Google Wins The AI War, Here's How…

Kasim: 00:00:11

I wanna talk to you about why I think Google wins the AI war.

Kasim: 00:00:13

Now, first and foremost, I'm not an AI expert, unlike everybody on Twitter.

Kasim: 00:00:17

However, I do have access to information that not everybody has access to, and it

Kasim: 00:00:22

happens to be pertinent at the moment.

Kasim: 00:00:24

So what I'd like to do is just introduce my thesis.

Kasim: 00:00:26

And in order to do that, I'm gonna borrow.

Kasim: 00:00:29

The favorite part of my keynote ever.

Kasim: 00:00:31

So if you've ever seen me speak, you already kind of know all this, but if

Kasim: 00:00:34

you haven't, this is gonna be new to you.

Kasim: 00:00:35

And what it requires us to do is examine what Google really is at its core.

Kasim: 00:00:40

if you were to have to explain Google to like a baby or an alien, you needed to

Kasim: 00:00:44

tell them in a soundbite, what is Google?

Kasim: 00:00:46

You'd probably say Google's a search engine.

Kasim: 00:00:48

And, you know, fine, sure.

Kasim: 00:00:49

Google has a search engine and maybe start it as a search engine.

Kasim: 00:00:53

Or is, referred to or thought of as a search engine.

Kasim: 00:00:55

But that's such a myopic view of Google.

Kasim: 00:00:57

Like Google is search, of course.

Kasim: 00:00:59

And if you think about the information that's being offered to Google every

Kasim: 00:01:02

time you search, it's absolutely insane.

Kasim: 00:01:04

Like Google knows what you are looking for, what you're afraid of,

Kasim: 00:01:09

what you're interested in learning, and this is where you kinda have

Kasim: 00:01:11

to think in three dimensions.

Kasim: 00:01:12

Google knows what you don't know.

Kasim: 00:01:14

It knows your knowledge graph.

Kasim: 00:01:16

It knows where you're ignorant.

Kasim: 00:01:18

It knows your prejudices.

Kasim: 00:01:20

It knows your, your echo chamber and your confirmation bias, like an unbelievable

Kasim: 00:01:26

and potentially very damaging.

Kasim: 00:01:29

Database of information is being collected on every single person

Kasim: 00:01:32

who engages with Google search.

Kasim: 00:01:34

At any point, think about all the things that you ask Google, and then think

Kasim: 00:01:39

about what those things mean and, and, and, and then extrapolate that out the

Kasim: 00:01:45

way that a machine learning mechanism would be able to extrapolate out.

Kasim: 00:01:49

You know, so if you're asking a question, something as benign as, Hey, does missing

Kasim: 00:01:52

a mortgage payment hurt my credit?

Kasim: 00:01:54

Now think about all the things that, that actually could potentially mean to Google.

Kasim: 00:01:58

You know, maybe you're worried about a job loss, maybe you're

Kasim: 00:02:00

worried about a potential divorce.

Kasim: 00:02:01

You're clearly having money problems.

Kasim: 00:02:03

And then you combine that with some of the other things that Google has access to,

Kasim: 00:02:06

and those dots begin to connect and create A network of information that's used to

Kasim: 00:02:11

advertise to you, to be honest with you.

Kasim: 00:02:12

But that's not the only thing that's used for, and that's the

Kasim: 00:02:15

point that I'm trying to make.

Kasim: 00:02:15

So Google is Google search fine, but it's not just a search

Kasim: 00:02:18

engine because it's also YouTube.

Kasim: 00:02:19

It's the largest video repository on the planet.

Kasim: 00:02:22

And also, by the way, the second largest search engine in existence.

Kasim: 00:02:25

because Google has YouTube.

Kasim: 00:02:27

A lot like having Google search.

Kasim: 00:02:29

Google has information on you, on how you emote what you wanna learn what

Kasim: 00:02:32

music you listen to, where you sit politically what pisses you off, if

Kasim: 00:02:35

you're, inside of YouTube and commenting on things like Google's cataloging

Kasim: 00:02:39

Temperament attached to specific comments and what they might potentially mean.

Kasim: 00:02:43

But it's not just YouTube, it's also Gmail.

Kasim: 00:02:46

has the largest email utility in existence, which is also turning

Kasim: 00:02:50

into one of the largest enterprise email utilities in existence.

Kasim: 00:02:53

And every single email you send and receive is being read by Google.

Kasim: 00:02:57

And you might say, cousin jokes on you, I don't use Gmail.

Kasim: 00:02:59

Doesn't matter.

Kasim: 00:03:00

You communicate with people that use Gmail, which means Google's

Kasim: 00:03:02

cataloging the way you speak your semantic index, the adjectives and

Kasim: 00:03:06

adverbs that appeal to you the most.

Kasim: 00:03:08

Every attachment that you send inside of Gmail is fair game, which

Kasim: 00:03:11

means all of your taxes and tax information, massively private, like

Kasim: 00:03:16

medical documents and everything that you have inside of Gmail, Google.

Kasim: 00:03:20

Reads and then appends to your user and you know, you start to

Kasim: 00:03:24

see this picture of a person being formed and it's unbelievable.

Kasim: 00:03:28

But it's not just Gmail, it's also Google Apps.

Kasim: 00:03:30

It's Google Drive.

Kasim: 00:03:31

Google Docs.

Kasim: 00:03:31

Google Photos.

Kasim: 00:03:32

Google Slides.

Kasim: 00:03:32

Google Sheets.

Kasim: 00:03:33

Google knows your net worth.

Kasim: 00:03:35

They know what your tax filing looks like.

Kasim: 00:03:36

They know what your children look like.

Kasim: 00:03:37

That's what crazy.

Kasim: 00:03:38

Go to Google Photos.

Kasim: 00:03:39

If you use Google Photos, it's unbelievable.

Kasim: 00:03:41

and I have two children.

Kasim: 00:03:42

One of 'em is seven years old.

Kasim: 00:03:43

He's about to turn eight.

Kasim: 00:03:44

And Google is able to tell me, Sammy, what Sammy looked like from birth to eight.

Kasim: 00:03:49

Like, it watches him grow, but it still knows.

Kasim: 00:03:52

It still knows.

Kasim: 00:03:54

Think about everything that you have inside of Google Drive right

Kasim: 00:03:56

now, and Google uses that info.

Kasim: 00:03:58

Why on earth would they give you that unbelievable plethora of free services?

Kasim: 00:04:03

Unless they're using it to catalog information about you Google Calendar.

Kasim: 00:04:06

Are you kidding me?

Kasim: 00:04:07

Google knows exactly what you do when you do it, how you do it, at what frequencies,

Kasim: 00:04:09

what intervals, when it repeats.

Kasim: 00:04:11

But it's not just Google apps, it's also Google Maps.

Kasim: 00:04:14

Everybody's Google in their pocket.

Kasim: 00:04:15

Google knows where you go, where you live, where your kids go

Kasim: 00:04:17

to school, how fast you drive.

Kasim: 00:04:20

It takes 30 minutes to get from point A to point B, but you got there in

Kasim: 00:04:22

25, so you get car insurance ads.

Kasim: 00:04:24

It's also Android.

Kasim: 00:04:25

Google owns the largest operating system in the world.

Kasim: 00:04:28

And again, you say, I don't use Android, I'm an iPhone user.

Kasim: 00:04:30

Fine.

Kasim: 00:04:31

You communicate with people that use Android.

Kasim: 00:04:32

Google knows who you call, who calls you, who you text, who you ignore,

Kasim: 00:04:35

what you say, what you download, what apps you're using but it's

Kasim: 00:04:38

not just Android, it's Chrome.

Kasim: 00:04:40

It's a browser.

Kasim: 00:04:40

Everything you've ever typed into that browser, everything you've ever logged

Kasim: 00:04:43

in to, everything you've ever done, when you've been logged in, if you go to your,

Kasim: 00:04:48

account, you know, and, I mean like Google gets to see all of that, but it's not

Kasim: 00:04:52

just Chrome, it's also Google Analytics.

Kasim: 00:04:54

Google Analytics exists on 90% of all known websites, effectively

Kasim: 00:04:59

adjusted from margin of error.

Kasim: 00:05:01

It.

Kasim: 00:05:01

Is the most prolific thing on the internet.

Kasim: 00:05:03

It knows every website that you've gone to, everything you've done on those

Kasim: 00:05:06

websites, how long you were there.

Kasim: 00:05:08

Let's get really uncomfortable for a minute.

Kasim: 00:05:09

Google knows who's in Narcotics Anonymous, who's cheating on their

Kasim: 00:05:11

spouse, who's about to declare bankruptcy, who's gonna get a divorce?

Kasim: 00:05:15

What kind of porn you watch, like Google knows.

Kasim: 00:05:18

Everything and knows everything about you, which culminates in

Kasim: 00:05:21

my favorite case study, which was published in April of 2015.

Kasim: 00:05:25

Google told me I'm pregnant.

Kasim: 00:05:27

Google told a woman she was pregnant before she knew she was

Kasim: 00:05:30

pregnant based solely off of her search and communication patterns.

Kasim: 00:05:33

Moore's Law says that machine computing power doubles every

Kasim: 00:05:36

18 months, so 2015 to today.

Kasim: 00:05:40

What, like, oh my, what could Google possibly know about us now?

Kasim: 00:05:44

And this is the point that I'm trying to make, and this is where

Kasim: 00:05:46

I think Google wins the AI battle.

Kasim: 00:05:47

AI this is where I, do everybody disservice by through and via

Kasim: 00:05:52

my ignorance, but the way that I understand it, and I've gone To school

Kasim: 00:05:55

as much as a person of my limited capabilities can go to school in ai.

Kasim: 00:05:59

And what's really interesting is, you know, I'm watching, like every interview

Kasim: 00:06:01

Sam Altman has ever done and I've read every book every peer reviewed

Kasim: 00:06:04

article that I can get my hands on.

Kasim: 00:06:06

And what's really interesting about it is the AI developers don't know how AI works.

Kasim: 00:06:10

they've said that overtly, and enough of them have said that to where I think

Kasim: 00:06:13

that's actually an accepted truth.

Kasim: 00:06:15

What they've done is they've built a neural network, a

Kasim: 00:06:18

digitized neural network.

Kasim: 00:06:20

And then they offer it inputs with trained outputs, and then they

Kasim: 00:06:24

allow the neural network to learn.

Kasim: 00:06:27

So the simplest, most benign example that was offered up in an interview, and I

Kasim: 00:06:30

wish I could, cite this, I gotta be better about cataloging all, learning I'm doing.

Kasim: 00:06:33

But it was a great, great, great example.

Kasim: 00:06:35

Is we showed Google a bunch of images of men and women

Kasim: 00:06:38

and or not Google, excuse me.

Kasim: 00:06:39

We showed the AI mechanism a bunch of images of men and women,

Kasim: 00:06:41

although this actually might have been Google's barred in terms of

Kasim: 00:06:44

the way that it was being trained.

Kasim: 00:06:45

I don't remember.

Kasim: 00:06:45

And then a human told bot, which when an image was a man and when an image

Kasim: 00:06:51

was a woman, and over time The machine started to be able to identify men and

Kasim: 00:06:55

women, but using its own knowledge graph.

Kasim: 00:06:59

So it's not a software program that says when somebody has long hair,

Kasim: 00:07:02

there are we, you know what I mean?

Kasim: 00:07:03

Cuz there's so many exceptions to the rule.

Kasim: 00:07:05

Instead, it was here's, you know, hundreds or thousands or tens

Kasim: 00:07:08

of thousands of images and it's, oh man, woman, man, woman, man,

Kasim: 00:07:11

woman, man, man, man, woman, man.

Kasim: 00:07:13

and then the machine pays attention.

Kasim: 00:07:14

And it builds, you know, who knows how many hundreds of thousands of

Kasim: 00:07:18

common denominators, like, cheekbones, jaw structure, the hue of the skin

Kasim: 00:07:21

height, shoulder width, et cetera.

Kasim: 00:07:23

And then it builds its own thought process around.

Kasim: 00:07:27

Assigning value valence.

Kasim: 00:07:30

AI's prowess exists upon the core premise of being trained.

Kasim: 00:07:35

And up until this point, I'd love to be challenged on this, by the

Kasim: 00:07:38

way, so somebody who knows better than me, please challenge on me.

Kasim: 00:07:39

But up until this point, everybody in the AI game.

Kasim: 00:07:42

Talks about, you know, I'm to chat, g p t, midterm, E, everything that

Kasim: 00:07:45

open AI's doing, everything that Microsoft's doing with Microsoft Chat.

Kasim: 00:07:47

Everything that Google's doing with their four public AI products.

Kasim: 00:07:49

Maybe there's more now.

Kasim: 00:07:50

But everything that AI does begins with the premise of having been trained First.

Kasim: 00:07:55

It was mid journey, got into a ton of trouble for this because there's

Kasim: 00:07:57

like hundreds of thousands of images that they use that they didn't

Kasim: 00:07:59

necessarily have the rights to use.

Kasim: 00:08:01

so then the question, the ethical question becomes, well, you're not.

Kasim: 00:08:03

Plagiarizing that image per se, but you're plugging these images into a

Kasim: 00:08:07

network that's using these images to learn how to create subsequent images.

Kasim: 00:08:10

So where does IP begin an end?

Kasim: 00:08:11

And if that's the case to be made, and I think it's a cogent case, that

Kasim: 00:08:15

means that the information that's being used to train the AI is at

Kasim: 00:08:20

least as of now the most important.

Kasim: 00:08:23

Variable

Kasim: 00:08:24

and Google has, I think, without question better information and more information

Kasim: 00:08:30

on everybody, and they've been collecting it for decades, which is crazy.

Kasim: 00:08:35

That makes me feel so old to say.

Kasim: 00:08:36

So a couple of.

Kasim: 00:08:37

Close runners up Apple.

Kasim: 00:08:39

If you think about what Apple has from an infrastructure standpoint

Kasim: 00:08:42

they have a ton of data on you.

Kasim: 00:08:44

Potentially Amazon, because of Amazon Web Services, Amazon e-comm for sure,

Kasim: 00:08:49

but that's, kind of like a Disinvited app, Amazon Web Services and everything

Kasim: 00:08:52

that it hosts, all that data ends up.

Kasim: 00:08:54

Being accessible to them kind of via backdoor.

Kasim: 00:08:57

And I don't know what that means and how much they can see or catalog,

Kasim: 00:09:00

but I'm sure it's all of it.

Kasim: 00:09:00

Right?

Kasim: 00:09:01

Like why wouldn't it be?

Kasim: 00:09:01

So, open ai, their training has been on open source libraries.

Kasim: 00:09:05

Basically it's, like the Wikipedias of the world.

Kasim: 00:09:08

When you start getting into proprietary AI with proprietary

Kasim: 00:09:11

data sets, nobody's proprietary data set is better than Google's.

Kasim: 00:09:15

you could combine every government on the planet and all the Big Brothers

Kasim: 00:09:18

Secret Watch programs, and the fact that the US government is, cataloging

Kasim: 00:09:21

every communication you and I have ever had, nobody's done it as well.

Kasim: 00:09:26

Or graph it as well as Google has.

Kasim: 00:09:27

And you know that because you see the way that Google ads works and because

Kasim: 00:09:31

of the way that Google Ads works, it intrinsically qualifies this storage

Kasim: 00:09:34

mechanism that Google has because they've, well, you've seen it in practice.

Kasim: 00:09:36

So in my mind, there's no way doesn't end up being the front runner for ai.

Kasim: 00:09:42

And, they didn't release first for the stated reason was And I believe

Kasim: 00:09:46

this, by the way, was they wanted to be cautious about reputation, basically,

Kasim: 00:09:52

is the way that I'm summarizing.

Kasim: 00:09:53

It was a little bit more complex than that.

Kasim: 00:09:55

But they didn't wanna take all the AI arrows.

Kasim: 00:09:56

And I think that that was a really intelligent decision.

Kasim: 00:09:59

Not being first to market was a really intelligent decision in this context,

Kasim: 00:10:02

because they had way more to lose than what's, effectively a nonprofit

Kasim: 00:10:06

that's released, an app that's.

Kasim: 00:10:08

Iterative in nature.

Kasim: 00:10:09

the folks behind chat, G P T could afford to be, they're young and

Kasim: 00:10:14

nimble so they could play that card.

Kasim: 00:10:16

Google had way more to lose, so it looks like Google's behind.

Kasim: 00:10:20

But man, I don't think that's the case.

Kasim: 00:10:21

And not only is that not the case, they have the building

Kasim: 00:10:24

blocks necessary to produce.

Kasim: 00:10:26

thing that I haven't said that I should have said this entire time I've

Kasim: 00:10:28

been focusing on what Google knows about you thing, about what Google

Kasim: 00:10:31

just knows in general, like Google has cataloged the internet better and

Kasim: 00:10:34

more effectively than anybody else has ever cataloged the internet, the

Kasim: 00:10:37

lexicon of the world's knowledge.

Kasim: 00:10:39

our intellectual library humanities.

Kasim: 00:10:40

Brain.

Kasim: 00:10:41

That's Google's claim to fame.

Kasim: 00:10:43

That's their number one unique selling proposition is they've

Kasim: 00:10:46

grafted knowledge better than anybody.

Kasim: 00:10:48

So I don't see a future where Google isn't a, and probably

Kasim: 00:10:53

the front runner in the race.

Kasim: 00:10:56

That is ai.

Kasim: 00:10:57

And I'd love to be challenged on that.

Kasim: 00:10:59

I'm not saying that in a perpetuate way, at least I hope I'm not.

Kasim: 00:11:02

I'd love to know if there's something, I don't know, like is there somebody

Kasim: 00:11:05

out there that has a better or at least competitive database of information?

Kasim: 00:11:09

I can't possibly think of anyone, you know?

Kasim: 00:11:12

Period.

Kasim: 00:11:12

Full stop.

Kasim: 00:11:13

So keep your eyes on Google for sure.

Kasim: 00:11:16

Cuz I think they, win the arms race on a long enough timeline.

Kasim: 00:11:19

I'd love to know what you think.

Kasim: 00:11:20

I appreciate you watching.