2:17

like your voice working and that's that's really

2:19

hard to do Which

2:21

is why I have programmed this

2:23

AI using a learning model

2:26

based on the 10-minute Bible hour podcast right

2:28

now Would you like me to play some of it for you? Oh, that'd be

2:30

spectacular. Thank you. All right, listen to this Herpes

2:33

aren't the worst experience of my life

2:36

It's probably the double herpes that

2:38

I would say is the worst experience

2:40

of my life. Also make sure to read your

2:42

Bible React to that. What do you think about

2:44

what I just did? I'm very disappointed I would

2:47

never sound a something like that. I don't know just

2:49

like you did that was it sounded

2:51

a lot like me Listen to this what all I mean

2:54

and this one if I were stranded

2:56

on a desert island with Hitler and

2:59

four monkeys and I had a submarine

3:01

that could only take the four monkeys off of the

3:03

island But if so, I'd have to eat Hitler. I

3:06

wish I wasn't on that island React

3:08

to that one. What did you think? I think those

3:10

AI learning models they degrade over time Yeah, make

3:13

less sense that one wasn't as good. But I mean

3:15

you've given me so much data, huh? I Hear

3:19

I'm gonna just throw one out there. I'm gonna just have

3:21

it make the most offensive Arrangement

3:24

of words I can listen to this and go What

3:30

do you think about that one?

3:31

Well, I mean it was just

3:32

it was just

3:33

all the swears in order. I Don't

3:36

think I literally have ever heard a swear that wasn't

3:39

in that list How long do you think it took me

3:41

to train that model in order to use your

3:43

voice like that? Year two years. Yeah.

3:46

Yeah, actually I didn't I just got you to play along

3:48

and just say things at a microphone No,

3:55

but you're you're using AI

3:57

to do what? You're well, at least you're

3:59

testing it Or is it even AI testing?

4:02

If it's even AI, and is any of the

4:04

AI we're using really even AI? It's

4:07

all just modeling at this point. I don't

4:09

think it is Skynet at this point, it's

4:11

my understanding. We're repurposing existing

4:14

material and reformatting it based

4:17

on pattern recognition, but I don't think we're

4:20

doing thought. I don't think it's true artificial

4:22

intelligence. It's sort of a workflow

4:24

optimization. Yeah. Put technology

4:26

at this point. But you're the first of Twitter

4:29

or whatever you call that platform now, just rife

4:32

with all kinds of threads about,

4:34

oh, I use this language

4:37

model to monetize

4:39

my whatever. If you're not doing this, you're not,

4:42

I don't know, man, these influencers today. But

4:45

what are you actually trying to do? You're taking

4:47

your existing back catalog of the 10 Minute

4:49

Bible Hour podcast, and

4:52

you are trying to do what with it? All

4:54

I'm looking for is help with editing. I'm

4:58

not actually trying to use an AI

5:00

editor, but I'm okay

5:02

with working with a remote editor who

5:05

does use AI to crawl

5:08

a bunch of different stock footage

5:11

services and grab things

5:13

that might work and give me ideas

5:15

to brainstorm and to get things going for

5:18

sort of a, I guess like a video

5:20

essay kind of format. You've seen those on YouTube where

5:22

you don't see the person's face. You

5:24

put together a little video essay and then they cover it with

5:27

stock footage and relevant images that

5:29

get the point across. Yeah. Even

5:32

though really it was a podcast. Yeah. What's

5:34

interesting is you've got a person talking and then every four

5:36

to five seconds or I don't know, maybe it's

5:38

even lower than that, three to six or whatever,

5:41

you have an image change. I think it's

5:43

interesting because the visual part of your brain

5:45

is so powerful that you look at these

5:48

images and you associate it with what the person is saying.

5:51

It's almost like someone listens

5:53

to the podcast and if you say cake, they

5:55

would just go to some stock footage

5:58

site and they would look up cake. Yes.

6:01

And then like five images of cake come up

6:03

and they pick the one that they want to go in there. Yeah.

6:06

Yeah. And so figuring out, you know, what

6:09

role AI could play in the first

6:11

pass or in collecting a bunch of good options

6:14

for stock images or footage, that's

6:16

not really the goal. The goal is I made

6:19

a series of podcasts that through the books of

6:21

the Bible stuff is eight to 13

6:23

minutes of pop. And it's

6:25

just, I think, a really, I hope, helpful

6:28

introduction and survey, a covering

6:30

of the themes and all of that in a playful, fun way

6:33

of each book of the Bible. So what I want to

6:35

do is put that on the YouTube channel, but

6:37

I want something to be happening visually that

6:39

adds more value to that. Charts,

6:42

facts, figures, maps, classical

6:45

art, pictures of artifacts

6:48

and relics and ruins and things like

6:50

that, just so people can see and feel

6:52

and connect with the Bible a little bit

6:54

more and have that kind of support and bolster

6:57

the words I'm saying. What's up going on the screen,

6:59

right? Pretty simple text on screen, Bible verses,

7:01

et cetera. I bought a set of books

7:03

a long time ago to use bookstore. I think

7:05

it was Matthew Henry's commentary. Oh, wow. Yeah.

7:09

The Bible. I have them in the other room. Yeah.

7:12

And so you just pick any book in the Bible and you're like, oh, I'm in Judges chapter

7:14

five or whatever. And you just go get Matthew

7:17

Henry's commentary and you open it up and

7:19

you get to see exactly what this person said

7:22

about that book, which is interesting.

7:25

Now what I think is interesting and what

7:27

you're doing is the modern format is

7:29

video, like the equivalent of a scroll

7:32

today or the equivalent of a book today is video.

7:35

I mean, podcasts are great, but ultimately

7:38

the authoritative, I don't know, the

7:40

thing you look for when you're trying to

7:42

solve a problem is the video. I don't know why. That's

7:44

just the way it is. You have an opportunity here

7:47

to build videos of,

7:49

I don't know, the Matthew Henry commentary

7:51

of the Bible, but it would be the Matthew Whitman commentary

7:53

of the Bible. And when

7:56

we turn into dust, that potentially

7:58

could live a few hundred years. years past us. So

8:02

you're trying to brute force, I don't

8:04

know if that's the way to say it, you're trying to optimize

8:07

a way to turn these

8:09

podcasts that you've set into a microphone

8:12

into video. Yes. And

8:15

you're trying to streamline that process and

8:17

people are employing modern artificial

8:19

intelligence techniques in order to, I'm

8:22

assuming, just write the words

8:24

that you're saying and

8:26

then convert those spoken

8:29

words into written words and then those written words

8:31

into images. Yes. That's what you're doing. That's

8:33

what I think is happening. And you're investing into this,

8:36

you're putting money into this. I've hired a lot

8:39

of people for an audition and

8:42

ultimately I haven't found the right partner

8:44

here. So I'm sort of back to square one,

8:46

but back to square one having learned what

8:48

isn't the answer. So I've gone to

8:51

different editing companies, I've hired individual

8:54

editors and all good

8:56

and talented people. I'm asking them for a weird project,

8:59

but I'm just giving them an audio file that's 10 minutes

9:02

long and saying, cover this and make

9:04

it educationally effective. Clearly,

9:07

some of the people I've hired just went

9:09

out and ran that through a transcription program,

9:13

then took the transcription and ran that through

9:15

some sort of AI web crawler, something

9:18

that would go and search websites and

9:21

then did an auto-populate of

9:23

whatever returns they got. And

9:25

they were just like,

9:27

here you go.

9:28

This is an edit of your thing. Other people

9:31

just clearly went and manually did the work

9:33

and tried to understand the material a little

9:35

bit more. Ultimately, nobody I worked with

9:38

thinks in these categories enough. It's just not

9:40

what they do. It's not their area of specialization and

9:42

it's fine. It's a weird niche. They

9:44

just haven't been able to add enough value to the

9:46

viewer for it to work. But the AI editing

9:49

has been really interesting because it's

9:51

just some word that I

9:53

say that the AI fixates

9:56

on. It's like, okay, well, here's a picture of that. So

9:58

I'm talking about... Abraham and

10:00

Isaac and there's a ram in the

10:03

bushes and it's a whole dramatic story

10:05

with life and death stakes And it's like

10:07

did somebody say Bush? Well,

10:09

he was president of the United States from 2001 till 2009 really during that time

10:16

He fought against the like wait, what what do we

10:18

do? Huh? What and I mean? It's

10:21

a little bit hyperbole, but kind of not it's

10:23

been that kind of thing where you can tell

10:26

that AI is involved because it'll Be pretty

10:28

close on a couple of cuts

10:30

like oh, yeah.

10:31

Well sure enough. That is that is Israel.

10:34

Mm-hmm Now it's modern

10:36

Israel. I'm I'm talking about the Old Testament

10:38

and you showed me a picture of Benjamin yet Netanyahu

10:41

and you know the Israeli flag

10:43

and maybe an Israeli fighter

10:45

jet So that doesn't really work for an Old

10:47

Testament survey that it would be awesome

10:50

if there were fighter jets in the Old Testament

10:52

Maybe Ezekiel Art

10:54

things running around So

10:57

you can just see what happens. It doesn't understand

10:59

the whole concept It doesn't understand the theme

11:02

or the motif or the feel of the video,

11:04

which I think is art at all So

11:06

it's just like you said word. This

11:09

is picture of word picture of word.

11:11

You said behold. There it is I serve

11:13

you I give you thing master and it's

11:15

clunky and bad. But the other thing that's interesting

11:18

is Okay, let me ask you this

11:20

when you edit what is your timing

11:22

philosophy Like you say something

11:24

and then you make your next cut. It's called see say

11:27

explain that please. Yeah, so see say is

11:30

Sometimes you want the it depends on if I want to

11:32

be ahead of the viewer or behind the viewer

11:34

or right on time so if I'm saying

11:36

a word that's difficult as azimuth

11:40

azimuth is a word that means

11:42

an angular deflection like He

11:45

rotated 15 degrees in azimuth. So you rotate

11:48

right and you point the gun 15 degrees to the right

11:50

azimuth Azimuth and elevation

11:52

those are the measurement angles that you would

11:54

take when you're you know targeting something So

11:57

if I were to say azimuth what I

11:59

would do is So

14:00

then you just show it. Yeah, the timing

14:02

of that matters. Yes, because You

14:05

don't think it matters, but it really really

14:08

does Say it again that a

14:10

bring to boy leaves his hand It's about that point

14:12

it would have been Abram Abram leaves his home

14:14

in or of the Chaldeans and makes his way west

14:16

Okay in that case makes his

14:19

way west. So there's an idea that's

14:21

gonna follow that I would not expect

14:23

the map to stay up on screen for another

14:25

two seconds It would be makes

14:27

his way west and then when the T

14:29

hits that I might give it three

14:32

or four frames and Then

14:35

boom screen change. Okay, and

14:37

what I would do there is I would

14:39

have a map of the ancient near east and I

14:41

have four of the Chaldeans over here in a little

14:43

picture of maybe Abram's face and Then

14:46

he makes his way west and he starts tracking

14:49

You know kind of you kind of bend up north to get around

14:51

the desert when you travel in that part of the world He

14:53

starts tracking west and then I would pull

14:55

in on the next part of the story until he arrives

14:58

Yeah, that's even better and

15:01

then maybe if I was covering the whole video I would go

15:03

to a picture of that place like

15:05

a nice big panoramic stock

15:07

photo That you just use the credits on

15:09

because I don't have access. I haven't been there

15:12

and You show this beautiful

15:14

picture. Maybe I even put the name of it up there.

15:16

This is where he went to and Visual

15:19

storytelling visual storytelling. Yeah, the next

15:21

thing here's the next thing I like the way you did that even better

15:23

because you he makes his way west So

15:26

in the same motion you're moving west then you start

15:28

pushing in the natural visual

15:30

question is where is he going? And

15:33

you didn't have to say it the image did it

15:35

for you. Mm-hmm. Yeah, I like that way

15:37

better. So The

15:39

AI can't do that Yeah,

15:44

there's nothing more to it man, it's just bad really

15:47

all of those satisfying timings that you talked But I

15:49

remember one time I edited a video where

15:52

we were flying a drone in front of this man

15:54

near me if I said Can I want to know? That

15:57

video is still available. Yeah patrons

15:59

who are I'm curious to see it. That's great.

16:02

Uh-huh, that happened. No, it wasn't that time at all. I

16:04

was flying a drone that didn't hit anyone in the face. It

16:06

was way up high. I was looking at the gargoyles

16:08

on top of the Catholic Cathedral. And

16:11

we got Salt Lake City in the background. It's a beautiful shot.

16:14

And there's a dove that flies through

16:16

the shot. No way. Behind

16:19

the bell tower, and then it comes out the other

16:21

side of the bell tower, and in an early

16:23

draft, I cut it before the dove

16:26

made it to the edge of the screen. Yeah, you can't do

16:28

that. Yeah, this was such a complex video.

16:30

There were so many things in it. And

16:32

I sent it to you. I was like, oh, take a look at this. I don't hear

16:34

anything. I don't hear anything. I don't

16:37

hear anything. Finally, I got a phone call. It's one in the morning.

16:40

Hey,

16:41

go in there right now and let me see the

16:43

end of that dove flying to the edge of the screen. Okay,

16:46

I'll call you back if I have more.

16:47

That's it. You didn't give me any other

16:49

feedback on anything. It was just,

16:52

I need to see the dove finish the flight. It was the

16:54

wrong cut. It was the wrong cut. You

16:56

were right. It did really well, and clearly it was because

16:58

we included the little... It was because of the dove. Yeah, yeah,

17:00

exactly. There are all these sensibilities that you have.

17:03

I share a lot of them. I mean, every editor

17:05

has their own taste,

17:06

but

17:07

there are all of these sensibilities that are kind of best

17:09

practice, and they feel satisfying.

17:12

Even if you don't edit video, your brain

17:14

knows when a thing is edited right, and it's

17:17

satisfying. Everything

17:19

I wanted to see, every time I had a next question,

17:21

whoever edited this and gave it to me, that's great.

17:24

Have a like. Have a nice comment even.

17:26

That was wonderful, but when it doesn't

17:28

happen, it's so disorienting.

17:31

It's frustrating. Frustrating. For

17:33

me, it's frustrating. It's like, why

17:35

did you do that? When I make

17:37

a music montage in Smarter Every Day

17:40

and I have slow motion footage. Gordon

17:42

McGlattery writes the music for Smarter Every Day.

17:44

His company's called A Shell in the Pit.

17:47

He writes gorgeous music. Super talented. Very

17:49

different. Sometimes it sounds like a video game.

17:52

Sometimes it sounds like acoustic guitar. He has

17:54

a large quiver

17:57

of arrows he can pull from at any one moment to

17:59

give you a certain And so when

18:01

I'm doing slow-mo, I make sure to hit

18:03

any changes on the beat I'm

18:06

thinking I'm thinking of one song he has

18:08

in particular. Can I play it for you real quick?

18:11

Yeah, sure

18:19

It is very pentatonic. Oh,

18:22

yeah. Yeah So

18:25

he's got music like so you slowly

18:27

ramp this up and then as

18:30

this music comes in more you're like, okay Well,

18:32

I'm gonna let it get more complicated visually and

18:35

then eventually Okay mechanical

18:37

thing happens now Yes,

18:40

right. Yes. This is a build montage Music,

18:44

you know what this makes me think of what

18:46

smarter every day the YouTube channel Yeah

18:49

In fact this the name of this song is

18:51

on the smarter everyday edit So he

18:53

he makes edits of the videos specifically

18:56

for smarter every day So this right here is

18:58

where it gets really complicated and we're gonna hit it

19:00

now Sorry, I waited

19:02

too long Wow,

19:08

it's like 8-bit Baroque. Yeah, he

19:10

does the order of the cosmos being

19:12

conveyed But

19:14

the intricacy on the deep great deep

19:16

hits of the music if I ever have to change

19:19

I would change and if there's a Right

19:21

here. We have to shift down Now

19:27

you get the idea so it's

19:29

important to hit on the music and each

19:31

different This is weird because

19:33

you're better at music to me each movement of

19:35

the music. I might change

19:37

a scene Go back to real-time

19:40

footage to show me tightening a ranch on the thing

19:42

that's about to do the thing so

19:44

I don't know. That's the kind of stuff I do. Yeah,

19:47

and it's great. You're a really good editor and George's

19:51

to George's great and you are too. Yeah,

19:53

my buddy Nate who helps me out with some of my church

19:55

visit video And it's he's

19:57

wonderful. I mean of the four of us. He's the only one who's

19:59

one em

19:59

Do shorts have

20:02

Emmys? He might. I don't know.

20:03

He's good. He's got quite the track record. Yeah,

20:06

he's good. Yeah. Nate's real good. Yeah,

20:08

so the not humanness of

20:11

the AI decision-making is

20:13

really evident on a few of these attempts.

20:16

And I think all of these attempts are a little bit human,

20:18

a little bit AI, but the cuts

20:20

just don't make sense. And so it's frustrating when

20:22

you watch somebody else's video that is on

20:25

YouTube just not do

20:27

the things it ought to do. If you just put that

20:30

there and put that there, it would make sense, but you didn't.

20:32

It's really frustrating when you're watching a draft

20:35

of your own material with your own voice.

20:37

And so then what's happened

20:39

in some of these rough edits that I've got back

20:42

is clearly the editors or

20:44

the editing AI software suites

20:46

that they're trying to use. They don't

20:49

recognize certain historical artifacts

20:51

or details and they

20:53

just take stab at it. And so

20:56

in some of these edits I'm getting these full

20:58

AI panels, AI written panels that

21:00

are just word salads, horrible

21:04

writing, atrocious, overly

21:06

busy. What do you mean? Just tortured

21:09

language. So for example, there

21:11

was, now just understand too, I've

21:14

sent three videos, one

21:16

that is a bread and butter video,

21:19

me sitting at my desk talking about history

21:21

stuff in the Bible. Another that is

21:23

a very typical video for me where I go and visit

21:25

a church and I ask somebody to tell me about

21:27

their understanding of religion and God

21:30

and Christianity and I ask questions and just listen

21:33

and marvel at their church building. And

21:35

then I sent out another video where it's

21:37

just my voice talking about the book of Genesis.

21:40

I'm comparing what I get back from different companies

21:43

on all of these edits. On one

21:45

of the edits I got back on this particular

21:47

church visit video. It's a church

21:50

that is all built

21:52

in the style artistically of the book of

21:54

Kells. Have you heard of the book of

21:56

Kells? It's something to do with

21:59

Ireland.

23:45

of

24:00

what worship should look like. It's too

24:02

busy. It just plastered over it? They

24:04

plastered over all of it. That's how- Wait.

24:07

That doesn't sound like you're being nice to people. That sounds

24:09

like you're- I don't know. Dude,

24:11

I wasn't there. It was World War

24:13

II. That's what they told you. That's what they told

24:15

me. Okay. So I'm going to choose to give them the benefit of the doubt.

24:18

Okay. That they were clearly doing so to

24:20

try to accommodate their Christian,

24:22

albeit different, Christian brothers and

24:25

sisters who were in a very difficult time. I

24:27

took it as a pretty neat story. And then in the 90s,

24:30

times had changed again. They do a restoration

24:33

on the church. And now all

24:35

of this, old St. Patrick's Church in downtown

24:38

Chicago, now all of the Book of Kelsusness

24:41

is on display. It's just beautiful. So I

24:43

had a few different editors take a run at this thing. And

24:46

one of the editors relied heavily

24:48

on AI for the writing. Not even

24:50

Google searches, I don't think. I'm just guessing.

24:53

You're not allowed to talk to the editors when you use these

24:55

companies. Oh, okay. Are they people

24:58

in the same country? Like are these people in America?

25:00

Uh-uh. No, I'm almost a hundred percent

25:02

sure they are not. Okay. I think

25:04

they are using AI

25:07

to look- like I think

25:09

they're entering something and saying, but now

25:11

frame that in a natural Midwestern American

25:13

style. Because you notice some of the AI

25:15

blemishes even in the written communication back

25:17

and forth between editor and myself. Really?

25:20

I think so. And I think these people are

25:22

native English speakers? Yeah,

25:26

maybe as a second language. Okay. They're

25:28

all obviously incredibly competent people,

25:30

but they're clearly working from overseas.

25:33

And that's what makes the economics of it probably

25:35

work. I don't know, you're not allowed to know what the economics

25:38

of it are or how it works at all. Maybe it's

25:40

a dubious problematic arrangement. I don't

25:42

know. What I do know is everybody I worked with is

25:45

really talented at all of these different companies, even

25:47

if they don't know the material. And why would I expect them to know

25:49

this material? It's very obscure. But

25:51

they would run into things clearly in

25:53

the editorial decision-making process and be

25:55

like, all right, AI, bail me out.

25:58

Write me a thing. about the

26:00

book of Kells. But here's the

26:02

thing, man. In the last year

26:05

of the evolution of AI and the integration

26:07

of AI, there are a lot of people who

26:09

I've read who are making compelling cases that

26:12

the writing side of it at least is getting worse.

26:15

Yes, I've heard this. What have you heard about

26:17

it? It just seems like in the early days,

26:19

it was learning off good stuff. I

26:22

don't know if this is it, but now the

26:24

things that are being input into the model are trash

26:27

and so you're getting trash out. So it's

26:29

almost like when you make, you remember back in the day, we

26:31

had Xerox machines, copiers, photocopiers?

26:34

Yeah, you make a photocopy of your butt? Yeah,

26:37

exactly. No, no. When you get a photocopy

26:40

of a photocopy, you'd be like, oh, it brought

26:42

over the blemishes from the blemishes and then,

26:45

and you could keep doing that over and over

26:48

and over and eventually the worksheet

26:50

your teacher would hand you be like, what is up with

26:52

it? I'm sorry, I can't read this number.

26:55

Yeah, have all the artifacts and flex and...

26:57

Yeah, I wonder if it's something like that. I

27:00

don't understand how it works well enough at all. I

27:02

mean, I'm tinkering with it as an end

27:04

user and I'm tinkering with it. Holy cow,

27:06

you just made me think something, dude. I'm sorry to interrupt you.

27:09

Go for it. That means we're gonna have to start

27:11

seeding. Like you're gonna be able to make

27:13

money with the seed for a

27:15

language model. Like, hey, here

27:18

is a, you got a good clean foundation

27:20

for a language model. You know what

27:22

I mean?

27:23

Like we're gonna be... A real transcode? Yeah, exactly.

27:26

That's interesting. So

27:28

the name of the game is going to be infect other people's

27:30

AI. Right. Scary.

27:33

And so even if we think of say a YouTube

27:35

channel's algorithm as being a

27:37

little bit AI-ish, I mean, it's machine learning. We've

27:40

talked before and we've looked at hard

27:42

empirical data to suggest

27:45

this video or this channel

27:47

is stuck. The AI is fixed. It doesn't

27:50

know what to do with this particular product.

27:52

And we've both talked with people who have arrived

27:55

at that place of being AI stuck or

27:57

algorithmically fixed, overfitted.

28:00

overfitted thank you and knew there was another term

28:02

and They go and

28:04

launch something else with the same talent the same material

28:06

They're still the same person and it works

28:09

the algorithm likes it Mmm, it goes so

28:11

we know that overfitting happens and

28:14

it looks like it's happening in Particularly

28:17

the language models the chatbots

28:19

over the course of the last year. It's been overused

28:22

with bad prompts And now it's

28:25

learning off of itself So

28:27

it's learning off of garbage it has written

28:29

if it is known echo chamber There

28:32

you go.

28:33

That's what I think is happening. Really?

28:36

I think you are reading stuff written by AI

28:38

so much more than you realize and

28:41

The more that happens the more that increases

28:43

the percentage of material that

28:46

the AI is Gleaning from

28:48

that is also AI It's

28:51

putting a second and third and fourth layer

28:54

between actual human minds writing

28:57

things and What we are modeling

28:59

from with our AI assistance One

29:09

being the least amount ten being the maximum

29:11

amount What level of candor do you

29:13

think makes sense for this little

29:15

spot where we talk about the fact that there isn't

29:17

an ad? on this episode was On

29:21

your scale was it? Yeah, I don't

29:23

know like from zero to one. That's not

29:25

in part in the scale What was

29:28

zero to one would be like we just

29:30

don't admit that there's no ad

29:32

on this episode Okay, and we just

29:34

hope nobody heard me just say that Oh Let's

29:38

go ten Yeah,

29:42

great. Yeah, there's no ad which means this one

29:45

is fully brought to you by The

29:47

handful of people who support this program to whom we are

29:49

super super grateful Yeah, patreon

29:52

is becoming increasingly more important

29:54

for notum questions, and I am also

29:57

grateful So I think

29:59

it's time to do something for the patrons again. Well,

30:02

I think so too. And I've been cooking

30:05

up ideas because it's not just notable questions.

30:07

I mean, the whole economy of everything

30:09

is making it such that Patreon

30:11

is kind of the way, it's the most important

30:14

thing for pretty much everybody who makes

30:16

stuff on the internet.

30:17

And so, yeah, I've been thinking the same stuff that...

30:20

The reason...

30:21

I'm definitely going to start this again. I absolutely hate

30:23

talking about things like this, dude. It

30:27

cripples me. So easy and so natural

30:29

for you and it just cripples me. I don't... Well,

30:32

let's just... I don't know. Let's just keep going.

30:34

Oh my goodness. Come on. I mean,

30:37

it's just the way it is, right? So I think

30:39

the reasons behind all this could

30:41

fill a whole episode. I think it

30:43

has a lot to do with inflation, which has a lot

30:45

to do with printing money, which

30:48

has a lot to do with all the things. Yes.

30:51

And so that affects marketing budgets. And

30:54

yeah, that's just where it's at.

30:57

Yeah, and it's

30:59

okay also. Yeah,

31:02

absolutely. It's okay that economies

31:04

ebb and flow and things like that. This is part of

31:06

reality, but the most

31:09

meaningful bedrock thing in terms

31:11

of how something like this happens is the

31:13

people who like it the most kicking in on it. That's

31:16

the way it works. Yeah, I also want to do an

31:18

episode on the Medici family. I

31:20

think that'd be cool. Whoa. Yeah.

31:23

Not now. I mean, that's a hard pivot

31:25

right there. Well, you just want to run it right now

31:27

and just keep going. No. It could

31:29

just be like a little bonus 90-minute episode

31:32

inside of this other episode. No, I

31:34

don't. I don't. But the fact that they... Why

31:36

do you want to talk about the Medicis? What's happening to you? Well,

31:39

the fact that they were patrons of

31:41

the arts and stuff like that, I think it's very interesting.

31:43

Oh. Yeah. Okay,

31:45

I see what you did. I'm grateful for the patrons.

31:48

It makes no damn questions possible. It pays

31:50

us. It pays Tina. And

31:52

I'm grateful. And I think we should send stickers out.

31:55

I would like to send stickers out. It's

31:57

time. I really like what you're doing on...

33:59

or

34:01

there's gotta be a good idea for something out there.

34:04

That's a really interesting idea. We could

34:06

do several things. We've talked about all kinds of stuff this

34:08

year. Yeah, it needs to

34:10

be this year.

34:12

And so then we send it out, and it's like, thanks

34:14

for supporting this last year

34:16

when we talked about the thing. And if you weren't

34:18

supporting this last year, then you don't get that sticker,

34:21

but you can get next year's sticker. Yep, but

34:23

it's an evergreen sticker. Think about something we talk about next

34:25

year. Like, it's a sticker that works all the time.

34:27

We don't put the year on it, but it's only like

34:29

a limited run. That's how we do it on Smarter

34:32

Every Day. I think it'd be awesome. Let's do it. Yeah,

34:34

okay, I like that a lot. And yeah, it is

34:36

time to send something out. I mean, that's always kind of been the model

34:39

is just hear something fun from time

34:41

to time and thank you, and it's a goodwill engine.

34:43

And yeah,

34:44

more than ever, that's what Patreon

34:47

is.

34:47

It's a goodwill engine, thick and thin.

34:49

It's really appreciated. Yeah, it is. So

34:52

if you wanna support, you can go to patreon.com

34:54

slash no dumb questions. And

34:56

you will, like, if you're a supporter at the

34:58

end of the year, you will get the sticker.

35:01

Let's do it that way, right? I like that a lot. Yep,

35:03

those are good ground rules, but we should probably go make

35:05

that sticker. But you gotta tell us what the

35:07

sticker needs to be first. That's, so it's not our fault.

35:10

The ball's in your court, y'all. You gotta tell

35:12

us what the sticker's supposed to be. Oh, we're off the hook. Look

35:14

at that. Thank you for your support. Obviously,

35:16

it means a lot, especially right now.

35:18

Thank you. Yeah, agreed. Is

35:21

that good? I think it's good. All

35:24

right. Do

35:26

we close file? That's close file. Closing three,

35:29

two, one, close.

35:35

I wonder what the first word to enter

35:37

the English language that was not

35:39

invented by a human mind will be.

35:42

Scrotulate.

35:43

Just a guess.

35:48

Excellent choice of prefixes,

35:50

my friend. Scrotulate.

35:55

Yeah, that was your brain. That was definitely your brain.

35:57

Oh, you got me again. Definitely your brain.

36:00

in one episode.

36:02

No, I think it's gonna happen.

36:05

I think there will be something

36:08

that authentically happens in our culture

36:10

that did not have a human brain origin.

36:12

It was a derivative product. I just

36:14

don't know when that's gonna happen, but it'll happen. Interesting.

36:18

Well, I can tell you this. What

36:21

the AI churned out on my videos

36:24

in terms of the written sections was

36:26

an absolute dumpster fire. Well,

36:28

you also are dealing with matters

36:31

of religion, faith, Bible.

36:34

So how did it do on the doctrine? Terribly.

36:37

What was the most heretical thing it did?

36:40

It's not even so much that it's heretical. It's just

36:42

a mishmash of a whole

36:44

bunch of incompatible ideas. And

36:47

so, you know, if you don't think hard about

36:49

it, I just here's

36:51

a Christian religious sounding phrase

36:53

and here's a Muslim one. I mean, they

36:55

both more or less sort of acknowledge a God

36:58

thing and there's some Mormon assumptions

37:00

over here and here's a phrase from

37:02

Hinduism and really there might be truth in all

37:05

of those particular phrases. There might

37:07

be beauty in all of those phrases.

37:10

But if you really look carefully at all

37:12

of the words and meanings behind all of those phrases,

37:14

you realize there are some mutually incompatible

37:17

belief systems being crammed into one thing

37:19

here and it just doesn't work. So

37:22

it would be like if you

37:24

saw something from your field that was just

37:26

a bunch of engineering drivel,

37:29

just a bunch of engineering philosophy I do see this. I've

37:31

flung together into a paragraph. You know where

37:33

I see this? Tell me. I see this on videos,

37:36

not just modern ones that have been

37:38

created by AI. I've seen footage

37:41

on the internet that I filmed

37:43

with my hands that

37:45

is being talked about by this other channel.

37:48

Yes. And I'm listening to them articulate

37:52

and speculate about what's going on and

37:54

say it with authority and

37:57

I was physically there when it was filmed and

37:59

I know

37:59

know it's horsepucky. And

38:01

I'm sitting there watching and I'm like, this

38:04

is a lie. And I look down at the view count

38:07

and I say, holy moly. And I look

38:09

at the channel and I'll be like, this channel has

38:11

a lot of viewers and they're propagating

38:14

misinformation right now. And

38:16

they're doing it as if it's true. And I've

38:18

thought to myself, you're an unethical

38:20

person. What

38:22

a gentle thing to think, given the circumstances.

38:25

Yeah. Yeah. I've seen that. I've seen other things

38:28

where it's a field that I know a lot about

38:30

and I've seen something spoken

38:33

with authority, even with an animation.

38:35

And it was not correct.

38:37

And I look at the subscriber count and

38:39

I look at the views and I'm like, wow,

38:43

this is just the thing I know about. What

38:45

are you doing with the things I don't know about?

38:49

Yeah. In my field, a lot of times when people are like,

38:51

hey, that was wrong. What they mean is I don't

38:54

hold that theological opinion on that

38:56

issue. But it's not one or zero

38:58

wrong. It just isn't. I mean, that's

39:00

just like, that's your read, that's your opinion.

39:03

But where it can be wrong is

39:05

if let's say we are talking about

39:07

a Presbyterian set of assumptions.

39:09

That's a certain brand of historic

39:12

Christianity, Presbyterians.

39:14

And they have a certain way of governing themselves.

39:17

If we went and we were talking

39:19

about what the Presbyterians think and

39:22

we talked about a Catholic form of

39:24

governance, well, then that would be wrong

39:26

because that isn't what the Presbyterians think

39:29

and it's incompatible with all of their other assumptions

39:32

about how church ought to be. It

39:34

doesn't mean it's universally wrong

39:36

and like the grand ethical, biblical,

39:39

what mind of God kind of sense.

39:42

It's just you're trying to take something that's incompatible

39:45

with a system and pretend

39:47

like it fits in that system. Well, it doesn't. It's wrong

39:49

when you put it there.

39:51

That can't work.

39:52

It's taxonomically wrong.

39:55

That makes sense.

39:56

Taxonomically, what does that mean?

39:58

Like in terms of.

39:59

organizing and ordering and categorizing

40:02

things. Okay, it doesn't fit with that, I

40:04

see, category.

40:06

That's the kind of stuff that the AI kept getting wrong

40:08

on these edits. It would say things that

40:10

are Christian from a different expression

40:12

of Christianity, or maybe we're

40:15

Muslim, but sort of sound Christian, Muslim

40:18

in their assumptions, and try to make

40:20

it fit with something Catholic. So the same

40:23

feeling I get when I read a phishing attempt

40:25

email, where it's like, you're

40:27

saying all the right words, but

40:30

you're, there's something off. I

40:32

don't know what this is. Did it even get that close, or was

40:35

it? I don't think it got quite that close. Really? There's

40:37

one moment where I'm looking at

40:40

the stained glass ceiling, depicting

40:42

the four authors of the Book of Kells.

40:45

They're symbols from Matthew, Mark, Luke, and John, the

40:47

four biblical authors of the stories of Jesus,

40:49

the Gospels, and it depicts their four symbols,

40:52

a beautiful stained glass, just magnificent.

40:55

And I'm looking at it, and the AI

40:57

editing cuts to a panel that

40:59

describes, I have no idea

41:02

what. It doesn't make any sense. It shows

41:04

a picture of something. We were just looking at it. We

41:06

don't need to cut to a picture. It shows

41:08

a picture of something that makes no sense from another

41:10

document that isn't in any way related

41:13

to the Book of Kells. And then it goes on to

41:15

describe what's happening in this document

41:17

with just tortured language, no

41:19

context. No, here's an image

41:22

of, it just Kool-Aid man's

41:24

through the wall. Here's an eagle. The eagle has

41:26

a hook. There's a man with a beard pulling

41:29

on the hook and a... Wait, what? What the heck is happening?

41:31

Oh, so they put this weird Celtic image

41:33

through an AI interpretive.

41:36

I don't know. Can you put an image in AI

41:38

and it will tell you what it is? Is that a thing you can do? I

41:41

don't know. Okay. What is going on with

41:43

the things I hired. I have no idea. I

41:45

wanted to see what would happen if I gave it this

41:47

input. You showed me a couple of things, and it looked like some...

41:51

I don't know. The short thing you showed me,

41:54

some of it looked okay. Some of it looked

41:56

like some kind of caterpillar Alice

41:58

in Wonderland drug. trip or

42:00

something. Okay that's fair. Would I imagine drugs to

42:02

feel like? Okay, alright. I don't know. So

42:06

what are you gonna do with this? Oh, did

42:08

you waste your money? No, certainly

42:10

not. I'm learning ways not to do it. So

42:13

I'm starting at the lowest hanging fruit.

42:15

Okay, well what happens if we

42:18

just hire an affordable editing

42:20

company that you can just Google? Editing company,

42:22

edit my videos. Can they do this? No. They

42:25

can't come close to doing it. That won't work. Alright,

42:28

well what happens if I hire

42:30

some different individuals? They can't because of money? Why

42:32

can't an editing company edit the videos? Because

42:35

they don't know the material. I see. It's

42:38

not fair to ask somebody for whatever

42:41

amount of money they charge to

42:43

go and learn all of the intricate

42:45

details of the Desert Wandering and the Book

42:47

of Deuteronomy and the beginning

42:50

of the conquest of Canaan at

42:52

the beginning of the Book of Judges to make a 10-minute

42:54

video. That's not fair. They're just not

42:56

conversant. So the amount of time it takes them

42:59

to vet an image or a

43:01

painting or, okay, what archaeological

43:04

site is this? Is it tell this or

43:07

is it tell that? And so

43:10

the mistakes that have been made

43:12

in these edits, they're innocent.

43:14

They just demonstrate

43:16

a lack of knowledge of the material. What's the process?

43:19

Do you say, hey, that

43:21

right there is a, I don't know, that's

43:23

a ball of fire. I don't even know what I'm

43:25

talking about right now. Well, I can fill in your

43:27

question. Okay. So for example, I

43:29

worked on a video about Susa, one

43:32

that ultimately I ended up kind of landing

43:34

the plane on and getting to where it needed to be. But it

43:36

was fun. That's a city, right? I had a lot of people take a run at

43:38

this ancient Persian capital of Susa.

43:41

I'm sorry I haven't watched the video yet. You're just

43:43

fine. Yeah, okay. Good.

43:45

And yeah,

43:47

I got different things back stylistically from different

43:49

editors I put it in front of and I've

43:51

already published the video and it turned out I

43:53

think fine, it's informative, it works.

43:56

But

43:57

the returns that were happening from search

43:59

engines and from the AI

44:02

were just anachronistic and incorrect. So for example,

44:04

I make a remark at one point where I say Xerxes

44:08

sitting upon his throne in the

44:11

palace, Susa thought to himself,

44:14

and the image that comes up is

44:16

like a clipart of Xerxes

44:19

sitting upon a portable throne overlooking

44:21

the Battle of Salamis in Greece.

44:24

Close, close. When he's

44:27

thinking, did like the statue of Rodin

44:29

come up, the thinker? That's the

44:31

kind of stuff that you get. Yeah. That's the kind

44:33

of stuff it does. It's every low-hanging

44:36

visual joke, and not even joke,

44:39

just association. So

44:41

what's the feedback loop? Okay, clearly it's wrong.

44:44

What do you do to correct it? Oh yeah. So anyways, I've

44:46

been using a program called Frame.io,

44:49

Frame.io, and the

44:51

interface is a lot like Adobe Premiere, which is what

44:53

you and I edit in. And

44:56

I just go through, and everywhere

44:58

that something's wrong, I have

45:00

to type a note. I can't go through

45:02

it verbally with these editors because

45:04

you're not allowed to talk with the editors ever.

45:07

No email, no Zoom calls. You

45:09

can't know their name. You just get

45:11

a first name that may or may not be true, a

45:14

picture that may or may not be them, and

45:16

everything is fed through a communication

45:19

channel with the American-based

45:21

project manager. Clearly,

45:24

they don't want their talent getting poached. I

45:26

don't blame them for that, but

45:28

it makes getting good edits super

45:31

hard because

45:33

you can't actually communicate.

45:35

So then I would mark the error

45:37

on the picture of Xerxes and be like, hey, this is great.

45:40

This is definitely King Xerxes, but

45:42

he's in Greece here, and

45:44

that's the wrong part of the planet. You know,

45:46

we need a different shot. And then

45:48

ultimately, I'd go into my archives

45:51

because I have an ever-growing database of pictures

45:53

I've taken in my travels, and

45:55

I would grab the really cool

45:58

relief of Xerxes from...

45:59

another old Persian capital, Persepolis,

46:35

and it's important to know, I mean you're

46:37

doing the Thomas Edison thing, I

46:39

made a thousand things that didn't work, you'll

46:42

have another tool in your toolbox, and so maybe

46:44

in the future, you're like, hey,

46:46

I have the simple thing, I just

46:48

need to cover the screen with XYZ, and then maybe it'll

46:50

give you a good first cut, and then

46:53

you can go in there, and it'll give

46:55

you ideas or something like that. Do you mind

46:57

me asking, how much does this cost for

46:59

a single like first pass edit? Do

47:02

you pay for a first pass of an edit? I've paid

47:05

a variety of amounts because I wanted to figure out... What

47:08

does what by me? And it's

47:10

been painful. I intentionally set

47:12

aside a budget for this, the budget hurts,

47:14

but I'm trying to solve a

47:18

bigger equation than just get a couple videos published.

47:20

I've got an idea,

47:22

I have a mission, I have a plan for what I want

47:24

to accomplish with my YouTube channel moving forward, and

47:27

that was kind of a question mark for a while because

47:30

the podcast sort of took over everything

47:32

I was doing, and that's great. I

47:35

mean, it's great, it's so much fun. I'm glad people

47:37

like it, I like making it, but now this YouTube

47:39

channel, finally, I'm like,

47:41

all right, I know what I'm doing, I know

47:43

what needs to happen next, but I want to build

47:45

out a whole system for how I do it before I tear

47:47

off in this direction. So I set money aside, probably

47:50

the least I paid for

47:52

an edit was $300. Wow.

47:55

Which hurt? Yes, $300 is a lot.

47:59

of money.

48:01

But in terms of sitting down and editing,

48:03

that's a lot of hours. Right.

48:07

Unless you super know the material, in which case

48:09

it's, oh, well, the right picture here is this, the

48:11

right picture here is that. But nobody knows that

48:13

material. Nobody studies that. The

48:16

most I paid for an edit was $1,500. You're

48:19

going out into unknown territory and you're

48:22

actually taking stock of the situation.

48:25

This was a huge investment.

48:27

Yeah. And I've been able to publish

48:29

almost nothing. Yeah.

48:32

Yeah. And next to what I have published,

48:34

I had to do the work on ultimately. Interesting.

48:37

I mean, do you get project files

48:39

and they still have not provided

48:41

me with those? I have asked. I haven't been

48:43

given those. So it makes it really hard to go and fix

48:45

their last few mistakes. Wow.

48:48

It's just a bad workflow. What that kind of stuff

48:50

is for is lowbrow, simplistic,

48:55

faceless, automated, crass.

48:58

Let's just try to get some money out of YouTube,

49:01

audience building channels where you make

49:03

a YouTube farm. For bot farms. 15

49:06

channels. Amazing things from

49:08

the internet today. This amazing

49:10

thing happened on the internet today. Yeah.

49:12

This man thought he could catch the cat, but

49:14

he couldn't. This man had to unclog

49:17

the drain. He did. Interesting.

49:19

That's reassuring actually. Like

49:22

this actually makes me feel good because

49:25

it makes, not that it's

49:27

hard to do what we do, but it makes

49:29

me feel good that for example,

49:32

my friends who might be older and watching

49:34

more things when they swipe on their phone, they

49:37

won't be tricked. It's harder to make

49:40

inauthentic material than I thought. Yes.

49:43

I thought you'd be able to plug

49:45

in words to a major politician's

49:47

mouth and have him or her say

49:49

something crazy and have it passes.

49:52

I mean, I know this can be done. It can be done.

49:55

I thought it would be far easier than this

49:57

or like narrating.

49:59

something. So it's

50:01

reassuring that it's not like pop, pop,

50:03

fizz, fizz. It's far more

50:05

hands-on. This is where I'd say it's at

50:07

right now. This isn't a very evergreen

50:09

observation because we know it'll be different in six months.

50:12

But right now, I think AI

50:14

language modeling is getting really bad, really

50:17

predictable, and people can smell

50:19

it. It feels sterile and

50:22

trite and safe and

50:25

institutional. Even when it tries

50:27

to sound human and friendly and colloquial, it

50:29

still sounds that – it's just a

50:31

stink on AI writing. I think

50:33

it's bad, and I wouldn't be too worried about

50:36

that replacing anybody's jobs right now. It makes the value

50:38

of authenticity go up. It does. I

50:41

think that AI video looks

50:43

like total garbage right now. There's

50:46

a show that came out a while back

50:48

called Secret Invasion, another one of these Disney

50:51

Plus utter flops, just

50:53

total failure. I think that show went so

50:55

badly, they're going to reevaluate their whole Star

50:58

Wars streaming. I mean, it

51:00

was awful. But their

51:02

opening was entirely generated by AI, and

51:05

it's horrid. It feels

51:08

completely inhuman and

51:10

like a mockery of a human-making

51:13

art. It's

51:14

bad.

51:15

However, I will say that

51:18

AI stable visual art,

51:20

it's really getting somewhere. What do you mean, like a single

51:23

image? Yeah. Okay. Yeah. It's

51:25

getting somewhere. It's the tells

51:27

and the things that used to give it away. As

51:30

we get into these new versions of Mid-Journey

51:32

in particular is what everybody's using. It's

51:35

getting good. I mean, you can

51:37

go in and specifically

51:40

circle and edit things you didn't like

51:42

in-app now. So if everything's

51:45

perfect – Do you pay for Mid-Journey? I've never used it.

51:47

I do. Yeah. Yeah. It's 80 bucks, 80

51:50

bucks a month. Wow. So I'm not going to pay for it forever.

51:53

But I'm using it really hard right now. I'm

51:55

trying to build out something that I'm going to use in a whole

51:58

series of videos that are coming up. There's

52:00

just a few pieces of specific art,

52:03

avatars, characters that I'm trying to build.

52:05

I'm really deep into the mid-journey world, learning

52:08

how to seed it with images, so that

52:10

rather than just randomly generating characters

52:13

that don't look the same, I want the same character from

52:15

image to image to image. I'm using

52:18

Adobe's generative AI

52:20

to select that, separate it from the background,

52:23

make any final augmentations. I'm

52:26

using AI a lot right now for image

52:29

art and photo images,

52:32

and I'm even figuring out how to use AI

52:34

tools to improve garbage

52:37

old photos of Nineveh before ISIS

52:39

destroyed it. The pictures just aren't any good

52:41

that we have of Nineveh because people took pictures

52:44

on potatoes of this thing when it emerged

52:46

from the sand, and then ISIS came

52:48

along. It never got photographed well,

52:51

even though it was out of the dirt during the age

52:53

of photography. Well, there

52:55

are ways to recapture that and reestablish

52:58

it and make it look like people took great photos

53:00

of it. I've been working on things like that

53:02

for future videos as well. I

53:05

think that's trending in a really impressive

53:07

direction in terms of what it could be. That also bothers

53:09

me about the SR-71 Blackbird. You

53:12

know the Blackbird, the coolest airplane ever? Yeah, the Cobra

53:15

Knight Raven. The

53:17

thing about the Blackbird is we had

53:19

this period where we went from film to

53:22

digital, and the first few years

53:24

of digital were just trash. Hot

53:27

trash. DV, progressive. I

53:30

mean, what was the scan where

53:32

it interlaced? That was it. Interlaced

53:34

videos like, why are we doing this? A

53:37

lot of that was when the SR-71

53:39

was flying, and so we didn't have really,

53:41

really, really good stuff from there.

53:44

Some of the best stuff is old film photos from

53:46

there. So anyway, I'm sorry. I just wanted

53:48

to lament that for a second. No, I mean, you're

53:50

on the same page. We went right to a perfectly logical

53:53

place. In your field,

53:54

that's something that just, because of the unfolding

53:56

of history, we just don't have a good look at.

53:59

Same thing for me.

54:00

with Nineveh for example.

54:03

So I think there's a ton of potential in all

54:05

of this. It's just, I

54:08

needed this by the way. Okay. I

54:10

needed to understand because the

54:12

last time we talked about AI I was like, it's coming

54:15

for us. I hate this. But

54:18

this is actually encouraging. Okay.

54:20

It's encouraging that people can sniff it out.

54:23

It's encouraging. So got

54:25

it. Still photos, need to be careful

54:28

what I'm looking at. However, the

54:30

video stuff and the writing stuff, that's

54:33

really, really good because it

54:35

lets me know that people aren't

54:37

going to be fooled quite to the degree that

54:39

I thought. And so this is encouraging. And I appreciate

54:42

the little, I don't know what you want to call this, this trek

54:44

into the unknown that you've done at your

54:46

great expense with money to figure

54:49

this out because it's, yeah, I

54:51

feel really good about it. So thanks. I'm

54:54

feeling better about where it's at too. I like

54:56

that the writing is failing. I

54:58

like that the visual images are getting better

55:01

and more controllable because I'm

55:03

seeing how it will advance knowledge

55:05

instead of destroy the concept of truth

55:08

and knowledge. I'm excited about where these iterations

55:10

are headed. And I'm also excited about

55:12

what it means for creativity and what

55:15

people do. I think there's going

55:17

to be more interest than ever in actual

55:19

human effort that goes into making

55:22

something beautiful, composing music,

55:24

trying to understand a thing and

55:27

understand it so well that you can spread it on the mental

55:29

desk in front of you and draw connections

55:31

between disparate things to find

55:34

meaning. Knowledge is great. Everybody

55:36

loves it. But meaning is

55:39

what we seem to be wired for. What does

55:41

it mean? Why is it like this? What

55:43

is the underlying truth of these physical

55:46

realities, these metaphysical

55:48

realities that we encounter in life? And

55:51

that's where the human thing happens. And

55:53

I think there's going to be more hunger than that than ever

55:55

as cheap AI things do

55:57

the busy work of creativity. those

56:00

who can still spread a series of disparate

56:03

things out on the table and make the connections. I think

56:06

there's job security in that and I think there's beauty

56:09

in that. When's the last time you watched

56:11

Inception, the movie?

56:14

Three weeks ago. Really? That's about how

56:16

long it's been since I watched it. Yeah, we watched it ahead of

56:19

Oppenheimer. So you know the

56:21

point in the movie where they had a totem and

56:24

the totem was the thing that would tell you if you were

56:26

in the real world or in a dream. Yes. It's

56:28

like AI doesn't have

56:30

the ability to cross that boundary and that

56:32

threshold. There's always something there

56:36

that lets us know that it's authentically

56:38

human and that thing that you're talking about, that metaphysical

56:41

place that the human mind occupies, I don't know,

56:43

I said a big word, I don't even know what metaphysical means, but

56:46

that place that the human mind occupies, it's

56:48

like AI hasn't

56:50

figured out how to cross that completely

56:53

yet and that makes me feel good. Yeah,

56:56

I think it takes a soul. I don't think it's getting there.

56:59

I think the greater fear for

57:01

me is not that it will cross

57:04

the line and become indistinguishable from humans.

57:07

It is that the majority of humanity will

57:09

regress back the other way across

57:12

the line and therefore become indistinguishable

57:14

from AI. So

57:17

more Brave New World, less Orwell,

57:20

or it's the book Amusing

57:22

Ourselves to Death, Neil Postman. But

57:25

that's another conversation. It is.

57:28

Bottom line on this, I eliminated

57:31

a whole bunch of options for how I want to

57:33

do this next big series of videos on

57:35

my channel and now

57:37

I'm going deeper down the well with different

57:40

strategies for working with humans.

57:42

Is it more important to have a great

57:44

editor who I can teach things about ancient

57:46

history or is it more important to have a great

57:49

ancient historian who I can teach things about

57:51

editing? It's kind of the Armageddon question.

57:53

Do you want the astronaut or do you want the explosive

57:55

minor expert to deal with the asteroid?

58:00

I love it. Thanks for the update.

58:30

Thank you.