part 2

0:00

S… Speaker 1 (part 2)

I don't have to call the phone.

0:00

S… Speaker 2 (part 2)

I'm going to call

0:05

S… Speaker 2 (part 2)

Vamsi and tell him that I'll be 30 minutes late.

0:07

S… Speaker 1 (part 2)

Okay.

0:08

S… Speaker 2 (part 2)

It'll call on my behalf.

0:09

S… Speaker 2 (part 2)

Because there are all these calling,

0:12

S… Speaker 2 (part 2)

AI calling apps.

0:13

S… Speaker 2 (part 2)

There is a tool called as Wapi.

0:15

S… Speaker 1 (part 2)

Wapi.

0:16

S… Speaker 1 (part 2)

Okay.

0:16

S… Speaker 1 (part 2)

Wapi.

0:17

S… Speaker 2 (part 2)

It has something called as,

0:18

S… Speaker 2 (part 2)

it has API integration.

0:19

S… Speaker 2 (part 2)

So, that is plugged into Jerry.

0:22

S… Speaker 2 (part 2)

Plus,

0:23

S… Speaker 2 (part 2)

11 labs,

0:23

S… Speaker 1 (part 2)

I have a voice.

0:24

S… Speaker 1 (part 2)

I don't have audio content.

0:25

S… Speaker 2 (part 2)

All my social media content is AI generated.

0:28

S… Speaker 1 (part 2)

So,

0:29

S… Speaker 2 (part 2)

when

0:33

S… Speaker 2 (part 2)

it talks, it will talk like me.

0:35

S… Speaker 2 (part 2)

It will sound like me,

0:36

S… Speaker 2 (part 2)

but it will say, I am webhouse AI executive.

0:38

S… Speaker 1 (part 2)

Yeah, I'll understand also.

0:39

S… Speaker 2 (part 2)

You'll understand.

0:40

S… Speaker 2 (part 2)

But people who don't know me,

0:41

S… Speaker 1 (part 2)

who don't talk to me,

0:42

S… Speaker 2 (part 2)

they will not know.

0:43

S… Speaker 1 (part 2)

So, I have to say.

0:43

S… Speaker 1 (part 2)

So,

0:44

S… Speaker 1 (part 2)

it will call.

0:45

S… Speaker 2 (part 2)

But that is not the crazy part.

0:47

S… Speaker 2 (part 2)

Let's say you don't pick up the phone Okay

0:50

S… Speaker 2 (part 2)

I have set up the guidelines that call,

0:53

S… Speaker 2 (part 2)

if the person doesn't pick up,

0:54

S… Speaker 2 (part 2)

call after five minutes.

0:55

S… Speaker 2 (part 2)

If the person still doesn't pick up,

0:57

S… Speaker 2 (part 2)

call after 30 minutes.

0:58

S… Speaker 1 (part 2)

Okay.

0:58

S… Speaker 2 (part 2)

If the person still doesn't pick up,

1:00

S… Speaker 2 (part 2)

if you have the email or WhatsApp,

1:02

S… Speaker 2 (part 2)

drop him an email or WhatsApp if needed.

1:04

S… Speaker 2 (part 2)

So there are tags,

1:06

S… Speaker 2 (part 2)

urgency tags.

1:07

S… Speaker 2 (part 2)

This is important.

1:08

S… Speaker 2 (part 2)

Make sure you call Vamsi and tell 30 minutes.

1:10

S… Speaker 2 (part 2)

If I just say call Vamsi and say,

1:13

S… Speaker 2 (part 2)

then there's one more playbook for that.

1:15

S… Speaker 2 (part 2)

Let's say no one.

1:20

S… Speaker 2 (part 2)

30 minutes late,

1:20

S… Speaker 1 (part 2)

you say,

1:21

S… Speaker 2 (part 2)

bro, do you want to eat anything?

1:23

S… Speaker 1 (part 2)

Let me know.

1:23

S… Speaker 2 (part 2)

That message could be released?

1:24

S… Speaker 1 (part 2)

No,

1:25

S… Speaker 2 (part 2)

no. You let

1:30

S… Speaker 2 (part 2)

me know what you want.

1:31

S… Speaker 2 (part 2)

If I say, yes,

1:32

S… Speaker 2 (part 2)

I want to eat, please order.

1:33

S… Speaker 1 (part 2)

Then again,

1:33

S… Speaker 2 (part 2)

it will call you back and tell you that.

1:35

S… Speaker 1 (part 2)

Crazy.

1:35

S… Speaker 1 (part 2)

Right?

1:37

S… Speaker 2 (part 2)

So, Jerry can do anything right now.

1:39

S… Speaker 1 (part 2)

Right?

1:40

S… Speaker 2 (part 2)

Let's say,

1:42

S… Speaker 2 (part 2)

I have a team.

1:43

S… Speaker 1 (part 2)

Like,

1:44

S… Speaker 2 (part 2)

data like analyst Chester,

1:46

S… Speaker 2 (part 2)

YouTube and all.

1:47

S… Speaker 2 (part 2)

And I was giving this example to your team.

1:50

S… Speaker 2 (part 2)

I treat Jerry like an actual very smart employee.

1:54

S… Speaker 2 (part 2)

I don't question.

1:55

S… Speaker 2 (part 2)

Can it do,

1:56

S… Speaker 2 (part 2)

can it not do?

1:56

S… Speaker 2 (part 2)

Let's say I want to understand what is working

2:00

S… Speaker 1 (part 2)

on my YouTube.

2:01

S… Speaker 2 (part 2)

I just tell Jerry,

2:03

S… Speaker 2 (part 2)

go into my YouTube because it already knows who I am.

2:06

S… Speaker 2 (part 2)

Context initially onboarding with HSN.

2:08

S… Speaker 2 (part 2)

It's a long process on OpenClock to set this up.

2:11

S… Speaker 2 (part 2)

Now I can just say,

2:13

S… Speaker 2 (part 2)

Jerry, figure out what is working on YouTube

2:17

S… Speaker 1 (part 2)

for me.

2:19

S… Speaker 2 (part 2)

And also what could be some outlier content pieces

2:23

S… Speaker 2 (part 2)

that I can create from my vidIQ account.

2:25

S… Speaker 2 (part 2)

Then it can open my YouTube with

2:29

S… Speaker 2 (part 2)

its own browser while I'm sleeping.

2:32

S… Speaker 1 (part 2)

Right.

2:32

S… Speaker 2 (part 2)

Or while wherever I am,

2:34

S… Speaker 2 (part 2)

doesn't matter.

2:34

S… Speaker 2 (part 2)

Then Slack or message.

2:35

S… Speaker 2 (part 2)

It can open my YouTube on the computer.

2:39

S… Speaker 2 (part 2)

on my back at my home,

2:41

S… Speaker 2 (part 2)

go to YouTube.

2:42

S… Speaker 2 (part 2)

If it understands that it can,

2:44

S… Speaker 2 (part 2)

it has to download those reports.

2:45

S… Speaker 1 (part 2)

Analytics download,

2:46

S… Speaker 2 (part 2)

outliers,

2:49

S… Speaker 2 (part 2)

whatever it needs to do.

2:50

S… Speaker 1 (part 2)

It can,

2:51

S… Speaker 2 (part 2)

it has access to its own browser.

2:53

S… Speaker 2 (part 2)

They can code,

2:54

S… Speaker 2 (part 2)

the product code.

2:55

S… Speaker 2 (part 2)

For example,

2:56

S… Speaker 2 (part 2)

data analysis.

2:57

S… Speaker 2 (part 2)

Data analysis found its insights.

3:00

S… Speaker 2 (part 2)

It has to present that data to me.

3:03

S… Speaker 2 (part 2)

So it has a habit of building into dashboards because that is what I say.

3:06

S… Speaker 2 (part 2)

Every time I ask you for data,

3:08

S… Speaker 2 (part 2)

always give me a dashboard,

3:09

S… Speaker 2 (part 2)

not a slab of text.

3:11

S… Speaker 2 (part 2)

So you can

3:18

S… Speaker 2 (part 2)

imagine, like you can think anything that you want to do.

3:21

S… Speaker 1 (part 2)

Yeah.

3:21

S… Speaker 1 (part 2)

Let's just imagine if this Jerry is a smart employee or a

3:25

S… Speaker 1 (part 2)

set of employees and Google employees,

3:27

S… Speaker 1 (part 2)

what would be the cost of them?

3:28

S… Speaker 2 (part 2)

Good question.

3:31

S… Speaker 2 (part 2)

Two ways of looking at it.

3:32

S… Speaker 1 (part 2)

Okay.

3:33

S… Speaker 2 (part 2)

You can almost run it for free.

3:35

S… Speaker 1 (part 2)

Okay.

3:36

S… Speaker 2 (part 2)

I'll tell you what it

3:40

S… Speaker 1 (part 2)

costs me.

3:40

S… Speaker 2 (part 2)

It costs me almost close to a lakh per month right now.

3:44

S… Speaker 2 (part 2)

I'll tell you why.

3:45

S… Speaker 2 (part 2)

I keep experimenting.

3:50

S… Speaker 2 (part 2)

Because I want to know what it can do.

3:52

S… Speaker 1 (part 2)

Because I want to know what it can do.

4:00

S… Speaker 2 (part 2)

So I keep trying a lot of things.

4:02

S… Speaker 2 (part 2)

Mine is an edge case.

4:05

S… Speaker 1 (part 2)

Okay.

4:05

S… Speaker 2 (part 2)

Right. I think if you would use it or someone like that will use it,

4:09

S… Speaker 2 (part 2)

I don't think it will cost more than 10 ,000 rupees,

4:11

S… Speaker 1 (part 2)

5 ,000 rupees per month.

4:12

S… Speaker 2 (part 2)

Depends on how much you use it.

4:14

S… Speaker 1 (part 2)

Now,

4:15

S… Speaker 2 (part 2)

you want the best out of an open claw,

4:18

S… Speaker 2 (part 2)

then you will spend that money and that money is going into API costs.

4:37

S… Speaker 2 (part 2)

The way you

4:45

S… Speaker 1 (part 2)

can do it for free.

4:46

S… Speaker 2 (part 2)

Yes, of course,

4:47

S… Speaker 2 (part 2)

it will not perform as well as this will do.

4:49

S… Speaker 2 (part 2)

But I think in one year,

4:51

S… Speaker 2 (part 2)

again, next time when we are talking,

4:53

S… Speaker 2 (part 2)

the free one will be as good as the paid one.

4:56

S… Speaker 2 (part 2)

And the free one is,

4:57

S… Speaker 2 (part 2)

like I told you,

5:00

S… Speaker 1 (part 2)

There's a cost to run all of this.

5:01

S… Speaker 1 (part 2)

The AI.

5:02

S… Speaker 1 (part 2)

The cost of this is AI.

5:04

S… Speaker 1 (part 2)

The compute,

5:05

S… Speaker 1 (part 2)

the tokens that AI uses.

5:06

S… Speaker 1 (part 2)

We need the cost of the chat GPT -1.

5:07

S… Speaker 1 (part 2)

And we use

5:12

S… Speaker 1 (part 2)

it. In the

5:17

S… Speaker 1 (part 2)

backend.

5:18

S… Speaker 1 (part 2)

What is it doing?

5:19

S… Speaker 1 (part 2)

It is using something called as tokens.

5:21

S… Speaker 2 (part 2)

Simplify it.

5:23

S… Speaker 1 (part 2)

In the words input,

5:24

S… Speaker 1 (part 2)

in the question,

5:25

S… Speaker 1 (part 2)

in the words,

5:26

S… Speaker 1 (part 2)

the answer is the cost.

5:28

S… Speaker 1 (part 2)

Let's say $1,

5:29

S… Speaker 2 (part 2)

$0 .20,

5:30

S… Speaker 1 (part 2)

$0 .10,

5:31

S… Speaker 1 (part 2)

whatever.

5:31

S… Speaker 1 (part 2)

That is the cost of compute.

5:34

S… Speaker 1 (part 2)

Because that AI model is

5:38

S… Speaker 1 (part 2)

running on a GPU.

5:40

S… Speaker 1 (part 2)

NVIDIA GPU,

5:41

S… Speaker 1 (part 2)

or Google Tensor Processing Unit,

5:44

S… Speaker 1 (part 2)

or TPU.

5:46

S… Speaker 1 (part 2)

And that is the cost.

5:48

S… Speaker 1 (part 2)

The hardware cost is what is being built to you.

5:51

S… Speaker 1 (part 2)

Because we don't have to worry about it.

5:53

S… Speaker 2 (part 2)

But today,

5:56

S… Speaker 1 (part 2)

AI models are becoming very,

5:58

S… Speaker 1 (part 2)

very smart.

5:58

S… Speaker 1 (part 2)

And from that AI model,

6:00

S… Speaker 1 (part 2)

we run into the computer.

6:03

S… Speaker 1 (part 2)

Not as powerful as,

6:04

S… Speaker 1 (part 2)

let's say, GPT 5 .4 or Gemini 3 .1 Pro or

6:09

S… Speaker 1 (part 2)

Opus 4 .6.

6:10

S… Speaker 1 (part 2)

These are three best models right now,

6:12

S… Speaker 1 (part 2)

right? Which are available to public at this point of time,

6:15

S… Speaker 2 (part 2)

right?

6:15

S… Speaker 1 (part 2)

Those are very big models.

6:17

S… Speaker 1 (part 2)

And the computer will run over.

6:18

S… Speaker 1 (part 2)

But if you have an email to rewrite,

6:22

S… Speaker 1 (part 2)

you don't have a model.

6:24

S… Speaker 1 (part 2)

To go and pick up Diet

6:28

S… Speaker 1 (part 2)

Coke in Banjara Hills with full

6:32

S… Speaker 1 (part 2)

traffic.

6:33

S… Speaker 1 (part 2)

It's better to go in a Rapido than your Ferrari.

6:36

S… Speaker 1 (part 2)

Correct?

6:36

S… Speaker 1 (part 2)

But we're using a Ferrari because we have a Ferrari.

6:39

S… Speaker 2 (part 2)

Right?

6:41

S… Speaker 1 (part 2)

Obviously,

6:42

S… Speaker 1 (part 2)

if you have a Ferrari,

6:43

S… Speaker 1 (part 2)

you should go use it.

6:44

S… Speaker 1 (part 2)

That's a different story altogether.

6:45

S… Speaker 1 (part 2)

So,

6:47

S… Speaker 1 (part 2)

we are using a very powerful model for things that

6:51

S… Speaker 1 (part 2)

you don't need it for.

6:53

S… Speaker 1 (part 2)

So, there was a study,

6:54

S… Speaker 1 (part 2)

right? 85 -90 % of the tasks need to use AI.

6:58

S… Speaker 1 (part 2)

Look at those

7:05

S… Speaker 1 (part 2)

tasks.

7:06

S… Speaker 1 (part 2)

80 % of them doesn't need the power of the AI model.

7:12

S… Speaker 1 (part 2)

Seeking information.

7:13

S… Speaker 1 (part 2)

This is question and answer.

7:14

S… Speaker 1 (part 2)

Practical guidance.

7:16

S… Speaker 1 (part 2)

Again,

7:17

S… Speaker 1 (part 2)

question and answer.

7:18

S… Speaker 1 (part 2)

Writing.

7:19

S… Speaker 1 (part 2)

Technical help.

7:20

S… Speaker 1 (part 2)

In technical help,

7:21

S… Speaker 1 (part 2)

we're including a smaller section under coding.

7:23

S… Speaker 2 (part 2)

Right?

7:24

S… Speaker 1 (part 2)

So 80 % of the tasks.

7:26

S… Speaker 1 (part 2)

This is the older report,

7:28

S… Speaker 1 (part 2)

by the way.

7:28

S… Speaker 1 (part 2)

This is two years old.

7:29

S… Speaker 1 (part 2)

OpenAI report is two years old.

7:31

S… Speaker 1 (part 2)

There's a recent anthropic report also,

7:33

S… Speaker 1 (part 2)

but very similar direction.

7:36

S… Speaker 1 (part 2)

Even it goes on,

7:37

S… Speaker 1 (part 2)

you don't need a very smart model.

7:38

S… Speaker 1 (part 2)

You don't need a very large model.

7:40

S… Speaker 1 (part 2)

Smart model is the wrong point.

7:42

S… Speaker 2 (part 2)

Not,

7:44

S… Speaker 1 (part 2)

if the model is large enough,

7:45

S… Speaker 1 (part 2)

doesn't make it just smart.

7:46

S… Speaker 2 (part 2)

Like models,

7:47

S… Speaker 1 (part 2)

larger models are good at everything.

7:49

S… Speaker 2 (part 2)

Right.

7:50

S… Speaker 2 (part 2)

But if you know,

7:52

S… Speaker 1 (part 2)

if you have a single model,

7:53

S… Speaker 1 (part 2)

you need coding.

7:54

S… Speaker 1 (part 2)

You need math solving.

7:57

S… Speaker 1 (part 2)

You need science.

7:59

S… Speaker 1 (part 2)

Because those are hard problems.

8:01

S… Speaker 1 (part 2)

So AI is trying to beat those things.

8:03

S… Speaker 1 (part 2)

But when it comes to writing,

8:05

S… Speaker 1 (part 2)

A small AI model,

8:07

S… Speaker 1 (part 2)

which can run on your computer,

8:08

S… Speaker 1 (part 2)

can do a 90 % of the job.

8:11

S… Speaker 2 (part 2)

Yesterday,

8:13

S… Speaker 1 (part 2)

he had a question like this.

8:15

S… Speaker 2 (part 2)

Okay,

8:16

S… Speaker 1 (part 2)

for example,

8:17

S… Speaker 1 (part 2)

if you have a phone,

8:18

S… Speaker 1 (part 2)

Google Edge AI app,

8:22

S… Speaker 1 (part 2)

Edge Gallery.

8:23

S… Speaker 1 (part 2)

Okay.

8:24

S… Speaker 2 (part 2)

Right.

8:24

S… Speaker 1 (part 2)

We will send all the links to people so you can download.

8:27

S… Speaker 1 (part 2)

Edge Gallery.

8:29

S… Speaker 1 (part 2)

A free app by Google.

8:30

S… Speaker 1 (part 2)

You have the ability on an Android phone.

8:34

S… Speaker 1 (part 2)

You have the ability to install AI models which are very small.

8:38

S… Speaker 1 (part 2)

For example,

8:38

S… Speaker 1 (part 2)

Gemma 4.

8:39

S… Speaker 1 (part 2)

Okay.

8:40

S… Speaker 1 (part 2)

Is a new AI model that came from Google.

8:42

S… Speaker 1 (part 2)

Which is a 1 billion parameter model.

8:44

S… Speaker 1 (part 2)

And the size of that model,

8:46

S… Speaker 1 (part 2)

hypothetically speaking,

8:47

S… Speaker 1 (part 2)

is 1 by 1000 of the size of Opus.

8:51

S… Speaker 2 (part 2)

Oh.

8:52

S… Speaker 2 (part 2)

Which you use.

8:53

S… Speaker 2 (part 2)

No,

8:53

S… Speaker 1 (part 2)

I don't know.

8:54

S… Speaker 1 (part 2)

I don't know.

8:54

S… Speaker 1 (part 2)

I don't know.

8:55

S… Speaker 1 (part 2)

I don't know.

8:57

S… Speaker 1 (part 2)

It is bloody good.

8:59

S… Speaker 1 (part 2)

It's 80 % there.

9:00

S… Speaker 1 (part 2)

And 70 % there.

9:02

S… Speaker 1 (part 2)

Rewriting tasks,

9:03

S… Speaker 1 (part 2)

this message pump,

9:04

S… Speaker 1 (part 2)

pump it up,

9:05

S… Speaker 1 (part 2)

this question comes in.

9:07

S… Speaker 1 (part 2)

That's what it is.

9:08

S… Speaker 1 (part 2)

It's running on your phone.

9:09

S… Speaker 1 (part 2)

Internet is also running on your phone.

9:11

S… Speaker 1 (part 2)

I have that app installed on my phone so that if I have some work to

9:16

S… Speaker 1 (part 2)

be done when I'm on a flight,

9:17

S… Speaker 1 (part 2)

I use that model.

9:21

S… Speaker 1 (part 2)

That's on your phone.

9:22

S… Speaker 1 (part 2)

It's working for free.

9:23

S… Speaker 1 (part 2)

Forever.

9:24

S… Speaker 1 (part 2)

And it has become 10 times more powerful than it was

9:28

S… Speaker 1 (part 2)

one year back.

9:29

S… Speaker 1 (part 2)

That was Gemma 4 which dropped.

9:31

S… Speaker 2 (part 2)

Is it available both on iOS and Android?

9:32

S… Speaker 1 (part 2)

iOS could be absent.

9:34

S… Speaker 1 (part 2)

Okay.

9:34

S… Speaker 1 (part 2)

I think Gallery is only Android,

9:37

S… Speaker 1 (part 2)

right? Edge Gallery is of Google,

9:39

S… Speaker 1 (part 2)

so it's only Android.

9:39

S… Speaker 1 (part 2)

But iOS could be absent and then links pump it up.

9:42

S… Speaker 1 (part 2)

You can run your own AI,

9:45

S… Speaker 1 (part 2)

which is...

9:46

S… Speaker 1 (part 2)

probably 15 -20 times more powerful than your phone.

9:49

S… Speaker 1 (part 2)

Okay.

9:49

S… Speaker 1 (part 2)

Phone is small.

9:50

S… Speaker 1 (part 2)

You can run very small models.

9:52

S… Speaker 1 (part 2)

Laptop has...

9:54

S… Speaker 1 (part 2)

Every laptop has GPU.

9:57

S… Speaker 1 (part 2)

Every laptop has secondary memory.

9:59

S… Speaker 1 (part 2)

In fact, if you're using...

10:00

S… Speaker 1 (part 2)

Using Mac, it's even better.

10:01

S… Speaker 1 (part 2)

Because they do something called as unified memory.

10:03

S… Speaker 1 (part 2)

Danwalla,

10:04

S… Speaker 1 (part 2)

you can actually run 10 -15 billion parameter models.

10:08

S… Speaker 1 (part 2)

Which are again insanely good.

10:11

S… Speaker 1 (part 2)

Like, Quen 3 .5.

10:13

S… Speaker 1 (part 2)

Incredible model.

10:15

S… Speaker 1 (part 2)

Gemma again dropped recently.

10:17

S… Speaker 1 (part 2)

Incredible model.

10:18

S… Speaker 1 (part 2)

Open source model this way.

10:19

S… Speaker 1 (part 2)

So you can run AI on your computer without internet

10:23

S… Speaker 1 (part 2)

access for free for life.

10:26

S… Speaker 1 (part 2)

Crazy.

10:29

S… Speaker 1 (part 2)

You can do probably

10:33

S… Speaker 1 (part 2)

70 % of the things that you want to do at a very,

10:37

S… Speaker 1 (part 2)

very high quality without spending a single rupee.

10:39

S… Speaker 1 (part 2)

And for laptop, you can use a tool like Olama or some

10:43

S… Speaker 1 (part 2)

UI,

10:45

S… Speaker 1 (part 2)

there's a tool called LM studio.

10:49

S… Speaker 2 (part 2)

Okay.

10:50

S… Speaker 1 (part 2)

And the other advantage of doing this is that when you use

10:54

S… Speaker 1 (part 2)

an open source model,

10:55

S… Speaker 2 (part 2)

then advantage is

10:56

S… Speaker 1 (part 2)

Your information is on your computer.

10:58

S… Speaker 1 (part 2)

What I

11:05

S… Speaker 1 (part 2)

am realizing about myself,

11:06

S… Speaker 1 (part 2)

and everybody has to realize this about themselves as well,

11:09

S… Speaker 1 (part 2)

we are opening up about things with AI

11:13

S… Speaker 1 (part 2)

that we will not with any human in the world.

11:15

S… Speaker 1 (part 2)

Deepest insecurities.

11:25

S… Speaker 1 (part 2)

because it used to give articles now AI gives you opinion

11:29

S… Speaker 1 (part 2)

and on top of that AI is sycrophantic sycrophantic

11:33

S… Speaker 1 (part 2)

and AI is like yes bro you're

11:37

S… Speaker 1 (part 2)

right I trust you you're awesome right

11:42

S… Speaker 1 (part 2)

so we are becoming more connected because you're getting acknowledgement

11:46

S… Speaker 1 (part 2)

from it as it is that you end up saying a lot of things that are very

11:50

S… Speaker 1 (part 2)

private to you

11:52

S… Speaker 1 (part 2)

Like last time,

11:53

S… Speaker 1 (part 2)

I think around six months back,

11:55

S… Speaker 1 (part 2)

around 25 million or 40 million

12:00

S… Speaker 1 (part 2)

chat GPT chats were leaked because of a breach.

12:05

S… Speaker 1 (part 2)

When I read that,

12:08

S… Speaker 1 (part 2)

I'm like,

12:09

S… Speaker 1 (part 2)

I don't want my messages to get leaked,

12:11

S… Speaker 1 (part 2)

man. So anytime,

12:13

S… Speaker 1 (part 2)

I made a habit after that,

12:15

S… Speaker 1 (part 2)

by the way.

12:17

S… Speaker 1 (part 2)

every time when there is something like this,

12:18

S… Speaker 1 (part 2)

I will use a model which is running on my phone or my computer.

12:22

S… Speaker 1 (part 2)

So that the moment I delete that app,

12:25

S… Speaker 1 (part 2)

it's gone.

12:25

S… Speaker 1 (part 2)

Nobody has access to my information.

12:27

S… Speaker 1 (part 2)

So it could be a lot of things,

12:30

S… Speaker 1 (part 2)

a lot of insecurities.

12:31

S… Speaker 1 (part 2)

This is how I'm feeling.

12:32

S… Speaker 1 (part 2)

Is it okay?

12:33

S… Speaker 1 (part 2)

Every time I say something like this,

12:35

S… Speaker 1 (part 2)

I also say that I'm not recommending you all these things because

12:39

S… Speaker 1 (part 2)

you will do something wrong with AI.

12:42

S… Speaker 1 (part 2)

As a result,

12:42

S… Speaker 1 (part 2)

you should delete it.

12:44

S… Speaker 1 (part 2)

This is your privacy.

12:45

S… Speaker 1 (part 2)

It's your right.

12:46

S… Speaker 1 (part 2)

You should know that this can happen so

12:51

S… Speaker 1 (part 2)

that you're taking steps to make sure your privacy is protected.

12:55

S… Speaker 2 (part 2)

That's it.

12:56

S… Speaker 1 (part 2)

For example,

12:57

S… Speaker 1 (part 2)

OpenAI recently said that especially for kids,

13:00

S… Speaker 1 (part 2)

if they feel that the kids are getting towards direction of suicidal,

13:05

S… Speaker 1 (part 2)

they will notify their parents.

13:07

S… Speaker 2 (part 2)

How do they identify?

13:10

S… Speaker 1 (part 2)

By looking at messages,

13:11

S… Speaker 1 (part 2)

conversations.

13:12

S… Speaker 2 (part 2)

So these are kid accounts.

13:15

S… Speaker 2 (part 2)

Okay.

13:15

S… Speaker 2 (part 2)

So naturally guardian accounts.

13:17

S… Speaker 1 (part 2)

Yeah, guardian accounts.

13:17

S… Speaker 1 (part 2)

YouTube kids alone.

13:19

S… Speaker 1 (part 2)

Claude also said

13:23

S… Speaker 1 (part 2)

if you do anything illegal,

13:24

S… Speaker 1 (part 2)

there's a possibility that this information will get to law enforcement.

13:28

S… Speaker 2 (part 2)

How crazy.

13:29

S… Speaker 1 (part 2)

So these

13:37

S… Speaker 1 (part 2)

things are there and it should be there.

13:40

S… Speaker 2 (part 2)

That's right.

13:40

S… Speaker 2 (part 2)

Last time I met a client.

13:42

S… Speaker 2 (part 2)

That's right.

13:43

S… Speaker 1 (part 2)

Safety guardrails.

13:45

S… Speaker 1 (part 2)

Safety guardrails have come in slowly.

13:47

S… Speaker 1 (part 2)

And it has to come.

13:48

S… Speaker 1 (part 2)

There's no two ways around it.

13:49

S… Speaker 2 (part 2)

Of course, of course.

13:49

S… Speaker 2 (part 2)

And cost also,

13:50

S… Speaker 2 (part 2)

two quick questions.

13:51

S… Speaker 2 (part 2)

One is,

13:52

S… Speaker 2 (part 2)

see, you don't have to pay a lot on your scale.

13:56

S… Speaker 2 (part 2)

Let's talk about your case only.

13:58

S… Speaker 2 (part 2)

Versus,

13:59

S… Speaker 2 (part 2)

the amount spent on AI tools or AI models or subscriptions versus,

14:03

S… Speaker 2 (part 2)

if you want to pay a lot on your scale,

14:06

S… Speaker 2 (part 2)

you should pay a lot on your scale.

14:09

S… Speaker 2 (part 2)

Just the opportunity cost.

14:10

S… Speaker 1 (part 2)

I think it will be like.

14:12

S… Speaker 2 (part 2)

Let people understand.

14:13

S… Speaker 2 (part 2)

20x I guess.

14:15

S… Speaker 1 (part 2)

It will be 20.

14:16

S… Speaker 1 (part 2)

I can't even.

14:17

S… Speaker 2 (part 2)

So.

14:18

S… Speaker 2 (part 2)

And that to 24 hours with you.

14:20

S… Speaker 2 (part 2)

Yeah.

14:20

S… Speaker 1 (part 2)

Anytime.

14:21

S… Speaker 1 (part 2)

Array bro like.

14:22

S… Speaker 1 (part 2)

The kind of systems.

14:24

S… Speaker 1 (part 2)

That we have managed to build.

14:27

S… Speaker 1 (part 2)

I don't think.

14:29

S… Speaker 1 (part 2)

I can hire a team

14:34

S… Speaker 1 (part 2)

which can do it.

14:34

S… Speaker 1 (part 2)

The way I want it to work like.

14:39

S… Speaker 1 (part 2)

You know that I create content now.

14:41

S… Speaker 1 (part 2)

Just like you get brand deals.

14:43

S… Speaker 2 (part 2)

You are not creating content.

14:44

S… Speaker 1 (part 2)

So my face creates content.

14:46

S… Speaker 1 (part 2)

My AI creates content.

14:47

S… Speaker 2 (part 2)

He's never there in front of the camera.

14:49

S… Speaker 1 (part 2)

Only in this podcast.

14:50

S… Speaker 2 (part 2)

Actually.

14:51

S… Speaker 2 (part 2)

Sorry.

14:53

S… Speaker 2 (part 2)

So.

14:56

S… Speaker 1 (part 2)

that line of business also works based on brand deals.

15:00

S… Speaker 2 (part 2)

The brands, yeah, yeah.

15:00

S… Speaker 1 (part 2)

Right.

15:01

S… Speaker 1 (part 2)

And so,

15:02

S… Speaker 1 (part 2)

the more brand deals,

15:03

S… Speaker 1 (part 2)

the better, higher quality brand deals,

15:05

S… Speaker 1 (part 2)

the better it is for the whole business.

15:06

S… Speaker 1 (part 2)

So, we have targets.

15:07

S… Speaker 1 (part 2)

Let's say,

15:08

S… Speaker 1 (part 2)

we are doing 100 rupees,

15:10

S… Speaker 1 (part 2)

we want to do 1 ,000 rupees next year.

15:12

S… Speaker 1 (part 2)

How do you scale?

15:13

S… Speaker 1 (part 2)

The only way to scale is to get more sponsors.

15:16

S… Speaker 2 (part 2)

Sponsors,

15:17

S… Speaker 1 (part 2)

one of the sponsors is always there.

15:18

S… Speaker 1 (part 2)

If you reach out,

15:20

S… Speaker 1 (part 2)

you reach out.

15:22

S… Speaker 1 (part 2)

So what we created,

15:24

S… Speaker 1 (part 2)

I'll give you this example.

15:24

S… Speaker 1 (part 2)

We created an AI agent.

15:26

S… Speaker 1 (part 2)

It's called the sponsorship bot.

15:28

S… Speaker 2 (part 2)

Okay.

15:29

S… Speaker 1 (part 2)

AI agent,

15:29

S… Speaker 1 (part 2)

content creators,

15:33

S… Speaker 1 (part 2)

50 to 80 members.

15:37

S… Speaker 1 (part 2)

Core AI will be 10,

15:39

S… Speaker 2 (part 2)

10,

15:40

S… Speaker 1 (part 2)

15 people like me.

15:41

S… Speaker 1 (part 2)

Yeah.

15:41

S… Speaker 1 (part 2)

Adjacent will be altogether 100 people.

15:44

S… Speaker 1 (part 2)

YouTube, Instagram,

15:45

S… Speaker 1 (part 2)

LinkedIn, Twitter,

15:46

S… Speaker 1 (part 2)

newsletters.

15:47

S… Speaker 1 (part 2)

250,

15:49

S… Speaker 1 (part 2)

300 properties.

15:50

S… Speaker 1 (part 2)

My AI.

15:53

S… Speaker 1 (part 2)

Every single day,

15:54

S… Speaker 1 (part 2)

I think around 10 or 12 times they do pulse check.

15:57

S… Speaker 1 (part 2)

Oh.

15:58

S… Speaker 1 (part 2)

Watch us every single video of every single piece of content that

16:02

S… Speaker 1 (part 2)

has gone out in any of these profiles.

16:04

S… Speaker 2 (part 2)

Of those 80 plus profiles.

16:05

S… Speaker 1 (part 2)

Of whatever profiles.

16:06

S… Speaker 1 (part 2)

I think 80, 90.

16:07

S… Speaker 1 (part 2)

My team keeps adding.

16:09

S… Speaker 1 (part 2)

Okay.

16:10

S… Speaker 1 (part 2)

So once she AI content created on a start,

16:12

S… Speaker 1 (part 2)

she gets added to that list.

16:13

S… Speaker 2 (part 2)

Okay. Right.

16:14

S… Speaker 2 (part 2)

Irrespective of the language.

16:15

S… Speaker 1 (part 2)

I don't care.

16:16

S… Speaker 1 (part 2)

Irrespective of the language.

16:17

S… Speaker 1 (part 2)

Right.

16:18

S… Speaker 1 (part 2)

But we mostly have been looking for English.

16:19

S… Speaker 1 (part 2)

Okay.

16:20

S… Speaker 1 (part 2)

We have not built the engine where we are telling AI,

сажетак

Питај ВИ о овом транкрипту