Secret Tech Club EP03: Copycats

00:00:07
Wow! Kang, you’re really talented!

00:00:10
Yeah, I used to be like,
best in school at drawing

00:00:14
Didn’t you say you knew a place
with the best coffee in town?

00:00:22
Maybe we could go on Saturday
or even a Sunday drink?

00:00:26
[crazy person laugh]

00:00:32
Hey, Niyat! Come over here!

00:00:35
So, what you think?

00:00:37
Impressive?

00:00:39
I used to doodle a lot as a kid

00:00:41
You know, I was the best in...

00:00:43
But do you know, generative AI can do it much better

00:00:46
Yeah, that’s … What?

00:00:48
Yeah, generative AI
learns from thousands of cat images

00:00:51
and then creates brand new ones

00:00:53
Ones that no one has seen before

00:00:54
Yeah, machine learning is clearly getting better

00:00:57
but as a human artist…

00:00:58
That’s not the same

00:01:00
Machine learning just looks at pictures
and say, “yeah, that’s a cat”

00:01:04
But generative AI looks at pictures and then
creates its own version from scratch

00:01:09
Isn’t that like stealing someone's homework?

00:01:13
Not really, it’s more like remixing

00:01:15
It learns the patterns, shapes,
colors, pictures

00:01:18
and then creates something unique
by combining them like an artist

00:01:22
who’s seen every cat in existence
and can imagine new ones

00:01:26
I’m doing art, though

00:01:28
My creations go far beyond photography

00:01:31
The generated images depends
on the training data

00:01:34
Feed the AI realistic cats,
and you will get lifelike images

00:01:38
Feed it with cartoons,
and you will get Disney cats

00:01:42
I don‘t think a whiteboard
doodle would pose much of a challenge

00:01:45
Don’t you think?

00:01:48
The more you know

00:01:54
I‘m also good at tech, you know

00:02:17
[Start]

00:02:18
Welcome to the Secret Tech Club

00:02:20
A video series where we
make tech fun and approachable

00:02:24
I’m Kang, and I’m joined by
my colleagues Niyat and Alexander

00:02:29
Today we will be diving into Generative AI

00:02:33
and how it  blows open the
door to creativity

00:02:37
and how it empowers
the next generation of artists

00:02:40
And I realize that that sounds

00:02:43
a lot more positive than you’d expect,
coming out from my mouth

00:02:47
after being thoroughly schooled by Niyat

00:02:50
in our sketch

00:02:54
How did it feel, Niyat, to crush me?

00:02:58
It was actually very funny

00:03:00
I can not remember
when I was laughing so much at work

00:03:05
Maybe on the other sketch
we had about Java

00:03:08
But yeah, I think it was a very nice

00:03:11
Alexander, maybe you don’t know, but…

00:03:13
when I showed her the storyboard and script,
and said “This is what I want to do”

00:03:17
She was a bit hesitant, and said
“I don’t know if my partner will allow this!”

00:03:22
I’m sure he understands!

00:03:23
He will understand
when he sees the final results

00:03:26
Because, in order to crush my ego,

00:03:29
she also had to crush my dreams

00:03:32
so that I would fall

00:03:33
just so that we could convey the
greatness of generative AI

00:03:37
Yeah

00:03:39
Of course I was thinking how
it would be perceived by other people

00:03:42
Like, not everyone
will understand it, maybe

00:03:46
It’s just good fun, okay?

00:03:48
In in reality, Niyat really
enjoyed my whiteboard doodle

00:03:51
She said “Wow, Kang”, for real

00:03:55
And Alexander,

00:03:55
you capitalized on the drawing

00:03:58
Yeah

00:03:58
I heard you laughing all day,
recording this sketch

00:04:02
and then I had to go in
and see what you actually did

00:04:05
What I saw was that cat

00:04:07
Yeah, we’re going to talk about that later

00:04:09
But before we look at
what you have to show today

00:04:12
I want to throw out a question
to keep in the back of our minds

00:04:17
“What happens when
anyone can create anything

00:04:21
without years of training?”

00:04:24
I don’t know if you have that feeling

00:04:27
When I first saw generative AI

00:04:29
and how it could
write an app in minutes

00:04:33
I was just thinking about

00:04:35
How I had spent
my whole life programming

00:04:38
All those summers, in the man cave

00:04:41
Getting good at it

00:04:42
And then

00:04:42
Now anyone can get access

00:04:44
that powerful feeling
to create, without an effort

00:04:48
Yeah

00:04:49
It’s like any art form,
like music, drawing, writing

00:04:51
They get the same treatment

00:04:53
It used to require talent
and hard work

00:04:55
You had to work so hard
just to get to a passable level

00:04:59
And now that is the baseline
for what everyone can do

00:05:02
Yeah

00:05:03
Talking about content,
you already see

00:05:05
If you’re going to LinkedIn
today, or recently

00:05:08
When anyone can create

00:05:09
It’s going to be a lot of content,

00:05:12
a lot of website applications,
that’s going to be created

00:05:16
And sometimes it’s just for fun

00:05:18
It’s an action hero of yourself!

00:05:20
I think, that’s the first step

00:05:22
Everyone gets to test it

00:05:24
And what’s important
is to find the real use cases

00:05:27
for organizations,
and innovation, and creativity

00:05:31
You two work with content, right?

00:05:33
I think this is a great setup

00:05:35
Because I feel
like I’m cheating if I use AI

00:05:37
to do the things I worked hard on

00:05:40
But at the same time
it makes me feel

00:05:42
like I have to keep up with you, because

00:05:43
you, for example, Alexander

00:05:44
You use all these new AI tools,
to generate images, videos, and so on

00:05:48
And I haven’t even touched it

00:05:50
And I like “Ah, you can do that?!”

00:05:52
And it’s getting better
and better by the week

00:05:54
Yeah

00:05:55
I feel stressed

00:05:57
Yes

00:05:57
You should

00:05:58
You should?

00:06:02
How do you feel?
Do you feel stressed?

00:06:05
Quite a bit,
but I also think that here at Zooma

00:06:08
we started really early
testing different kinds of tools

00:06:11
I remember years ago
when we gathered some Zoomers

00:06:15
and we just went through the tools,
and tested them together

00:06:18
And then talked about it
and yeah, I think it will

00:06:23
maybe not replace us,
but we can use it as a sure moment

00:06:24
maybe not replace us,
but we can use it as a shoulder mate

00:06:25
And I think it’s very important
to have also critical thinking here

00:06:30
like we know that
it can give you false information

00:06:33
So that’s why it’s always important
to review everything before you publish

00:06:38
Alexander,
could you tell us a little bit more

00:06:40
about your cat experiment
and some learnings, maybe?

00:06:44
Yeah

00:06:44
So I’ve written a blog post about it

00:06:48
It’s on The Onlinification Hub

00:06:51
As you saw in the intro of this podcast

00:06:54
Kang, he drew
this cat on the whiteboard

00:06:58
I took an image of that

00:07:00
Told ChatGPT to convert
the figure on the whiteboard

00:07:03
into a clear illustration

00:07:05
What we got out was this one

00:07:08
Then, I asked it to illustrate
the lounging cat on a beach chair

00:07:12
sipping a coconut drink

00:07:15
And you see it maintains
the same facial expression, and

00:07:20
here it’s a bit more happier, yeah

00:07:22
Who wouldn’t be in that situation

00:07:25
And next thing to do was
to create a colorful comic magazine cover

00:07:30
featuring this character

00:07:33
It’s a bold vintage style comic

00:07:36
And at this point,
I was pretty proud of what I did

00:07:39
so I sent it to Kang

00:07:41
And what did you think, Kang?

00:07:43
I was very impressed

00:07:45
One thing that I realized, I think

00:07:47
after it was posted

00:07:49
I didn’t notice
that the first sketch had two tails

00:07:54
Did you see that?

00:07:56
No, you told me that

00:07:57
It probably mixed up one leg

00:08:00
Yeah, probably

00:08:01
This comic strip,

00:08:03
you’ve probably seen
a lot of those on LinkedIn

00:08:05
It’s very easy to
generate in ChatGPT, for example

00:08:09
I wanted to push it to
a more photorealistic image as well

00:08:13
You asked about my learnings, Kang

00:08:15
And this is the first learning

00:08:17
You don’t go from a sketch to final image

00:08:22
Because then you let
the AI fill in the gaps on the way

00:08:25
And you see here they
have nothing to do with each other

00:08:28
except they are two cats

00:08:30
But if you want to maintain

00:08:32
and just do slight changes

00:08:35
You can actually create this into a character

00:08:37
So the same workflow making the illustration

00:08:41
and then vectorizing that illustration

00:08:44
with a few details

00:08:47
Then a soft brushwork

00:08:49
And the fur becomes more realistic

00:08:51
Then you push that to more realistic image

00:08:55
And then finally
you have full photorealism

00:08:57
with detailed images

00:08:59
I can show you quickly

00:09:00
This is Sora

00:09:02
And it’s a standalone product from OpenAI

00:09:05
where you can create images and videos

00:09:08
I go in here every morning

00:09:10
to see what the
community creates and what is trending

00:09:14
And that day I saw that

00:09:17
it was trending to create animals as humans

00:09:22
So I thought

00:09:24
I don’t know if Kang is yet impressed by this

00:09:27
Maybe I can try this

00:09:29
So I created this image

00:09:32
And we pushed that to full circle

00:09:34
This was when I get pretty impressed

00:09:37
how well ChatGPT
actually mixes images, etcetera

00:09:40
I wanted it to stand
and face its first illustration

00:09:44
And this image is
all generated in ChatGPT

00:09:49
When you sit there
and you generate, it’s hard to stop

00:09:52
So we took it even further

00:09:54
This cat wanted to erase its original

00:09:57
and create a message

00:10:00
“You don’t need to where you start

00:10:03
you can go endless possibilities”

00:10:05
Did you input two images for this?

00:10:07
or did it already know the context?

00:10:10
Yeah

00:10:10
It remembers the chat’s history

00:10:12
So what I did was

00:10:13
I used those two as references

00:10:15
I marked those two

00:10:16
“Make this cat sit
on a sofa reading that original series”

00:10:21
You created a character, basically

00:10:22
Yeah, exactly

00:10:24
And what we could do, later on

00:10:26
This cat can actually drive away

00:10:28
We can make it into a video, and

00:10:31
it’s all within minutes, actually

00:10:34
But, how does it relate to our customers?

00:10:37
What do they get from this?

00:10:39
Are you just
playing around at work, Alexander?

00:10:41
Sometimes it feels like that, but

00:10:45
You asked me if it feels like I play around

00:10:49
Sometimes it actually does

00:10:50
So I had to write another article

00:10:53
because colleagues, and some friends asked

00:10:57
“What’s the value
for an organization using this?”

00:11:01
This is just an example of how,

00:11:03
when you come up with ideas,

00:11:05
can actually steer
the end result of the AI outputs

00:11:09
You know, two years ago

00:11:11
it was all very random

00:11:13
You had to explain
in words what you were visualizing

00:11:17
and then it was a person
with eight fingers looking like this

00:11:23
And now
you’re can in detail, go in and actually,

00:11:27
“No, this is how I wanted”

00:11:28
So I thought we should do a live experiment

00:11:31
Should we do it now, or?

00:11:32
Yeah, we can do it

00:11:33
You want us to draw on the whiteboard?

00:11:36
Yeah

00:11:39
How about you draw?

00:11:42
Okay

00:11:43
Last time I did it, so now it’s your turn!

00:11:45
How is your drawing skills, Niyat?

00:11:48
So let’s imagine that

00:11:49
one of our customers are going to a trade show

00:11:52
and we need to come up with an idea

00:11:54
for what the booth is gonna look like

00:11:56
And they have two products

00:11:59
Yeah

00:11:59
So let’s create, like,

00:12:01
One

00:12:02
Yeah, that’s good

00:12:03
On one side,
we have a product, “Product 1”

00:12:07
And on the right side we have “Product 2”

00:12:16
And on the top

00:12:17
can you create like,
you know, a big banner

00:12:23
Yeah, and you can write out “Logo”

00:12:35
That’s good

00:12:37
And at the entrance here in the middle

00:12:41
Let’s do a reception, like

00:12:43
just the table, yeah like that

00:12:46
And a round table is perfect

00:12:49
Yeah

00:12:50
I’m excited

00:12:54
So do you think this is a good idea, Kang?

00:12:57
Yeah

00:12:58
You come there
and you see the two products clearly

00:13:01
You see the logo, and some sales stuff

00:13:04
So, that’s really good

00:13:06
We take a photo of it

00:13:17
Do you have anything to talk about

00:13:18
while I’m evolving this

00:13:19
What I’m doing is evolving

00:13:20
How long will it take?

00:13:22
It will take a while

00:13:23
I’ll do this in three steps

00:13:25
Like we did with the cat

00:13:27
While we’re waiting for Alexander

00:13:29
to finish his task

00:13:31
For you personally,

00:13:32
what impact has AI had for you?

00:13:34
Is there like a
single moment when you realized that

00:13:37
this changes everything?

00:13:42
Maybe there have been several moments

00:13:46
So I think I am more efficient at my work, and

00:13:51
yeah, also in some cases raised the quality

00:13:54
For example, when I started six years ago

00:13:57
with Zooma as a content creator

00:13:59
our boss Anders, he was asking me

00:14:01
“Niyat, we have so many articles

00:14:03
but now I want you to create a Slog”

00:14:05
A Slog?

00:14:06
Yeah, a Slog, like

00:14:08
It’s kind of a recorded blog article

00:14:11
So I have recorded all the authors,
and we have created

00:14:15
some audios that I then
edited and then uploaded to the article

00:14:20
which took a long time
because we have like, a lot of articles

00:14:25
Yeah

00:14:25
We have this Zooma book, right?

00:14:27
So you read…

00:14:28
Yeah

00:14:29
… through the whole book?

00:14:30
Yeah, for example

00:14:32
Now it’s very easy to create with AI

00:14:35
This post narration

00:14:36
We have like an AI voice

00:14:38
reading out loud your content

00:14:40
So it’s done by one click

00:14:43
It would be nice to hear the author

00:14:46
to read it, right

00:14:47
That would be the dream, but

00:14:48
It actually takes a lot of time to edit

00:14:51
when you record, you won’t do it perfectly

00:14:54
Not everyone is like, professional

00:14:55
Yes, and I think that’s pretty close

00:14:58
with tools like HeyGen, for example

00:15:00
where you just record
some words or some sentences

00:15:03
and then it automatically generates

00:15:06
whole articles with your voice

00:15:08
and another thing is AI translation

00:15:13
So before I asked all the authors

00:15:15
to translate their articles, sentence by sentence

00:15:18
because we have
The Onlinification Hub in English

00:15:21
but we have it also in Swedish

00:15:24
but now with tools like DeepL

00:15:27
where you can easily upload your article

00:15:30
and also upload some keywords

00:15:33
and it translates your article automatically

00:15:35
It is also integrated into HubSpot, our CMS

00:15:39
We have our website in 7 or 8 languages

00:15:42
So now, when we want to translate articles

00:15:45
you can do that very easily

00:15:47
by one click

00:15:48
But the way you create articles
have changed a bit, right?

00:15:53
Because you have
interviewed a lot of the colleagues

00:15:55
for the #MeetAZoomer series

00:15:58
Yeah

00:15:58
Are you using the
automatic transcription on Teams?

00:16:01
Yes

00:16:01
Then you can take it further in ChatGPT

00:16:03
to create drafts, and

00:16:04
Yeah, so it’s much more easy

00:16:07
and doesn’t take so
much time to create these articles

00:16:10
because you can
get the help from ChatGPT

00:16:13
You can use it also as a shoulder mate

00:16:15
It can shorten several parts

00:16:18
I found myself being sceptic

00:16:20
You know, for example “automatic audio”

00:16:22
The AI reading the blog post

00:16:25
but at the same time, I listen to it!

00:16:27
You know

00:16:28
I think if the articles
were just completely AI generated

00:16:32
Then they would not be interesting at all

00:16:35
But because it’s about something real

00:16:37
Then people don’t really care

00:16:40
Yeah, exactly

00:16:42
How’s it going, Alexander?

00:16:44
It’s good, I’m on the last image right now

00:16:47
You said that this would take seconds?

00:16:53
The process of pitching this idea

00:16:55
to a customer today would be that

00:16:57
Yeah, we sit in a one hour meeting

00:16:59
Then we brief an art director

00:17:01
And then they create that,
and we feedback

00:17:04
“Now we don’t want
the table to be like that”

00:17:07
They change that, and…

00:17:09
I mean, compared to that

00:17:10
it feels like seconds

00:17:13
So here we have it

00:17:17
It’s exactly like how you drew it, Niyat

00:17:21
I’m impressed

00:17:21
Wow, me too

00:17:25
We just need to put the the product into it

00:17:27
Yeah, exactly

00:17:29
So what we did here was that we
we did it in three steps

00:17:33
We took a photo of this layout

00:17:35
We made it into more cleaner shapes

00:17:39
And we took that image
to make it more with the

00:17:42
realistic materials and 3D generate it

00:17:47
Then we finally

00:17:49
took that image into a real setting

00:17:56
Now, whenever we have a Zooma meeting

00:18:00
everyone has to
create images for their slides

00:18:03
And that’s a way…

00:18:04
…yeah… why did you
come up with that idea, Niyat?

00:18:07
To force everyone to do it?

00:18:11
…to get everyone into it, and…

00:18:14
That was my idea!

00:18:17
No, no…

00:18:18
Your idea?

00:18:19
Yes!

00:18:21
In the workshops we had

00:18:22
I remember that!

00:18:23
Do you wanna say it?

00:18:25
No…

00:18:30
Because sometimes it’s easier

00:18:31
if you just get into it

00:18:34
and start with the tool, and

00:18:37
I mean, we are all
curious, but sometimes it is

00:18:39
a barrier to to use many different tools, so

00:18:43
I thought, okay,
then maybe it should just be mandatory

00:18:46
Yeah

00:18:47
Some people think that AI generated images

00:18:50
“That’s for the marketing department to do”

00:18:53
And I have friends working

00:18:55
you know, with finance, or HR

00:18:59
and they don’t really,
you know, are into these tools yet

00:19:02
But it’s so important just to start testing

00:19:07
and you will come up
with use cases within the organizations

00:19:11
For example

00:19:12
Imagine that

00:19:13
a customer support person

00:19:17
receiving a lot of complaints

00:19:19
about the packaging of the product

00:19:22
That person who
has that knowledge and the input

00:19:25
she can just take
a photo of the packaging, and

00:19:29
you know, tweak it and, send it to

00:19:32
innovation team, or something

00:19:34
like, “Here’s an idea that I have”

00:19:36
So, your point is that generative AI
is not only for creative, artist type of stuff

00:19:42
but it can be used on a bigger organization

00:19:45
He’s basically
encouraging people to be more creative

00:19:47
Exactly!

00:19:48
Yeah

00:19:50
I think that people who weren’t visual before,
can be more visual now

00:19:55
because there is a
tool that can bridge the gap

00:19:58
from being verbal to visual

00:20:01
Our colleague Stellan wrote me a while ago

00:20:04
He sent a link to a concert with the text:

00:20:07
“Kids today miss out on (all the) bands.

00:20:11
No one wants to go through the work
of getting along with 3 to 4 other people,

00:20:16
writing songs, rehearsing in the studio,
traveling around in a bus

00:20:21
with a negative bank account.

00:20:23
But man, when they did

00:20:24
it turned out absolutely amazing.”

00:20:27
Okay, so here’s my analysis

00:20:30
The form a band,

00:20:31
each individual has to
master an instrument first, right?

00:20:35
And find 2 to 3 other like-minded people

00:20:40
and hope that something cool is produced,
after countless hours of playing together

00:20:45
And that’s a lot of conditions
that must be met

00:20:47
But with AI,
you can bypass all the initial struggle

00:20:50
Jump directly to the creation part

00:20:53
It doesn’t eliminate the need for performers,

00:20:56
but if you’re more interested, like, in

00:20:57
experimenting with
the sounds and expressions

00:21:00
you don’t need
to rely on a whole team for that

00:21:03
Because you’re not into music creation

00:21:05
you have also not tried Suno, or?

00:21:08
I’ve heard about it, yeah

00:21:10
So that’s what I want us to do today

00:21:12
I want to create a song together in Suno

00:21:16
to give you a taste of music creation

00:21:20
So, okay, the process…

00:21:24
Usually I don’t have a process

00:21:25
but for today we going to have a process

00:21:27
So I’m thinking first,
we want to create the lyrics for our song

00:21:32
that is about what we have talked about

00:21:35
What should the prompt be?

00:21:37
“Create song lyrics about…

00:21:41
Kang, Niyat and Alexander

00:21:46
talks about generative AI”

00:21:57
Anything else?

00:21:58
Yeah, AI translation

00:22:05
Health and translation

00:22:08
Transcription And,

00:22:11
that has changed the process

00:22:14
What about the host?

00:22:15
Did he ask good questions, or?

00:22:18
“Asked bad questions”

00:22:21
Okay

00:22:23
“Keep it lighthearted”

00:22:28
Okay, let’s see

00:22:31
“Funky pop, nerdy rap fusion”

00:22:34
Sounds like my type of song!

00:22:35
You like that genre?

00:22:38
Hey, you

00:22:43
Now I’m gonna take this

00:22:44
So this is Suno

00:22:46
There’s a simple prompt,
where you  can just write

00:22:49
but I’m gonna use the custom one,
because I want to paste the lyrics here

00:22:53
“Prompt party”

00:22:56
Style tag

00:22:57
“Funky pop

00:23:01
nerdy rap fusion

00:23:04
Influences from Chinese music,

00:23:11
Eritrean music, German music,

00:23:16
and Swedish surfer music”

00:23:24
Let’s see

00:23:33
Is this now the song already?

00:23:36
With the text?

00:23:38
[AI song playing]

00:23:49
Can I listen to that?

00:23:51
Oh, that’s a good beat!

00:23:55
They create two versions

00:24:18
That’s some good text

00:24:30
I’m so amazed

00:24:31
What a cool tool

00:24:32
Did you know that it was that good?

00:24:34
No

00:24:35
And we didn’t give it that much input

00:24:37
but it sums up
what we talked about today

00:24:40
Do we have anything else to say
before we wrap up?

00:24:44
Before we RAP up?

00:24:47
Yeah

00:24:47
Until next time

00:24:49
Keep learning

00:24:50
Bye bye