0:00

It's time for Twitter this

0:02

week in tech. Lisa Schmeiser

0:04

is here from No Jitter,

0:06

Daniel Rubino, from Windows Central,

0:08

Attorney Kathy Gellis. We just

0:10

learned the United Kingdom is

0:12

asking Apple for a back

0:14

door into its end-to-end encryption.

0:16

What will Apple do? We're

0:19

going to talk about the

0:21

Cosma bill, which is a

0:23

plan by Congress, to ban

0:25

social media for all people

0:27

under 13. And we'll talk

0:29

a little bit

0:31

about what Doge

0:34

is doing and

0:36

whether it's legal,

0:38

that and a

0:41

whole lot more

0:43

coming up next

0:46

on Twitter. Podcasts

0:48

You Love. From

0:51

People You Trust.

0:53

This is Twit.

0:56

2025. Self-driving government.

0:58

It's time for Twit this

1:01

week at Tech. The show will

1:03

we cover the weeks. Tech News.

1:05

Doing it a little early this

1:07

week. Apologies to people who

1:09

like to watch the stream

1:11

live. I hope you got

1:13

the message. The memo arrived

1:15

because there's apparently a football

1:18

game on later in the

1:20

day. Alicia Schminzer. Are you

1:22

all excited about the big

1:24

Super Bowl? I like it better

1:26

when the 49ers play. And the reason

1:29

I like it better is because the

1:31

roads empty out and I can go

1:33

anywhere in the Bay Area without

1:35

traffic. It's true. This one is

1:37

going to be so low rated

1:39

I think because it's a replay

1:41

of two years ago that I

1:43

think your hike is off. Editor

1:45

in Chief at No Jitter, it's

1:47

great to see you. And I

1:49

see more and more people using

1:51

the Blue Sky handle nowadays Elshmeiser

1:54

at the Blue Skyiser. Hello, good

1:56

to see you. God your hair looks

1:58

good. Are you have you been? What's

2:00

your conditioning routine

2:02

like? I don't know. It is

2:05

a complicated... Is it? Well, you

2:07

know, you gotta have a regime,

2:10

right? You know, I just found

2:12

out you can get a shampoo

2:14

in a bar like soap. Oh.

2:17

And it saves plastic. I

2:19

know. Just a little

2:21

bit of guys. It's

2:23

like, they make shampoo

2:25

conditioner and body wash.

2:27

5 and 1 baby. Mine was

2:29

at two pace. Irish Spring

2:32

5 and 1, it cleans

2:34

everything. Including grout. Good to

2:36

see you Daniel. Thank you

2:38

for being here. Also our

2:40

attorney at large, IP attorney,

2:43

a contributor to Tech

2:45

Dirt. You probably room

2:47

reading her. Somewhat upset?

2:49

What's the word? Post

2:51

on Tech Dirt Critical?

2:53

See, see, see, council.com, great to see

2:56

you, Kathy go. It's always a

2:58

question. Now I realize, C-G-C-O-U-N-S-E-L.

3:00

Oh, we left out a

3:02

C. In the letter. She's not

3:04

a counselor. She's not a counselor.

3:07

She's not, she's a counselor. That's

3:09

right. I also recommend never get

3:11

a domain name that is like,

3:13

oh, I forget if it's homin

3:16

or homophone, but it was a

3:18

bad idea. But it was a

3:20

bad idea. You have to get

3:22

all the other spellings, C-O-U-N, C-I-L,

3:24

and all of that. M-O-U-S-E-O-N, that

3:26

too. Big story today. There

3:28

are a lot of big

3:30

stories today, actually. We've got

3:33

a jammed show, so we're

3:35

going to jam along. UK

3:37

has ordered Apple to give

3:39

it access to encrypted cloud

3:42

data. This is from the

3:44

Financial Times. The British government

3:46

has ordered Apple to grant

3:49

secret access. That's the interesting

3:51

part. How do we know this? I think

3:53

there was a leaker to its

3:56

customers' encrypted cloud storage. This is

3:58

from the Snoopers Charter, which was

4:00

a bill passed the UK investigatory

4:02

powers act last month Apple received

4:04

a quote technical capability notice requiring

4:06

them to create a backdoor to

4:09

a iCloud now this was not

4:11

the government has not said this

4:13

apple has not said this they're

4:15

not allowed to it's one of

4:17

those according to people familiar with

4:19

the matter the move would enable

4:22

UK well You know, I started

4:24

saying UK, but it isn't just

4:26

UK. The move would enable global

4:28

law enforcement and security services. Everybody

4:30

would be affected. It's extraterritorial, which

4:33

means UK law enforcement could access

4:35

the encrypted data of Apple customers

4:37

anywhere in the world, including in

4:39

the United States. We, you know

4:41

what? This has been going on

4:43

for years. The FBI's been asking

4:46

Apple to do it. Apple said

4:48

no. Australia's tried this, but this

4:50

is the first time a nation-state

4:52

is actually, as far as I

4:54

know, told Apple explicitly, we need

4:56

a back door. I feel like

4:59

it had happened before a number

5:01

of years ago, and I thought

5:03

that it was kind of put

5:05

to rest back then, so I

5:07

was surprised to see this news

5:10

where it just kind of came

5:12

to the fore of like, oh

5:14

yeah, we're revisiting this issue and

5:16

ignoring every reason anyone ever told

5:18

us for why it was a

5:20

terrible idea. So, was there a

5:23

big deal about this with the

5:25

San Bernardino shooter? Yeah, and Apple

5:27

said no to the FBI. Yeah,

5:29

yeah. Go ahead, Daniel. I'll just

5:31

say, I think they used to

5:34

have like kind of like holes

5:36

in their operating system where a

5:38

third party software was able to

5:40

access this stuff and then Apple

5:42

patched all that and like, yeah,

5:44

well, so Apple has in the

5:47

past been able to do this.

5:49

In fact, they told the FBI

5:51

in the case of the San

5:53

Bernardino terrorists. that oh if you

5:55

had just taken his phone home

5:57

it would have uploaded the contents

6:00

to iCloud and we could have

6:02

given it to you. So as

6:04

recently as whenever that was eight

6:06

years ago, ten years ago, Apple

6:08

would and did do that. They've

6:11

done it in other situations as

6:13

well. Trump's campaign chief

6:15

was using WhatsApp, which

6:17

was encrypted to

6:19

message. This is in the

6:21

2016 election. But WhatsApp

6:24

uploaded unencrypted text to the

6:26

Apple-i-cloud. and the government got

6:29

it. We know that that

6:31

happened because if you read

6:33

the indictment, they quoted his

6:36

what's app messages from iCloud.

6:38

But Apple patched, they did, they

6:40

patched that hole. They fully encrypted,

6:42

as far as we know, they

6:44

fully encrypted everything. In order for

6:46

iCloud to work, they have to have a

6:48

key because, you know, for a

6:50

variety of reasons. If you have

6:52

fully encrypted cloud storage, there's a

6:54

lot of abilities that go away. which

6:56

is why most cloud storage companies,

6:58

including I think Microsoft, offer, correct

7:01

me if I'm wrong, Daniel, kind

7:03

of an enclave, an encrypted enclave,

7:05

and then the rest of it,

7:07

well, you know, we have the

7:09

keys to it, isn't that right? Yeah,

7:11

it comes like vaults. Yeah, yeah. Yeah.

7:13

Apple does not offer that. Apple just

7:15

says we have the keys, but we're

7:17

not going to give them to anyone. This

7:20

is a huge issue, as we know, from

7:22

back doors in the past. It's not

7:24

been so long ago that

7:27

it was revealed that salt

7:29

typhoon Chinese hackers had access

7:31

to the phones of many

7:33

officials in government high officials

7:35

in government because Going way

7:38

back to the 90s the

7:40

government mandated that there would

7:42

be wiretap capability and digital

7:45

technologies At the time

7:47

the FBI director said no, but

7:49

it but don't worry because this

7:51

backdoor will never leak out well

7:53

And as a result, we've got Chinese

7:55

hackers in our phone network and can't

7:58

do much about it, incidentally. So

8:00

this is always a bad idea.

8:02

I think every real security guru

8:05

says there's no way back doors

8:07

stay secret. It's all fascinating with

8:09

the aspect. Girls this week, they're

8:11

having a bad week. It's a

8:14

bad week for security in general.

8:16

Yeah. Go ahead Daniel. Sorry, I

8:18

was going to say that. You

8:21

know, what's also scary about this

8:23

is the fact that I mean

8:25

mentioned this earlier that, you know,

8:27

if. they do this, then they

8:30

can access data from Americans indirectly,

8:32

right? And this is what the

8:34

NSA has always done. NSA can't

8:37

spy. And with the Patriot Act

8:39

can't spy directly on Americans. But

8:41

if they're spying on someone else

8:43

in another country who happens to

8:46

have access to Americans, that data

8:48

gets collected as well. And so

8:50

it's an indirect way. So I

8:53

wouldn't even be surprised if the

8:55

US government is like. Shrugging their

8:57

shoulders like, you know, we're not

8:59

saying you guys shouldn't do this

9:02

because it would probably help them

9:04

as well They wanted the whole

9:06

eyes, you know A consultant advising

9:09

the United States and encryption matters

9:11

according to the Washington Post who

9:13

broke this story said Apple would

9:15

be barred from warning its users

9:18

If this happens Apple You know,

9:20

had rolled out, and this is

9:22

what started this, a couple of

9:25

years ago, advanced data protection. A

9:27

switch you could turn on, most

9:29

people don't, because it eliminates some

9:31

of, as I said, some of

9:34

the features of iCloud and other

9:36

things, but they offered it, and

9:38

they specifically encouraged government officials to

9:41

use it, because they didn't even

9:43

have the keys to this. This

9:45

was fully encrypted. trying to do

9:47

this a couple years earlier backed

9:50

off after according to the post

9:52

after objections from the SBI this

9:54

is back in the first term

9:57

of president Trump The service is

9:59

an available security option for Apple

10:01

users in the US and elsewhere

10:03

I'm not sure if it's available

10:06

in the UK Probably what will

10:08

happen is Apple will make it

10:10

no longer available in the UK

10:13

But that doesn't solve the problem

10:15

in the US Ron Wyden, Senator

10:17

Ron Wyden of Oregon, a Democrat

10:19

on the Senate Intelligence Committee, said

10:22

the US has to dissuade Britain.

10:24

Trump and American tech companies letting

10:26

foreign governments secretly spy on Americans

10:29

would be unconscionable and unmitigated disaster

10:31

for America. I know why you're

10:33

laughing. I know why you're laughing.

10:35

I know why you're laughing. New

10:38

what he was saying. Unmitigated disaster

10:40

for Americans privacy international security. Signal

10:42

President Meredith Whittaker said using technical

10:45

capability notices to weaken encryption around

10:47

the globe is a shocking move

10:49

that will position the UK as

10:51

a tech pariah rather than a

10:54

tech leader if they implemented the

10:56

directive will create a dangerous cybersecurity

10:58

vulnerability in the nervous system over

11:01

a global economy. You know why

11:03

Meredith Whittaker is commenting on this

11:05

because next... It's signal. Because signal

11:07

there's no back door and signal.

11:10

The UK's obviously. Yeah, she's fantastic.

11:12

She writes a lot of great

11:14

stuff on this topic. Yeah. This,

11:17

you know, given all the news

11:19

coming out of DC, this will

11:21

probably not get the coverage it

11:23

should. But if you're watching our

11:26

shows, you understand why this is

11:28

such a big deal. But I

11:30

think it's up for Apple to

11:33

somehow figure out how to resist

11:35

on its own. We don't have

11:37

a foreign policy at the moment.

11:39

We barely have a domestic policy

11:42

at the moment. So there's, the

11:44

conventional wisdom about how we would

11:46

approach this is not available to

11:49

us. This is not something where

11:51

you get together, you speak to

11:53

your government, and you sort of

11:55

point out like, okay, I know

11:58

you're tempted, but here's the consequences

12:00

of. if this happens. This is

12:02

not a government that is equipped

12:05

to consider the consequences of decisions

12:07

that it makes. And at this

12:09

point, given how much power is

12:11

embodied in any particular administration and

12:14

getting exercised in any particular way,

12:16

it's probably all for incursions for,

12:18

you know, privacy, it seems to

12:21

be the administration's position in an

12:23

awful lot of technology areas these

12:25

days. Android is also at this

12:27

point end to end encrypted in

12:30

its backups. It does not offer,

12:32

it does not have the keys.

12:34

So, Remember that this came out

12:37

because of a leak. It's very

12:39

likely that the same order went

12:41

to Google and went to signal

12:43

and went to meta for WhatsApp.

12:46

Because, I mean, why would they

12:48

just target Apple? Google's been encrypting

12:50

backups for Android phones since 2018.

12:53

Google declined. This is interesting. Google

12:55

declined to say whether any government

12:57

had sought a backdoor. Google is

12:59

turning to such a like. Weird

13:02

company, right? They were like the

13:04

good guys for so long and

13:06

then of course they had that,

13:09

you know, we don't do evil

13:11

thing and they got rid of

13:13

that and then they were like

13:15

And now this week they're like

13:18

we think it's fine to use

13:20

AI to make super smart weapons

13:22

Yeah, yeah, yeah, yeah, that's a

13:25

line of revenue They got rid

13:27

of DEA, well maybe a little

13:29

evil now we're all about the

13:31

evil Defined evil? Well, to be

13:34

fair, getting rid of DEAI seems

13:36

to be the thing. It's interesting

13:38

because it's in anticipation of trouble

13:41

from Washington. Nobody's told them to

13:43

do that yet. But Washington's in

13:45

advance. It's missing the states. So

13:47

what I'd really like to gently

13:50

push back on is instead of

13:52

calling it DEA, which reduces it

13:54

to an acronym and a buzzword,

13:57

let's just point out that what

13:59

companies are companies are. Volunteer. doing

14:01

right now is downplaying or limiting

14:03

efforts to make the workplace more

14:06

diverse, more equitable, and more inclusive.

14:08

That's it. Companies are explicitly saying

14:10

we no longer value diversity, we

14:13

no longer value equity, we no

14:15

longer value inclusion. Like those

14:17

are the words, let's use the words,

14:19

not the acronym. No, I agree with

14:21

you, but why, and this is a

14:24

softball, why is it important that the

14:26

company like Google be diverse? McKenzie has

14:28

done tons. Okay, sorry, I'm gonna pop off

14:30

for a minute and I hope that's okay,

14:32

everybody. That's why I asked the softball

14:35

question. Kathy and Daniel. McKenzie

14:37

has nearly a decade of

14:39

research examining the performance of

14:41

publicly traded companies when you

14:43

have a diverse workforce, especially

14:45

diverse leadership from management up

14:47

to sea level and board

14:49

of directors, compared to companies

14:51

with a homogenous culture. And

14:53

the difference in profit is

14:55

in the double digits percentage,

14:57

as is the difference. and

14:59

employer retention and the difference

15:01

in overall productivity. It's actually

15:03

in the best interest of any

15:05

shareholder to demand the EI because

15:07

it boosts overall company performance and

15:10

the bottom line. Why somebody would

15:12

attack diversity equity and inclusion and

15:14

say we would prefer to take

15:17

these out of our company, you're

15:19

basically saying we'd like a company

15:21

that's less profitable. And I think

15:24

if we're going to talk about

15:26

this, we need to talk about

15:28

McKenzie's research. There's been research

15:30

from Pew, there's been research from

15:33

other institutes focused on taking a

15:35

look at the participation of

15:37

populations in the workforce and what

15:40

the net effect in bottom line

15:42

is. And the research, if you do

15:44

a... What is it called when you

15:46

take a look at multiple studies and

15:48

come to this conclusion? Yeah, it's

15:51

fairly unambiguous. Diversity,

15:53

equity, and inclusion

15:55

brings to bear a wide variety

15:58

of talents, of strength. of

16:00

viewpoints of perception of market opportunities that

16:02

homogenous populations simply don't have. It's better

16:04

for business, it's better for society, it's

16:06

a win-win. And that's why it matters.

16:08

And our discord says that this administration

16:10

is exchanging DEA for homogeneity, exclusion, and

16:12

inequity in labor, which interestingly is Hyle.

16:14

Someone, I wonder if that's pointed, satirical

16:16

commentary on it. But somebody also was

16:19

saying how no, D.I. will just have

16:21

merit. And I think what we need

16:23

to push back on is this idea

16:25

that D.I. was foreclosing merit. D.I. was

16:27

a way of making sure you captured

16:29

merit. Yeah, so that's the point they're

16:31

making. You should promote people, you should

16:33

hire people based on ability alone. Talent.

16:35

Yes, that's how you get it by

16:37

making sure that you're you're tapping into

16:40

it from from problem. I would also

16:42

raise the point that if you're going

16:44

to develop AI and you do it

16:46

only with white men, you're going to

16:48

have a problem that if you make

16:50

games and you only have white men

16:52

writing them. Well, we already know what

16:54

problems you have. In fact, Gamergate has

16:56

now won. It's seven years later and

16:58

Gate won. Well, you take a look.

17:01

I can remember going to CES a

17:03

couple years. And by the way, none

17:05

of the companies that I'm going to

17:07

mention, I'm not going to mention names.

17:09

They're not in business. I've checked, but

17:11

I can remember going and taking a

17:13

look at just a curiosity or somebody

17:15

who's like, I have a smart sensor

17:17

for a diaper. So when you attach

17:19

it to a diaper, an alert goes

17:21

off on your app. and this way

17:24

you know changed the baby which because

17:26

you're in the other room in another

17:28

state you're out of town what first

17:30

of all babies are not hesitant about

17:32

letting you know in the day trust

17:34

me you know I said so does

17:36

this sensor send an alert to every

17:38

caregiver responsible for the baby and they

17:40

said no it's a closed ecosystem so

17:42

it's one sensor one app, one phone,

17:45

if you want multiple users, you have

17:47

to attach multiple sensors. And I was

17:49

like, Ben in a household that functions

17:51

with people. Is that baby walking funny?

17:53

Well, she's got four sensors in her

17:55

door. Yeah, no, this guy thought this

17:57

was a reasonably, perfect, prominent solution. Guy

17:59

is the word. And I was like,

18:01

you don't have kids, do you, or

18:03

a partner, or and. Have you ever

18:06

changed a diaper in your life? Well,

18:08

it was a lot of personal technologies

18:10

where it solved a very specific problem

18:12

for a very specific demographic, or there

18:14

was somebody who perceived a market need,

18:16

but because they didn't have lived experience

18:18

or access to people lived experience, they

18:20

went ahead and wasted time and money

18:22

and effort, creating something that was completely

18:24

unusable for a target audience. Amazing. This

18:27

is a core problem when you have

18:29

a homogenous workforce. That's a great example.

18:31

they won't be able to perceive anything

18:33

outside their experience and they won't even

18:35

know how to get outside their experience

18:37

after a while. I want to make

18:39

sure that I get out the whole

18:41

idea about why merit and DII are

18:43

synonymous and not paradoxical. Because somebody is

18:45

suggesting, well, skin pure pigment has nothing

18:47

to do with merit. Well, A, D,

18:50

I has more to do than just

18:52

skin pigment. But sure, of course not.

18:54

None of these qualities that D, I

18:56

make sure, are mixed up in our

18:58

company so that you have all sorts

19:00

of people from all sorts of walks

19:02

of life, physiology, culture. religion, etc. The

19:04

fact that they can all come together

19:06

and bring their contributions is important because

19:08

it also produces as Lisa was saying,

19:11

a more powerful company because it knows

19:13

how it has a bigger bag of

19:15

tricks to pull from to be able

19:17

to produce to do its business. But

19:19

also the reason why these efforts are

19:21

important is because there are structural issues

19:23

that prevent some of these populations from

19:25

getting in the door to be able

19:27

to make those contributions. And DEA I

19:29

was about removing those barriers to make

19:32

sure that those people could be included.

19:34

because we need

19:36

them to be included

19:38

because it will

19:40

have better business. Well,

19:42

I mean, selfishly,

19:44

the businesses will make

19:46

more money if

19:48

they have access to

19:50

more people with

19:53

more ideas. And there's

19:55

all sorts of

19:57

things. When there's things

19:59

interfering with that,

20:01

it's good to knock

20:03

down the things

20:05

that are interfering with

20:07

it, just because

20:09

we want a better

20:11

culture in country

20:13

that might be more

20:16

mixed up and

20:18

equal. What is the

20:20

percentage of women

20:22

CEOs in this country?

20:25

I don't know, but not high. It's very

20:27

low. Single digit. So

20:30

the other thing, some have pointed out and I

20:32

agree with them, is that a lot of the

20:34

companies that had so -called DEI initiatives were just,

20:36

it was lip service. Basically,

20:38

companies are companies, right? They're not, they

20:40

don't have a heart. They don't

20:42

have a conscience. They're there to maximize

20:44

profits. That's almost their responsibility, their

20:47

corporate responsibility. So when the wind blows

20:49

one way and everybody says, oh,

20:51

you got to do DEI, they write

20:53

up a thing and they put

20:55

it in their mission statement and done.

20:57

They find the stock art where

20:59

it's smiling people of different colors already.

21:01

Yeah. Every front page of every

21:04

website has a black person, a white

21:06

person, a woman, and a Chinese

21:08

person. It's just what you do, but

21:10

is it lip service or is

21:12

it genuine? I think in Google's case,

21:14

it was somewhat genuine. I think

21:16

in a lot of the companies, was

21:18

somewhat genuine. Yeah. Yeah. I mean,

21:20

even in Turkey's case, it was genuine.

21:23

Like they had programs. They had

21:25

shirts. They they did LGBT outreach. Like

21:27

they understood one of the other

21:29

things about being a business that values

21:31

DEI, it just is values and

21:33

not just an acronym is because they're

21:35

your customers and they've got money

21:37

and they can give you the money

21:39

if they think that you're selling

21:42

something worth them giving you money for.

21:44

I remember Martin Luther King Jr.'s

21:46

son on TV saying, well, if you're

21:48

not to include us, we're not

21:50

to include you. And I think that's

21:52

a very important point too. A

21:54

company that very kind of publicly says,

21:56

we're going to roll back our

21:59

hiring goals. And this is what Google

22:01

did. on Tuesday they removed the language in their 10K to the

22:03

SEC which said quote we are and this is all they're

22:05

saying we are convert committed to making

22:07

diversity equity inclusion part of everything we

22:10

do and growing a workforce that is

22:12

representative of the users we serve that

22:14

seems pretty anodyne it doesn't seem

22:16

no because that tells you who they want to

22:19

that tells you who their target audience the users

22:21

are I think you need to pay attention to

22:23

that phrase. So, okay, that's interesting.

22:25

By the way, Google's efforts did

22:28

increase the number of women, the

22:30

number of people of color working

22:33

at Google by a few percentage

22:35

points. It didn't, it didn't overnight

22:37

make it equitable. I think one

22:40

of the things you're going to have

22:42

to look for in the next year

22:44

or two, they've had a return to

22:46

office push too, haven't they? Yeah. One

22:48

of the emerging trends

22:51

in workplace research, Slack

22:53

has shown this, I wish

22:55

I could pull the names

22:57

of the vendors off the

23:00

top of my head, but

23:02

something that's coming up,

23:04

as people take a look

23:07

at the impact of return

23:09

to office policies, is

23:11

the impact disproportionately

23:14

affects women,

23:16

parents, There

23:19

is a suggestion that these policies

23:21

are discouraging equitable participation

23:24

in the workplace because

23:26

as they come back to the office, it's

23:28

not as productive or merit-based environment

23:31

as it was when you were

23:33

purely producing in a collaborative remote

23:35

environment. there's external social stressors, especially

23:37

if you're a parent and you're

23:39

responsible for child care, things like

23:42

that. So this is something I'd

23:44

be very, very curious to look

23:46

at with Google over the next

23:48

few years is they might be

23:50

hosting on numbers that were boosted

23:53

by previous diversity equity inclusion initiatives

23:55

earlier, but let's see what those

23:57

numbers look like in 2026 and

23:59

2027. after a full year of RTO and

24:01

after a cultural downplaying of

24:04

diversity equity and inclusion initiatives. Because

24:06

I think I can see a

24:08

change. Both Google and Microsoft have

24:10

said hybrids not going anywhere. Yeah.

24:12

But Google does say you got

24:14

to come to work three days

24:16

a week. Yeah. And I think

24:18

Microsoft, is it the same at

24:20

Microsoft Daniel? Do you know? Because remember,

24:22

the Amazon said five days a

24:24

week, Amazon said, you're back in

24:26

the office, you're sitting at that desk.

24:29

Great way to shake people out. Well,

24:31

I wonder how many people they lost,

24:33

whether that, what is the cost of

24:35

having. Well, I think that's why Google

24:37

and Microsoft both. The best people before

24:39

you lose the people want to shake out.

24:42

Yeah, like the great thing about a layoff

24:44

is you can target your cuts, whereas

24:46

when you're trying to shake people by

24:48

attrition, the best people leave. The best

24:50

people say no, I'm not coming back.

24:52

Well, that's the most hireable. Well, that's

24:55

right. That's what I mean. Yeah, I'm

24:57

not sure that that equates, but yes.

24:59

It's the people who can get another

25:01

job where they don't get another job

25:03

where they don't have to work in- When

25:05

you look at, okay, this is a collapse

25:07

of a whole bunch of things that should

25:10

not have collapsed, but let's look at

25:12

where the pushback power is. And

25:14

so one of the places that

25:16

there's pushback power is among consumers

25:18

and the public and enough people outraged

25:21

about it that it can certainly create

25:23

some market force pressure and we should

25:25

look into doing that. But the other

25:28

thing high as a lawyer I want

25:30

to know is these companies who thought it

25:32

was in their interest to do

25:34

it. I don't know if that

25:37

quite plays out because even if

25:39

they thought that their business depended

25:41

on the benevolence and quotes of

25:44

Trump, that's not the only

25:46

authority that governs them. They

25:48

are still exposed to states

25:51

and states have laws. They

25:53

have enforcement powers and they've

25:56

got courthouses that private aggrieved

25:58

people who believe they've been

26:00

discriminated against can still access. So

26:02

all they've done is in trying

26:04

to protect themselves against whatever they

26:07

thought was gonna happen or improve

26:09

their position with respect to the

26:11

Trump administration, they've really undermined their

26:13

position with respect to every other

26:15

power and authority and I don't.

26:17

quite know why they thought that

26:19

the you know when they run

26:21

the math on that that they're

26:24

going to be better off this

26:26

way and I think it's time

26:28

to maybe make them see they

26:30

are not in fact better off

26:32

this way. All right. I had

26:34

with a friend this week because

26:36

they were like oh I'm so

26:38

worried that Apple is still sticking

26:41

to D. I and so is

26:43

Costco and I was like don't

26:45

be it's a numbers game and

26:47

clearly they've done the math they've

26:49

run the math and decided that

26:51

visibly sticking to this stance and

26:53

appealing to a customer base is

26:55

going to work out better for

26:58

them than anything the Trump administration

27:00

could possibly do. Like, as Kathy

27:02

pointed out, there's a cost benefit

27:04

analysis to this, and it is

27:06

a little bit astonishing that a

27:08

lot of these places aren't taking

27:10

a look at how their, how

27:12

their public posturing is going to

27:15

land with different state enforcement or

27:17

even international enforcement, because, I mean,

27:19

the EU is not shy about

27:21

regulation. I have to say there's

27:23

a case to be made that

27:25

EU regulation, especially their anti-tech regulation

27:27

like this UK law, is harmful

27:29

to the EU in terms of

27:32

being an innovator. I think the

27:34

EU is behind. This is a,

27:36

let's go back to England for

27:38

a moment. The British public wants,

27:40

87% of the British public would

27:42

back a law requiring AI developers

27:44

to prove their systems are safe.

27:46

before release, which by the way,

27:49

eliminates every single public AI that's

27:51

out there, because all of them

27:53

have been jailbroken. 60% are in

27:55

favor of outlawing the development of

27:57

smart than human AI models. Good

27:59

luck. Well, good luck

28:01

in developing the models too. I would

28:04

say. Well, we can dispute that. I

28:06

think by the, I think it's pretty

28:08

clear to me that by the end

28:10

of this year, AI coding will be

28:13

equivalent of to the best coders out

28:15

there right now. It will replace coders.

28:17

Coders will become something else. We'll see.

28:20

You know, we'll see. That's that's the

28:22

low-hanging fruit. I think there's a lot

28:24

of things that AI can do better

28:27

than a human. So. Maybe not in

28:29

Britain. Leah, when you launch the AI

28:31

show, feel free to be honest. We

28:33

did. We have. We took this week

28:36

in Google and renamed it Intelligent Machines.

28:38

And every week we're going to have

28:40

a guest from the AI. This week

28:43

we had a guy who was the

28:45

head of sales at Open AI. He

28:47

was very aggressively, aggressively in favor of

28:50

AI and AGI and all of that.

28:52

We're going to have in... couple of

28:54

weeks, Ray Kurzweil, who of course created

28:56

the term intelligent machines on it, has

28:59

been saying the singularity is near for

29:01

decades. You may, I do not disagree

29:03

with him at this point. I think

29:06

we've seen amazing progress, don't you think?

29:08

In a human intelligence. And the way

29:10

the thing about the AI stuff is

29:13

once it gets to the point where

29:15

it can recreate itself. and create better

29:17

AI from AI. That's why an AI

29:19

coder might be the first step towards

29:22

a singularity because then it's right in

29:24

its own code and it can iterate

29:26

much faster. Yeah, the issue with all

29:29

this is China, right? Because. The EU,

29:31

UK, can talk about we need restrictions,

29:33

we need laws, we need to slow

29:36

this down, we've got to monitor it.

29:38

It's fine. Sounds great. The United States

29:40

is a little bit in between there

29:42

because we stand the most to, you

29:45

know, gain from having unrestricted AI. But

29:47

China's just out there doing whatever it

29:49

wants. And that's the problem. If you're

29:52

going to have a system where everybody

29:54

is like, like, we need to have

29:56

rules and regular. And China is like,

29:59

go ahead, I can tell that we're

30:01

just going to go ahead and just

30:03

do our thing. They're going to end

30:05

up winning, whatever winning is here, you

30:08

know, the more advanced AI system, they're

30:10

just going to get there first. And

30:12

it's considered to be like the next

30:15

nuclear arms race, right, where we're going

30:17

to get the best AI. And up

30:19

to this point, the United States has

30:21

had, it's been the leader in tech

30:24

and Europe is far behind. And China

30:26

is wririllingling the United States at the

30:28

United States at this point. they have

30:31

some tricks up their sleeves apparently. Now

30:33

we can debate about how much of

30:35

that was stolen or, you know, some

30:38

smoke and mirrors, but at the same

30:40

time, they created a pretty impressive generative

30:42

AI system, you know, pretty quickly. Yeah.

30:44

Well, that's a point. Any very, very

30:47

attractive destination for researchers at this point,

30:49

since we are currently in the US

30:51

embracing an administration and a culture of

30:54

devaluing education. in research investment. So where

30:56

do you think the brains to develop

30:58

this stuff are going to go? Well,

31:01

no, I'm realizing there is a ceiling

31:03

on that on what AI can do.

31:05

And I really want to push back

31:07

on the term of best AI. You

31:10

know, it could produce certain amounts of

31:12

code that will run sort of efficiently

31:14

in a couple of context, but that

31:17

doesn't make it best if the UI

31:19

is terrible. If it doesn't understand all

31:21

sorts of contextual parameters, if it doesn't

31:24

understand how it ultimately intersect interfaces with

31:26

human behavior. Like you need the humans

31:28

to be able to sort of direct

31:30

it. You can't just kind of have

31:33

it pull. We kind of treated as

31:35

a magic one that just pressed this

31:37

little button and all the problems that

31:40

the human beings couldn't quite solve is

31:42

going to magically be solved by a

31:44

computer and it can do some stuff

31:47

but it can't solve them because where

31:49

is it even going to have learned

31:51

the understanding that billions of people haven't

31:53

still managed to figure out. So I

31:56

think there's a feeling. Who makes the

31:58

most profit from it right now the

32:00

systems aren't very profitable right? No expense

32:03

But that's what's interesting about China, right?

32:05

that's growing in enterprise AI is enterprise

32:07

AI that there's been two years worth

32:10

of dumping in all of this money

32:12

and all these resources into co-pilots and

32:14

AI assistants and automated workflows and now

32:16

so-called agentic AI but no one's buying

32:19

and workers are super resistant to using

32:21

AI so that's that's another part of

32:23

the story too is it's not doing

32:26

what people have been

32:28

promised. If you're not understanding the

32:30

problems to solve, you're not going

32:32

to solve them. And I shouldn't

32:34

say none because there are certain

32:37

API applications that have taken the

32:39

trouble to sort of figure out

32:41

here is the problem and here's

32:43

how AI solves it. I don't

32:45

want to stand in the way

32:47

of those, but that is not.

32:49

the buzz. That is not the

32:51

thing that people are talking about.

32:53

That is not where all the

32:56

money is getting dumped into and

32:58

all the essentially hysteria. We are

33:00

growing the tulip bulbs around AI,

33:02

which is not designed to be

33:04

problem specific. It is designed to

33:06

be magic wand specific. And if

33:08

you think you've conjured one, I

33:10

think you're going to be wrong.

33:13

You're not going to solve the

33:15

problems because you never understood the

33:17

problems you were supposed to be

33:19

solving in the first place. The

33:21

framing LITs are an AI as

33:23

blooms taxonomy of learning where you

33:25

move from simply being able to

33:27

remember and retain a piece of

33:30

knowledge to being able to apply

33:32

that knowledge in contextally appropriate situations

33:34

to doing more cognitively complex things

33:36

which depend on application synthesis analysis

33:38

and. contextual flexibility. And Kathy, I

33:40

think we're in a little bit

33:42

of agreement here. AI doesn't show

33:44

a lot of ability to discern

33:47

and adapt to specific contexts or

33:49

to even understand that there are

33:51

that problems exist in different contexts

33:53

and therefore would have different solutions

33:55

depending. It's super great for for

33:57

highly structured automated tasks or for

33:59

queries where you need to. to

34:01

find patterns and looking at those

34:04

patterns helps you find a solution.

34:06

But with really complex things, I

34:08

don't know that we're to the

34:10

point where you have a set

34:12

of artificial intelligence tools that are

34:14

able to take a look at

34:16

a problem like how do we

34:18

free it more nitrogen for growing

34:21

agriculture again and understand the constituent

34:23

parks, break it down, do the

34:25

brute force computing and then put

34:27

it all together. And how do

34:29

you abdicate so much control over

34:31

our world to something that? doesn't

34:33

get hungry, doesn't get sad, doesn't

34:35

grieve, doesn't love, doesn't have anything

34:38

that is part of the human

34:40

existence. It's going to be creating

34:42

solutions where the solution may be

34:44

offing like the entire population, but

34:46

human beings in theory would react

34:48

to that and say we've got

34:50

a problem here. It's interesting. That

34:52

goes full circle back to our

34:55

DEAI, except at this time, we're

34:57

going to have to tell AIs

34:59

to start including us. I need

35:01

to take a break. We have

35:03

a great panel and lots to

35:05

talk about. Lisa Schmeiser is here.

35:07

It's so nice to have you

35:09

from No jitter.com, where she's the

35:12

editor-in-chief. Windows central.com is where you'll

35:14

find Daniel Rubino. Also, editor-in-chief. Good

35:16

to see you. Off the off

35:18

the rails this week. I mean,

35:20

no, I don't mean what's a

35:22

good way to say that you

35:24

have just been knocking it out

35:26

of the park Talking about what's

35:29

going on with Doge and we'll

35:31

we'll get to that in a

35:33

little bit lots more to talk

35:35

about You're watching this week in

35:37

tech our show today brought to

35:39

you by threat locker love these

35:41

guys because they are solving a

35:43

massive issue with security imagine hardening

35:46

your security And never have to

35:48

worry about zero day exploits or

35:50

supply chain attacks or ransom wherever

35:52

again. That's what threat locker can

35:54

do. Worldwide companies like JetBlue trust

35:56

threat locker to secure their data

35:58

and keep their business operating. operations

36:00

flying high, if you will. Now, look,

36:02

every security company says,

36:05

oh, you never have to worry

36:07

about threats again. What makes

36:09

threat locker different? It's something

36:11

called zero trust. And it

36:14

really, really works. Imagine taking

36:16

a proactive, and this is

36:18

the key, these three words,

36:20

deny by default approach to

36:22

cybersecurity. What does that mean? It

36:24

means every action is blocked, every

36:27

process is blocked, every user.

36:29

is blocked unless explicitly authorized

36:31

by your team. It means

36:33

nobody can get inside your network

36:35

and do whatever they want. They

36:37

have to be approved. Threatlocker helps

36:39

you do it and this is so

36:42

important nowadays gives you a full

36:44

audit of every action. That's great

36:46

of course for compliance but also

36:48

helps you with risk management. Because

36:50

if something does happen you know

36:52

exactly where, when and how.

36:54

And Threatlockers 24-7 U. U.S. base

36:56

support team fully supports you in getting...

36:59

The whole thing set up and onboarding

37:01

and beyond. This is so key, you can

37:03

stop the exploitation of

37:05

trusted applications within your

37:07

organization. Keep your business

37:09

secure, keep it secured

37:11

from ransomware. Organizations across any

37:14

industry will benefit from

37:16

threat lockers ring fencing,

37:18

isolates critical and trusted

37:20

applications from unintended uses

37:22

or weaponization, limits attackers

37:24

lateral movement within the

37:26

network. And by the way... Works

37:28

on heterogeneous networks as well. It

37:30

works for Max and PCs. And

37:32

it's very affordable. That's the other thing.

37:35

You might say, well, that sounds

37:37

like big enterprise, big iron. No. Every

37:39

business needs threat locker. Get

37:41

unprecedented visibility and control of

37:43

your cyber security quickly, easily,

37:46

and extremely cost-effectively. With threat

37:48

lockers, zero trust, end point

37:50

protection platform. You need this.

37:52

Visit threat locker.com. You get

37:55

a 30-day trial and... Learn

37:57

more how threat locker can

37:59

mitigate. Even Zero days, even

38:01

completely unknown threats. And help you

38:03

with compliance. threatlocker.com. We really like

38:05

these guys. And I think you

38:08

need to visit threatlocker.com. And if

38:10

you're going to Zero Trust World,

38:12

I wish you have a great

38:14

time in Orlando this week. I

38:17

wish I could be there. They're

38:19

putting on a big conference with

38:21

lots of great and interesting stuff.

38:23

threatlocker.com. You find out more about

38:26

it at Zero Trust World. Threatlocker.

38:28

All right, thank you threat locker

38:30

for supporting Twitter Super Bowl later

38:32

today I Thought this was really

38:34

interesting. I tried to sign up

38:37

so as you may or may

38:39

not know only 38 states allow

38:41

sport gambling betting On sports we

38:43

had a referendum in California and

38:46

it failed I thought it would

38:48

win for sure because everybody likes

38:50

to bet on games And if

38:52

you see the ads on the

38:55

Super Bowl today and every football

38:57

game, every sporting event, it looks

38:59

like draft Kings is legal everywhere.

39:01

Not exactly. You have to go

39:03

to Las Vegas and I think

39:06

it was actually Las Vegas that

39:08

got it overturned in California. They

39:10

want you to come there. Well,

39:12

the crypto folks have found a

39:15

way around it. I don't know

39:17

how legal this is. crypto.com and

39:19

Calci. have both done an end

39:21

around on regulations to allow you

39:24

to bet on the Super Bowl.

39:26

Because they call it a trade.

39:28

Robin Hood thought about doing this

39:30

and decided not to. crypto.com, a

39:33

Singapore-based company, was inviting users in

39:35

the US to quote trade their

39:37

own prediction. No, not the word

39:39

betting is never used on sports

39:41

events, including who win the Super

39:44

Bowl. So what happens is. crypto.com

39:46

has effectively created a contract, a

39:48

swap contract. There's a market for

39:50

yes or no positions in the...

39:53

of the NFL playoffs, college football

39:55

bowl games. For every yes, there's

39:57

a corresponding no. Prices constantly move.

39:59

Sounds a little bit like betting

40:02

in January, for example, a yes

40:04

for the Kansas City Chiefs to

40:06

win this afternoon Super Bowl cost

40:08

$56.75, but if they do, you

40:10

pay $100. The no side, $46.75.

40:13

And you get $100. If the

40:15

Eagles win. And crypto.com gets the

40:17

VIG. I'm sorry, that sounds like

40:19

a betting term. They get $3.50.

40:22

Fees. Fees, it's just fees. I

40:24

have two comments to make. One,

40:26

a lot of these, oh, make

40:28

all this money or do this

40:31

one thing with this one correct

40:33

trip. Like all these people keep

40:35

like thinking they've discovered something and

40:37

it turns out to be securities

40:39

fraud. Like. Not anymore, not anymore,

40:42

not anymore. We got rid of

40:44

Gary Gensler. I don't think the

40:46

SEC is going to get involved.

40:48

I don't know. This is what

40:51

they're doing, right? They started this

40:53

in December right after the election.

40:55

Right, well, so maybe nobody enforces

40:57

it, but basically there's definitely this

41:00

ethos of like, oh, look, I

41:02

have innovated a solution. No, you've

41:04

basically figured out something that already

41:06

was against the law. Oh, you

41:08

didn't realize is that. You didn't

41:11

know what the law was. Well,

41:13

I think that's why Robin Hood

41:15

decided not to, right? That's possible.

41:17

Maybe it's more of a PR

41:20

hit, because I honestly don't think

41:22

that the SEC has now constituted

41:24

is going to do anything. Well,

41:26

that's a separate problem. In fact,

41:29

I'm wondering if I should even

41:31

file taxes this year. Yeah, that's

41:33

a separate problem, too. But I

41:35

had that thought from before. And

41:37

again, who's going to happen if

41:40

I don't? A little too soon

41:42

to make that. Maybe wait a

41:44

year? Okay, I'll try to next.

41:46

No, no, I mean, wait a

41:49

week even and see where we

41:51

are. Will there be an IRS

41:53

on April 15th? It's unknown at

41:55

this point. Do we have a

41:58

treasury department I think is the

42:00

bigger question? Well, IRS is part

42:02

of the treasury, right? Well, they

42:04

give the money. to the Treasury

42:06

even if they're, that money. What

42:09

department is the IRS under? It's

42:11

kind of Treasury. It's probably Treasury,

42:13

but I don't know, I didn't

42:15

realize I needed to know. Anyway,

42:18

that was one comment. I'm sure

42:20

we'll swing around back to that.

42:22

But the other comment, I'll definitely

42:24

get back to that. You notice

42:27

I'm burying it a little bit,

42:29

because I don't want people to

42:31

tune out too. I don't remember

42:33

what happened back then. It's the

42:35

one where they go off the

42:38

rails. It's dark bif Tanner and

42:40

he's basically made all this money

42:42

illicitly because he was. Bif might

42:44

be the CEO of crypto.com. I

42:47

don't know. crypto.com says quote. We

42:49

don't offer sports betting products. We

42:51

offer tradable cryptocurrency commodities and tradable

42:53

financial products which differ from products

42:56

offered by sports books. That's

42:59

my story and I'm sticking to

43:01

it. By the way, this is

43:03

the number one betting day on

43:06

the US calendar. Estimated $1.5 billion

43:08

in legal wagers on the Super

43:10

Bowl. The Supreme Court struck down

43:13

a federal prohibition, you might remember,

43:15

in 2018, creating this $15 billion

43:17

industry. 38 states in Washington DC

43:20

allow it. And Americans now legally

43:22

wage your more than $12 billion

43:24

a month. Yeah, the social effects

43:26

are really disruptive. I have to

43:29

think that this is, that gambling

43:31

addiction is going to become a

43:33

real problem. And you know what

43:36

is interesting? This came from England.

43:38

This came from the richest woman

43:40

in England. I've talked about this

43:43

before. Got rich because her dad

43:45

owned, remember in England, you're going

43:47

to have these betting shops, right?

43:49

bookie shops and you can go

43:52

and you could bet. And her

43:54

dad owned a bunch of them

43:56

and she decided, well let's get,

43:59

let's get digital. But she did

44:01

something that was super smart that

44:03

ended up making her the richest

44:06

woman in England richer than JayK

44:08

Rolling or the Queen. I guess

44:10

the Queen is no longer with

44:12

us, but the King. So Denise

44:15

Coates is trying to get in

44:17

the United States, but she created

44:19

the prop bet. And you've seen

44:22

if you watch the ads on

44:24

the football games. You'll see draft

44:26

kings and all these others Kevin

44:29

Hart talking about you can bet

44:31

on whether that kick is going

44:33

to be good You can bet

44:35

on whether they're gonna the snow

44:38

will hit the ground You can

44:40

bet on almost anything instantly and

44:42

get an instant reward and to

44:45

me this is a recipe for

44:47

disaster You know that's always been

44:49

the the story of a of

44:52

a real somebody with a real

44:54

gambling problem is they'll bet on

44:56

anything anything anything at all You

44:59

know, well, look at the positive

45:01

side. The younger generation has no

45:03

money to really bet, so. They're

45:05

doing it on credit and debt.

45:08

That's even worse. Oh my God.

45:10

Do you want to see this,

45:12

the, the graph? This is, this

45:15

is the graph of legal, of

45:17

legal. Yeah, that's a big problem.

45:19

There's very strong whiffs of things

45:22

are going very wrong with the

45:24

sports competitions themselves. Yeah. Yeah, there's

45:26

a lot of, there's, you know,

45:28

my son, who is a huge

45:31

fan of the Green Bay Packers,

45:33

is not going to watch the

45:35

Super Bowl today. He says it's

45:38

all rigged. Yeah. Now whether it

45:40

is or not, the fact that

45:42

the NFL profits from sports gambling.

45:45

Remember Pete Rose never got into

45:47

the Hall of Fame because he

45:49

bet on baseball? I think your

45:51

son's a really handy bellwether for

45:54

how gambling is just going to

45:56

undermine the nature of sports fans.

45:58

and further separate from being a

46:01

community original thing to being, come

46:03

on, make me some money thing.

46:05

Well, and that's something to worry

46:08

about. One of the things also,

46:10

I was thinking with the, with

46:12

that prop bet thing, you already

46:14

have problems even with the stock

46:17

markets, which in their own

46:19

way is their own form of

46:21

betting, but at least there's some

46:23

form of asset underneath it. you

46:25

have issues where the speed by

46:27

which the information can get exchanged

46:29

and the bet can be placed

46:31

has distorting values and there is

46:33

money to be made by arbitraging

46:35

the advantages that somebody has based

46:37

on the speed that they're able

46:39

to get things done. And we're

46:41

talking like speed in terms of

46:43

split seconds. So for all of

46:45

these things, it's not that you're

46:48

truly dealing with natural odds.

46:50

You're dealing with something that

46:52

is inherently those odds are flexing

46:54

based on advantages built in that

46:56

people aren't calculating for and are

46:59

not getting calculated for. So they

47:01

aren't fair bets. Well, I mean, that's

47:03

also the stock market. It's also, but

47:05

it becomes a big problem and it

47:08

becomes a problem that the stock market

47:10

actually has to try to solve

47:12

for. Well, that's true. And I'm

47:14

just saying these companies are regulated

47:16

by the commodities, the CFTC, what

47:18

does that stand for? The commodities,

47:20

something trading commission. which is a

47:22

separate governmental organization, which

47:25

by the way Elon Musk

47:27

is trying to dissolve right

47:30

now. The Commodities Future Trading

47:32

Commission did actually sign off

47:34

on what they call events

47:37

contracts, allowing people to trade

47:39

on whether Taylor Swift would announce

47:41

her next album or whether a

47:44

new movie would tank at the

47:46

box office. crypto.com submitted

47:48

filings to the CFTC.

47:50

They didn't ask permission, by the way.

47:53

This doesn't work that way. They informed

47:55

the commission of their intent. And

47:57

then it's kind of, there's a

47:59

fast track. process. These companies get

48:01

to self-certify their derivatives contracts. And

48:04

then the CFTC, if they

48:06

decided to, could shut it down.

48:08

Nobody did, of course. The firm

48:11

filed its paperwork five days before

48:13

Christmas. Probably thinking, you know, no,

48:15

nobody's going to be there. Actually,

48:18

it was the day before the

48:20

big government, the threat of a

48:23

government shutdown. They picked the timing

48:25

pretty wisely said Peter Malashev a

48:27

partner at a DC law firm

48:30

right before Christmas after the elections

48:32

So anyway I thought well, let

48:35

me see if I can

48:37

get a crypto.com account quick so

48:39

that I can just you know

48:41

show you what it looks like

48:44

to Trade on the result

48:46

of the Super Bowl unfortunately after

48:48

submitting a picture of my driver's

48:51

license, a video of me turning

48:53

my head left and right,

48:55

giving him my cell phone number,

48:58

they said, we'll get back to

49:00

you in one to five business

49:02

days. So I don't think I'm

49:05

going to get to bed on

49:07

the Super Bowl today. I just

49:10

thought it's got that. So there's

49:12

some security. I'm safe. Yeah. Hey,

49:14

good news. There's going to

49:16

be some good ads on the

49:19

on the game, the big game.

49:21

Open AI is a. doing its

49:24

first ad, there were I

49:26

think three different AI ads last

49:28

year, including anthropic, but no open

49:30

AI. Google, of course, talked about

49:33

its AI. This year, Google

49:35

was going to run a Super

49:37

Bowl ad, I think they still

49:40

are, that said that Gouda was

49:42

the world's most popular cheese. According

49:44

to Gemini. It is not. It

49:47

does not make up 50 to

49:49

60% of the world's cheese consumption.

49:52

That is not true. It's going

49:54

to say how many rocks

49:56

are in it. It isn't part

49:58

of the glue. Right. These are

50:01

almost glue anywhere. Google. has edited

50:03

Gemini's AI response in a

50:05

Super Bowl commercial, this is according

50:08

to the verge, to remove that

50:10

incorrect statistic, the ad shows the

50:12

small business owner using Gemini

50:14

to write a website description about

50:17

Gouda. In the edited video, Gemini's

50:19

response now skips over the specifics

50:22

and just says Gouda is one

50:24

of the most popular cheeses in

50:26

the world. At this point I

50:29

doubt that, but yeah. Well, you

50:31

know why? Because you know those

50:34

little red wax covered baby

50:36

bells? That's gooda. Ah, well, there

50:38

you go. And it's delicious, by

50:40

the way. And you're accepting sponsorship

50:43

offers from Baby Bell. I

50:45

would love some cheese. I'll take

50:47

my money and cheese. Google Cloudaps

50:50

president Jerry Dishler said, it's not

50:52

a hallucination. That's not, it

50:54

reminds me, remember Martin Short used

50:56

to do that lying lawyer, it

50:59

was smoking all the time, it

51:01

had this sweaty lip, I knew,

51:04

I knew that, I always knew,

51:06

it's, it's not a hallucination, it's

51:08

grounded in the web. Apparently, this

51:11

came from a website called cheese.com,

51:13

which is filled with, according

51:15

to the verge, what seems to

51:18

be SEO optimized blogs. Gouda is,

51:20

according to the Evie Baker Professor

51:22

of Agricultural Economics at Cornell,

51:24

most assuredly not the most widely

51:27

consumed cheese in the world. So

51:29

they've edited it out. So you

51:32

will see the ad on

51:34

the Super Bowl if you watch

51:36

that, but no. In fact, that

51:38

business owner was real, the website

51:41

was real, and he has now

51:43

removed the Gouda claim as well.

51:46

For the camera wasn't on me,

51:48

but I'm sure our listeners could

51:50

hear my eyes roll when you

51:53

were talking about this in

51:55

the SEO, my optimized blogs. Yeah,

51:57

well, Yeah, cheese.com, apparently not the

52:00

cheese authority you might have thought.

52:02

A disclaimer beneath Gemini's response

52:04

says, it's not intended to be

52:06

factual. So maybe you're right about

52:09

AI. Why would you? Reminds me,

52:11

the humane. Remember, humane had

52:13

their, their, their, their, their, their,

52:16

their, their, pan. Or, and they

52:18

made a prediction. I had somebody

52:21

do with the stars and, or

52:23

an eclipse or something, and, and,

52:25

oh, wrong. Well, right. Google rolled

52:28

out, The demo had incorrect

52:30

facts. Gemini hallucinated in

52:33

its rollout. Yeah, Gemini

52:35

too just came out this

52:37

week and people are saying

52:39

great things about it. I don't

52:41

know. See, I just, I don't want

52:43

to poo-poo you guys. We will

52:45

find out, but I honestly think

52:48

that we are seeing such

52:50

amazing progress with AI that

52:52

it is almost inevitable that

52:55

we are going to see.

52:57

human level intelligence in the

52:59

next few years, maybe sooner

53:01

or later. I'm not entirely sure

53:03

what that something is. Oh, I

53:05

could be wrong. Something we should

53:07

actually want. Yeah, I can be

53:10

as confidently wrong as people are.

53:12

From a pure computing

53:14

standpoint. this stuff is cool like

53:16

and and we are amazing we've

53:18

not seen anything like this within

53:21

the in terms of is it

53:23

anything that can substitute for intelligence

53:25

is it anything that is useful

53:27

valuable and also not dangerous I

53:29

I would not put you know

53:31

all my chips on that prediction

53:33

I'm not putting all my chips on

53:35

it but I am I feel fairly you

53:38

know I wear now I wouldn't put any

53:40

chips on it I keep showing this this

53:42

is this is a This is announced at CES,

53:44

this is the B computer, it's recording

53:46

this conversation and everything that happens to

53:48

me. And then it gives me an

53:50

AI summary at the end of every

53:52

day with action items and all that

53:55

stuff. And yeah, it makes a lot of

53:57

dumb mistakes when I was watching a movie

53:59

the other night. and it thinks I'm

54:01

rehearsing for a role in Richard III.

54:03

So it's a little confused, because I'm

54:06

not. But there's some stuff in it

54:08

that's kind of amazing, and I feel

54:10

like it's making my journal for me.

54:12

I honestly wish I'd had this for

54:15

my whole life, because it would be

54:17

so cool. But there are mistakes in

54:19

it. You just, you have to, I

54:22

think you have to understand how to

54:24

use it, really. I don't mind it

54:26

as a tool that the humans wield.

54:29

What I bristle at is this idea

54:31

that it is going to be the

54:33

tool that can just have a certain

54:36

degree of autonomy that humans do not

54:38

need to wield it anymore. And especially

54:40

when it comes to atomic weapons. Well,

54:43

you also talk about transparency should be

54:45

part of that too, at least. with

54:47

a human workflow you do have certain

54:50

levels of built-in accountability either in an

54:52

implicit in social way or in an

54:54

explicit organizational way and one of the

54:57

things that's super disturbing about the way

54:59

we're mainstreaming AI utilization in general is

55:01

is the whole it's a black box

55:04

don't you worry you're pretty little monkey

55:06

heads about how it works right no

55:08

I agree with you but it's well

55:11

it's kind of a black box transparency

55:13

and accountability and not in a give

55:15

us your trade secrets way but rather

55:18

when you do have a screw up

55:20

like glue pizza, what are you doing

55:22

about this? How did it happen? How

55:25

can we be sure it will never

55:27

happen again? No, I think that's the

55:29

wrong attitude to be honest. I respectfully

55:32

disagree because that, first of all, you

55:34

can't fix that. It is a black

55:36

box even to the guys who are

55:39

writing this stuff. They don't know exactly

55:41

how it comes to these conclusions. And

55:43

I think AI safety is a mistake.

55:46

What we, you can't expect the, first

55:48

of all, we know it doesn't work.

55:50

Every AI has been jailbroken. We did

55:53

a whole show on it on Tuesday

55:55

with Steve Gibson. It's almost impossible to

55:57

have AI safety. What it requires is

56:00

human intelligence intermediating it. So I agree

56:02

with you, you should not give control.

56:04

of our nuclear oxygen to an AI.

56:07

There's got to be a human in

56:09

between the AI and the launch button.

56:11

But I don't think you can make

56:14

the AI not make mistakes, not hallucinate,

56:16

or be safe. I think that that's-

56:18

There just needs to be a built-in

56:21

level of review and accountability. That's currilacking.

56:23

Yeah. But I think what we have

56:25

to do is train humans, train ourselves

56:28

to use AI appropriately. I really like

56:30

the idea of DEI for AI, by

56:32

the way. I'm going to... Diversity, inclusion

56:35

for AI. It comes down to data

56:37

sets. Well, who's responsible for picking the

56:39

data sets for training? Right. And more

56:42

importantly, how is their work being checked?

56:44

But this was the breakthrough, I think,

56:46

that made this all possible, is that

56:49

we have... I remember 30 years ago,

56:51

I did stories on it. There was

56:53

a woman who was trying to create

56:56

an AI and had... an army of

56:58

100 stenographers typing in every fact she

57:00

could find. Well, in the intervening three

57:03

years, we did it for her, we

57:05

created the internet and put everything we

57:07

could think of into it, and then

57:10

the AI has access to it. So

57:12

I don't think it's a question of

57:14

data sets as much. I mean, it

57:17

might be for face recognition, things like

57:19

that. Generally yeah it's it's whatever they

57:21

can get their hands on right yeah

57:24

yeah well actually that's one of our

57:26

stories is meta yeah with meta in

57:28

the books and sure thing so meta

57:31

what what's the story here there's a

57:33

really really great book that I want

57:35

I want to recommend to anyone who

57:38

wants to talk about AI called code

57:40

dependent by wired reporter Madhamita Mirgia and

57:42

she travels the country and the globe

57:45

and goes to different countries where AI

57:47

sweatshops to set up where people are

57:49

specifically tagging images or specifically tagging different

57:52

pieces of data to train the AI

57:54

and she points out there is no

57:56

unless you have people who dig into

57:59

this you have no idea how what

58:01

data is being used to train you

58:03

have no idea how it's being tagged,

58:06

you have no idea how it's being

58:08

structured, or if it can be even

58:10

used as we get to more and

58:13

more sophisticated iterations of AI, because structure

58:15

versus unstructured data. I think we've run

58:17

out of, I think we've actually, the

58:20

biggest problem is we've run out of

58:22

stuff to train AI with. Yeah. Well

58:24

then let's get to your next story,

58:27

which is a doozy. Oh my gosh.

58:29

Oh, we got big. You know what?

58:31

I'm, I, this is, this like show

58:34

is like a mountain. We've started at

58:36

the little tip of the iceberg. We're

58:38

getting to the middle. There's, there is

58:41

some stuff to talk about. What's going

58:43

on in Washington DC is somewhat shocking.

58:45

Kathy, you've written quite a bit about

58:48

that. We have lots more to talk

58:50

about. It's good to have you, Kathy

58:52

Gellis, writing for Tech, read her, read

58:55

her, her articles this week, her articles,

58:57

this week. I'm amazed that you're not

58:59

like. bouncing off the walls here. I

59:02

am apoplectic. I'm just amazed that I

59:04

am still sitting here, but give a

59:06

time. I can see the apoplexy forming

59:09

also here from Windows Central Daniel Rubino

59:11

editor-in-chief. Always a pleasure to have you

59:13

on. Thank you, Daniel. And of course,

59:16

Leisha Schmeiser from No jitter.com. Our show

59:18

today brought to you by Kota. Have

59:20

you used Kota? Kota is so cool.

59:23

turning your back of a napkin idea

59:25

into a billion dollar startup. You know,

59:27

have you ever thought about that? It's

59:30

going to take countless hours of collaboration,

59:32

of teamwork. And it's hard to build

59:34

that team. It's a team that's aligned

59:37

on everything for values to workflow. But

59:39

there is a tool that helps. It's

59:41

exactly what CODA. was made to do.

59:44

CODA, CODA, it's an all-in-one collaborative workspaces,

59:46

started literally itself as a napkin sketch,

59:48

and now in the five years since

59:51

launching in beta, CODA has helped 50,000

59:53

teams all over the world get on

59:55

the same page. With CODA, you get

59:58

the flexibility. Actually, let me give you

1:00:00

the address right now, because I know

1:00:02

you want to just look at it,

1:00:05

and you can't. Go to coda.io,

1:00:07

coda.io/Twitter. And while you're looking,

1:00:09

let me explain. With CODA,

1:00:11

you get the flexibility of

1:00:13

docs, the structure of spreadsheets,

1:00:15

the power of applications, and

1:00:17

the intelligence of AI, all

1:00:19

together in a seamless workspace

1:00:21

built for enterprise. It

1:00:23

facilitates deeper collaboration, quicker

1:00:26

creativity, gives you more time

1:00:28

to build. You've got to take a

1:00:30

look at this. It's really mind-blowing. It's

1:00:32

incredible. If you're a startup team and

1:00:34

you're looking to increase alignment and agility,

1:00:36

it's a really hard thing to do,

1:00:38

especially, I mean, you can say return

1:00:40

to office, but frankly, most of us

1:00:42

are still remote, right? Cota brings you

1:00:45

all together instead of a physical

1:00:47

office. You have a virtual workspace

1:00:49

that everybody is involved with. Is

1:00:51

it as attached to us, is

1:00:53

utilizing. COD can help you move

1:00:55

from planning to execution and record-to

1:00:57

execution and record time. You've got

1:00:59

to try it for yourself. Here's

1:01:01

the good news. Now you're

1:01:03

at cota.io/Twitter, right? You're

1:01:05

looking, this is interesting. How

1:01:08

about six months free? Six months

1:01:10

free of the team plan for

1:01:12

startups. What about that? Six

1:01:15

months half a year free. C-O-D-A-O-

1:01:17

slash, that's enough to get

1:01:19

that napkin pretty far off

1:01:21

along the line to a

1:01:23

startup, to a unicorn. Get

1:01:25

started for free, get six.

1:01:27

free months of the team

1:01:29

plan. That's the best offer

1:01:31

ever. kota.io/to what we love

1:01:33

these guys. Thank you, Kota

1:01:35

for supporting us. Thank you,

1:01:37

Kota. So let's talk about

1:01:39

this meta story. I'm not

1:01:41

really kind of sure what

1:01:43

happened. Of course, meta, like

1:01:45

every other AI company is

1:01:47

trying to train on as much

1:01:49

data as possible. There is a

1:01:51

rich trove of data that isn't

1:01:54

in fact on the internet. It's

1:01:56

called books and if

1:01:58

you could ingest library

1:02:00

right it'd be good for

1:02:02

your AI to train on

1:02:04

every possible book out there

1:02:07

and of course authors and

1:02:09

publishers not too happy about it

1:02:11

there is a copyright case against

1:02:14

meta raised by book book

1:02:16

authors they allege meta illegally

1:02:18

trained its AI models not

1:02:20

on books it bought but

1:02:22

on pirated books last month

1:02:25

meta admitted to torrenting Controversial

1:02:27

data set known as Libgen

1:02:29

has tens of millions of

1:02:32

pirated books. We kind of

1:02:34

knew this because you were

1:02:36

able to find bits and

1:02:39

pieces of those books in

1:02:41

meta and llama. Yesterday, meta's,

1:02:43

well actually it was I

1:02:46

think Thursday, meta's unredacted emails

1:02:48

were made public for the

1:02:51

first time. The evidence is

1:02:53

meta torrented 81.7 terabytes

1:02:56

of data. Through the site

1:02:58

Anna's archive, including 35.7

1:03:00

terabytes of data from Z

1:03:03

library and Libgen, the court

1:03:05

filing said, Meta also previously

1:03:07

tormented 80.6 terabytes of data

1:03:10

from Libgen. All onto a

1:03:12

laptop! Turrenting from a corporate

1:03:15

laptop doesn't feel right, says

1:03:17

Nikolai Balashkov, a meta research

1:03:20

engineer. writing in April. This

1:03:22

was the message, 2023 message,

1:03:25

adding a smiley emoji in

1:03:27

the same message she expressed

1:03:30

concern about using meta IP

1:03:32

addresses to load through Torrance

1:03:34

pirate content. By September, Bashlakov

1:03:37

dropped the emogies consulting

1:03:39

the legal team directly and

1:03:41

emphasizing in an email that

1:03:43

quote, using Torrance would entail ceding

1:03:46

the ceding, ceding an email that

1:03:48

quote using Torrance would entail ceding

1:03:50

the ceding the ceding the files.

1:03:52

i.e. sharing the content outside, that's what you

1:03:54

don't have to do that, but that's kind

1:03:57

of the, you know, polite thing to do.

1:03:59

This could be... He said, he wrote,

1:04:01

legally not okay. Man, this is why

1:04:03

companies don't want you to put this

1:04:05

stuff in emails. Oh,

1:04:08

emails prove that Meta knew

1:04:10

it was illegal. Bashlikov's

1:04:14

warnings landed on deaf ears. The

1:04:18

authors who were suing say the evidence

1:04:20

showed Meta chose instead to hide its turning

1:04:22

as best it could while downloading and

1:04:24

seeding. They did in fact say, you don't

1:04:26

have to seed, by the way, if

1:04:28

you download. But they did, seeding terabytes of

1:04:30

data from multiple shadow libraries as recently

1:04:32

as April of last year. They

1:04:36

didn't use Facebook's servers while

1:04:38

downloading the dataset to, quote,

1:04:40

avoid the risk of anyone

1:04:42

tracing the seeder downloader. This

1:04:44

is from an internal message

1:04:46

from Meta researcher, Frank Zhang.

1:04:50

He described the work as being in stealth

1:04:52

mode. They modified

1:04:54

settings so that the smallest amount

1:04:56

of seeding possible could occur. Oh,

1:04:59

well. I

1:05:02

think they're busted. Mark

1:05:04

Zuckerberg said, I didn't do it.

1:05:06

I didn't know anything about it.

1:05:08

Come on, Mark, where's that masculine

1:05:10

energy that would, to meet the

1:05:12

moment. He claimed to have no

1:05:15

involvement in the decisions to use

1:05:17

libgen to train AI models. Well,

1:05:19

it wouldn't really matter, I think,

1:05:21

whether he did or he didn't.

1:05:23

is the CEO. He's so much

1:05:25

the person of the, there's some

1:05:27

open questions whether the persona of

1:05:29

the company really can be separate

1:05:31

from some of its leaders, especially

1:05:33

when their personal actions are so

1:05:36

closely tied to them in a

1:05:38

way that corporate structures are not

1:05:40

really operable. I think for him,

1:05:42

he's in better shape than, say,

1:05:44

Musk up the road, but that

1:05:46

raises a question. I don't think

1:05:48

somebody will, I'm sure, try to

1:05:50

pick at that. I have a

1:05:52

couple of concerns about the story

1:05:54

and things that are in the

1:05:57

story. This is, by the way,

1:05:59

I'll give credit to Ashley Belanger,

1:06:01

senior policy. reporter at Arts Technica writing this story.

1:06:03

Go ahead. I, so if they seated, I

1:06:05

do kind of face palm at them,

1:06:07

but a couple of, I don't

1:06:09

necessarily have the same concern if

1:06:11

they downloaded because I don't think

1:06:13

there should be a distinctive... Is

1:06:16

that the right to read you've

1:06:18

spoken about? It would still go into

1:06:20

that, like if you were allowed to...

1:06:22

Does it change it that it's pirated

1:06:24

material? No, because it... I don't think

1:06:27

it's something that it essentially would be

1:06:29

that the copyright holder could say no

1:06:31

you can't read this book unless you

1:06:33

paid but the copyright really only applies

1:06:36

to making copies of it itself. Maybe

1:06:38

it kind of is an AI adjusting

1:06:40

something making a copy isn't that making

1:06:42

a copy? I don't I don't think so

1:06:44

and all and the lawyers who are defending

1:06:47

these companies are very clear that what you

1:06:49

really have is not making a copy you

1:06:51

really have it learning and just storing

1:06:53

it. than me reading and remembering something

1:06:55

I read from a book. I'm not

1:06:57

making a copy of my brain. Yeah,

1:06:59

the lawyers are arguing that, but I

1:07:01

think based on how the technology actually

1:07:04

behaves, that it's really more of a

1:07:06

learning function, in which case, if you're

1:07:08

trying to create an artificial intelligence, it's

1:07:10

going to function and learn the way

1:07:12

a human intelligence would and store information

1:07:15

that it's gleaned in some way that

1:07:17

it can use it again. So in

1:07:19

that sense, I think the downloading is

1:07:21

not particularly dispositive. You know, I take,

1:07:23

I don't, I don't represent meta so

1:07:26

I can throw stones at it as

1:07:28

much as I want, but. Oh yeah, my

1:07:30

legal senses are like, oh, seriously,

1:07:32

dude, like, okay, fine, if I

1:07:34

had to defend it, I defend

1:07:36

it, but I think it's a harder,

1:07:39

it's a harder left. I also, I

1:07:41

have, I have torn it for years,

1:07:43

you don't have to seed. There's

1:07:46

a checkbox in the client. As

1:07:48

the lawyer, I don't want to

1:07:50

blur these two things together. I

1:07:53

think the legal analysis of whether

1:07:55

downloading was okay is a different

1:07:57

legal analysis for whether seating. that

1:08:00

bothers me about this case, there's some

1:08:02

earlier reporting where the, um, some of

1:08:04

these emails had been privileged and the

1:08:07

privilege was essentially punctured through assertions of

1:08:09

the crime fraud exception to attorney client

1:08:11

privilege and I'm a little bit uncomfortable

1:08:13

with and maybe more than a little,

1:08:16

but I, there would be more details

1:08:18

about that happening. You're an attorney and

1:08:20

of course, you're thinking like an attorney.

1:08:22

And that is important in the trial,

1:08:25

but it's not important to our discussion,

1:08:27

right? Well, but it kind of is.

1:08:29

Well, no, let's stipulate that they did,

1:08:31

in fact, bit torn a bunch of

1:08:34

pirated books. I mean, the discussion we're

1:08:36

having is, is that okay or not?

1:08:38

You say it's okay. No, I think

1:08:40

it does matter, because the whole way

1:08:43

that you could puncturing privilege is a

1:08:45

really. No, I understand. From a legal

1:08:47

point of view, that's important. But it

1:08:49

doesn't matter. We can still talk about

1:08:52

it. But it's predicated on the idea

1:08:54

that it is so wrong what was

1:08:56

being protected by this privilege that you

1:08:59

could then have this. That could be

1:09:01

the consequence of it. And I'm thinking

1:09:03

that I'm not sure it is wrong

1:09:05

enough that it should have had that

1:09:08

collateral consequence. Well, it may get them

1:09:10

off the hook. We're not debating whether

1:09:12

they should be what should happen to

1:09:14

the trial. We're just talking about the

1:09:17

fact that they did it and whether

1:09:19

that's okay. Do you think it's okay,

1:09:21

Lisa? You know, I

1:09:23

am just going to be quiet

1:09:25

and let the lawyer and let

1:09:27

the Leo talk to each other

1:09:29

about it. How about you Daniel?

1:09:31

No, this is this is a

1:09:33

real lesson for me. This is

1:09:35

this is a master class and

1:09:37

legal excellence. I'm just here to

1:09:39

learn. Yeah, I've learned from Kathy

1:09:41

about this so-called right to read,

1:09:43

which I think is very interesting.

1:09:45

And if you say an AI

1:09:47

is is not copying but learning

1:09:49

from something, that's something we do.

1:09:51

Well Kathy is it illegal for

1:09:53

me to read a pirated book?

1:09:55

I don't know what right the

1:09:57

copyright holder would be able to

1:09:59

assert against. you. They might try

1:10:01

to say that your personal... They

1:10:03

could go after the piracy site.

1:10:05

Yeah, I think a lot of

1:10:07

the file sharing litigation originally had

1:10:09

been more that making available was

1:10:12

the thing that stepped on the

1:10:14

copyright. So that's what the reading

1:10:16

is a lot harder. Right. They

1:10:18

never went after people for listening.

1:10:20

Right. And it's a much harder

1:10:22

reach to be able to do

1:10:24

that because that's not that's not

1:10:26

controlling the copying in the same

1:10:28

way. Right. It's more. It's a

1:10:30

video games. Everything I was torn

1:10:32

today never went after the people

1:10:34

who were just downloading it. And,

1:10:36

you know, getting back to Leo,

1:10:38

you have pointed us out the

1:10:40

seating part. That's the uncomfortable issue,

1:10:42

right? It's just like your ISP,

1:10:44

if it catches you. using bit

1:10:46

torrent, it's not so much because

1:10:48

you're downloading, it's because you're seeding

1:10:50

and you're sharing copyright of material.

1:10:52

That's the legality. So when it

1:10:54

comes to this issue here with

1:10:56

meta and Facebook and books, it's

1:10:58

the seating that feels a little

1:11:00

uncomfortable that they're doing, but the

1:11:02

ability to, and I think the

1:11:04

courts have very kind of upheld

1:11:06

this a little bit, just came

1:11:08

back to that, you know, ability

1:11:10

to read. It's true. AI. If

1:11:12

you could ask AI, like co-pilot,

1:11:14

like co-pilot or Gemini, or Gemini,

1:11:16

be like, you know, you know,

1:11:18

print up the entire book for

1:11:20

Stephen King, his latest book, and

1:11:22

it gave you the entire thing.

1:11:24

Okay, that would be an issue,

1:11:26

right? Because that would be then

1:11:28

distributing an entire book for free,

1:11:30

you know, and free. But it

1:11:32

doesn't do that. I mean, really,

1:11:34

that a lot of... New York

1:11:36

Times says that that Open AI

1:11:38

did in fact regurgitate New York

1:11:40

Times articles. But in order to

1:11:42

do that, they had to jump

1:11:44

through hoops. They had to print

1:11:46

the first two paragraphs and say,

1:11:48

what would the first two paragraphs

1:11:50

and say, I'm not convinced that

1:11:52

AI is there to, or is

1:11:54

even able to make, you know,

1:11:56

you can't get war and peace

1:11:58

out of an AI. It's also

1:12:00

worth talking a little, well actually,

1:12:02

I guess it was, the copyright

1:12:04

office came out with a study

1:12:06

on its second level of study

1:12:08

on AI and where copyright law

1:12:10

needs to be, although I think,

1:12:12

actually the one that just came

1:12:14

out wasn't on liability for output,

1:12:16

I think it was based on

1:12:18

potential protection available for output. And

1:12:20

it basically said, it's gonna be

1:12:22

case specific and you'll have to

1:12:25

look at how much originality went

1:12:27

into causing it to generate something.

1:12:29

Just fascinating stuff. By the way,

1:12:31

Amazon is going to have a

1:12:33

big event revealing something we have

1:12:35

all been waiting for. February 26th,

1:12:37

you're gonna see Panos Panos Panay,

1:12:39

the former beloved devices guy from

1:12:41

Microsoft, show the new Amazon echoes

1:12:43

with AI, I think, right? This

1:12:45

is, Panos, how many people is

1:12:47

Panos stolen from Microsoft? Didn't more

1:12:49

people follow him over to Amazon?

1:12:51

Yeah. Ralph Green recently went over

1:12:53

there, their head designer. So we

1:12:55

should see more German brutalist designs

1:12:57

for election devices, which I'm all

1:12:59

for. We have made an excess

1:13:01

square. It's like, how is that

1:13:03

bad? That sounds pretty awesome. It

1:13:05

says now three by two because

1:13:07

that's the problem. Awesome designer. Yeah,

1:13:09

I love that surfaces look. I

1:13:11

think they're great. Yeah, this is.

1:13:13

interesting only because Panos Panos and

1:13:15

Amroth Gray and Urb now part

1:13:17

of that division because they do

1:13:19

have some amazing ideas and they

1:13:21

tend to be very innovative and

1:13:23

if anything you know Amazon and

1:13:25

Alexa needs a real kick in

1:13:27

the pants because they have some

1:13:29

real issues their products are fine

1:13:31

but the whole point of theory

1:13:33

which isn't saying much yeah which

1:13:35

isn't so much, but the whole

1:13:37

point was to get you to

1:13:39

use Alexa to buy more stuff

1:13:41

from Amazon or supposed to be

1:13:43

a shopping system. Everybody just ends

1:13:45

up using it as a timer

1:13:47

to play music and they're losing

1:13:49

money on all the hardware. So

1:13:51

there's no reason to use it.

1:13:53

Billions. Reason to use it. Yeah.

1:13:55

They've lost, I think they've got

1:13:57

to figure out, 10 billion or

1:13:59

something over years. Yeah. So they

1:14:01

got to figure out how to

1:14:03

make this interesting enough to, especially

1:14:05

to, especially to charge, to charge,

1:14:07

But they've explored charging people $5

1:14:09

to $10 a month for that.

1:14:11

That was the rumor. Yeah. Yeah.

1:14:13

So like, they got to monetize

1:14:15

this, right? So how useful is

1:14:17

this going to be in your

1:14:19

house versus what's on your smartphone?

1:14:21

Which is real, you know, that's

1:14:23

where the real war has been.

1:14:25

But if Android has theirs, and

1:14:27

of course you have Apple has

1:14:29

theirs, but there's no room there

1:14:31

for Amazon, just like there was

1:14:33

no room for Microsoft on it

1:14:35

ways. I don't. for a smarter

1:14:38

Amazon echo. I have to pay

1:14:40

zero dollars for an echo because

1:14:42

I don't want an echo. You

1:14:44

don't have any echoes anyway. I

1:14:46

don't have any smart home devices

1:14:48

because I don't trust the companies

1:14:50

enough in terms of protecting my

1:14:52

data or respecting my privacy. So

1:14:54

every room in my house has

1:14:56

a Siri, has an echo, and

1:14:58

has a Google voice assistant. Every

1:15:00

room in my house. And I

1:15:02

think that's great. And I wear

1:15:04

this bracelet that's sending everything that

1:15:06

happens to some unknown AI in

1:15:08

the sky. I don't even know

1:15:10

where it goes. We're gonna, by

1:15:12

the way, interview the creators of

1:15:14

this on the 19th of February.

1:15:16

And that's the first thing I'm

1:15:18

asking of is, where's this going?

1:15:20

I think the privacy thing is

1:15:22

always. Go ahead. Just because the

1:15:24

privacy thing is an interesting discussion

1:15:26

because it is in theory extremely

1:15:28

important. In reality though, it's not.

1:15:30

We've seen over and over again

1:15:32

after years reporting on this stuff

1:15:34

that people really just don't kind

1:15:36

of care. They'd rather have convenience

1:15:38

or I don't care. They already

1:15:40

put all their information out there

1:15:42

online. There's not a week that

1:15:44

goes by where most of us

1:15:46

don't get a letter in mail

1:15:48

saying your data's been breached somewhere,

1:15:50

right? Like this is just like.

1:15:52

I stand in front of my

1:15:54

echo show and dance naked just

1:15:56

for the fun of it. Right.

1:15:58

Yeah, exactly. Hoping that somebody in

1:16:00

China is forced to watch it.

1:16:02

There's a joke, right, with younger

1:16:04

generations. They put so much out

1:16:06

there when they're young, that like,

1:16:08

like, you know, for myself, somebody

1:16:10

came out from photos of me

1:16:12

in college, would be a little

1:16:14

embarrassing, right? Because of my generation.

1:16:16

But if you're younger, you really

1:16:18

put everything out there and exposed

1:16:20

yourself. There's nothing left you can

1:16:22

really do. But even, you know,

1:16:24

nudity is not that big of

1:16:26

a deal anymore. So it's like,

1:16:28

there is an interesting counter argument

1:16:30

to this, which is if you

1:16:32

completely expose yourself as much as

1:16:34

possible, there really is no risk

1:16:36

of your data leaking ever or

1:16:38

privacy concerns, right? You just let

1:16:40

it all go. want that you

1:16:42

know Amazon's been talking about this

1:16:44

for a while it's been slowed

1:16:46

down because according to reports it

1:16:48

was incredibly stupid and just really

1:16:51

was like so bad that they

1:16:53

couldn't release it according to Reuters

1:16:55

executives have scheduled a go no

1:16:57

go meeting for putting AI into

1:16:59

echo for Valentine's Day writers says

1:17:01

there they will make a final

1:17:03

decision on the street readiness of

1:17:05

echoes generative AI revamp according to

1:17:07

the people and an internal planning

1:17:09

document seen by Reuters. So it

1:17:11

isn't yet known whether they'll release

1:17:13

this and there have been you

1:17:15

know there have been reports of

1:17:17

it not being very good. I

1:17:19

couldn't get worse in Syria. Syria

1:17:21

has literally gotten worse with Apple

1:17:23

intelligence than it was before. John

1:17:25

Gruber and and another blogger post

1:17:27

asked a asked Siri to tell

1:17:29

it who won every Super Bowl

1:17:31

from zero through 60 and it

1:17:33

was no it was horrible it

1:17:35

was terribly wrong whereas in the

1:17:37

past it would have just said

1:17:39

I don't know but look here's

1:17:41

what I found on the web

1:17:43

about that which would have been

1:17:45

accurate there's a lot that's gotten

1:17:47

bad as AI has been it

1:17:49

minimum sucking all the oxygen in

1:17:51

the room, but also now also

1:17:53

getting embedded in all sorts of

1:17:55

software. I think if you counted

1:17:57

what has not gotten worse, let

1:17:59

alone what has actually gotten better,

1:18:01

the count, you know, very little

1:18:03

has not gotten worse. from all

1:18:05

this. Well, remember, you were, you

1:18:07

were talking about eating rocks and

1:18:09

putting Elmer's glue on pizza. That

1:18:11

was a Google search result. Google

1:18:13

has now again started testing a

1:18:15

new search AI mode internally. And

1:18:17

the reason is there's intense pressure.

1:18:19

I don't use Google anymore. I

1:18:21

use perplexity. I use AI search.

1:18:23

And I bet a lot of

1:18:25

people do. The search results are

1:18:27

better. Google's got to see this

1:18:29

as an existential threat to its

1:18:31

business. Yet Google's existential threat is

1:18:33

making its search engine which had

1:18:35

been industry leading now crap. Well

1:18:37

that's true they did it to

1:18:39

themselves didn't they did it themselves

1:18:41

they were doing they produced a

1:18:43

really good product for years and

1:18:45

years and years and then decided

1:18:47

Let's not. Let's go change it.

1:18:49

Let's give up on everything that

1:18:51

made it good. And instead of

1:18:53

trying to make it better, we

1:18:55

will just make it different. And

1:18:57

they've made it different in a

1:18:59

way that just can't compete with

1:19:01

what they originally had. But they

1:19:03

keep drawing anyway. And it's terrible.

1:19:06

And people just lose trust with

1:19:08

the company. And we don't like

1:19:10

it as an AI company. And

1:19:12

we don't even like it as

1:19:14

a search engine anymore. From 9

1:19:16

to 5 Google. Here here's a

1:19:18

screenshot. How many boxes of spaghetti

1:19:20

should I buy to feed six

1:19:22

adults and ten children and have

1:19:24

enough for seconds? Are they American

1:19:26

or European children? Yeah, really. Americans,

1:19:28

you need like 80 boxes. Adding

1:19:30

the children's portions and the adults

1:19:32

portions gives you a total estimate

1:19:34

of 38 to 54 ounces. Increasing

1:19:36

this by 25 to 50% for

1:19:38

second servings puts you in the

1:19:40

range of 47.5 to 81 ounces.

1:19:42

Most boxes of spaghetti are one

1:19:44

pound. Anyway, it says you should

1:19:46

get three to five boxes of

1:19:48

standard-sized spaghetti to feed six adults.

1:19:50

This is an example, yeah. Or

1:19:52

compare wool down in synthetic jackets

1:19:54

in terms of insulation, water resistance,

1:19:56

and durability. You know, that is

1:19:58

the kind of... we would prefer

1:20:00

to do then, you know, where

1:20:02

can I buy down jackets? Yeah,

1:20:04

or what's the best down jacket,

1:20:06

which we know is a terrible

1:20:08

yourself. Really good. There's there's an,

1:20:10

you know, you point that out

1:20:12

though, and now I'm thinking there's

1:20:14

like wire cutter and Tom's guide

1:20:16

and all of these other sites

1:20:18

that are making their bones on.

1:20:20

offering that kind of a value

1:20:22

of buying advice is journalism. But

1:20:24

they should be expert, based on

1:20:26

expert research. Yeah, no, that's right.

1:20:28

Yeah, because this AI will combine

1:20:30

all of those results. It might

1:20:32

give. I will take their content

1:20:34

and match it up and spin

1:20:36

it out. Yeah, but badly. So

1:20:38

actually, as long as it's bad,

1:20:40

no, they shouldn't be scared, except

1:20:42

it's very annoying because it's hard

1:20:44

to find. My favorite way of

1:20:46

searching Google. smart enough to send

1:20:48

me to the site that best

1:20:50

answers the natural language. How's that

1:20:52

working for you? Reasonably well. I

1:20:54

mean, maybe I'm only using a

1:20:56

narrow range of questions and it

1:20:58

seems to go better with some

1:21:00

than others. And sometimes that may

1:21:02

be based on what is available.

1:21:04

So I think the more interesting

1:21:06

thing is what if when there's

1:21:08

something not on. If there's something

1:21:10

on point, I really don't want

1:21:12

Google to get in the way

1:21:14

of it. If there's nothing on

1:21:16

point, then the question is, what

1:21:19

should Google do, if anything, to

1:21:21

gapfill that? Right. The other issue

1:21:23

that's gonna come up with something

1:21:25

like this, though, is you're gonna

1:21:27

see a, I mean, you're already

1:21:29

seeing a flood of AI glorish

1:21:31

in terms of content writing, where

1:21:33

there are companies devoted to using

1:21:35

generative AI models to pump out

1:21:37

things that can. rank very highly

1:21:39

in search results and say absolutely

1:21:41

nothing of use or value. There's

1:21:43

no primary sources. There's no report

1:21:45

and there's no nothing. And with

1:21:47

this kind of search that result

1:21:49

that you've put in there, Leo,

1:21:51

I didn't see any sources for

1:21:53

citations. I didn't see any facts

1:21:55

that backed it up. And it

1:21:57

actually was a screenshot from Google

1:21:59

by the way. But no, no,

1:22:01

I'm just saying. It would be

1:22:03

super easy to game the results

1:22:05

just by, you know, just by

1:22:07

having AI generate all sorts of

1:22:09

all sorts of content that gets

1:22:12

tossed into the pool of training

1:22:14

data or gets tossed into the

1:22:16

pool of results that any sort

1:22:18

of AI is going to use

1:22:20

to try to aggregate synthesize and

1:22:22

put back a good answer. I

1:22:24

just saw an ad. You have

1:22:26

to take a look at the

1:22:28

data and see where it's coming

1:22:30

from. You really do. I've got

1:22:32

Fox running in the background. I

1:22:34

just saw an ad for a

1:22:36

company called Ramp featuring Saekwan Barkley.

1:22:38

And the whole thing it does

1:22:40

is it looks at all the

1:22:43

things you do. Like in his

1:22:45

case you had a leg day.

1:22:47

And it said, well, we didn't

1:22:49

get a receipt. Give me the

1:22:51

receipt for that leg day. And

1:22:53

Zakuan took a picture of it

1:22:55

and said, okay, got it. And

1:22:57

that's done with AI. But it's

1:22:59

more than just AI. It's AI

1:23:01

that's also watching you every moment

1:23:03

of the day to see what

1:23:05

you're up to. So it can

1:23:07

say, hey, you just took an

1:23:09

Uber. Where's that receipt? This is

1:23:11

the world we're getting into. And

1:23:13

I think you were right when

1:23:15

you said, Daniel, Daniel people don't

1:23:17

care. people. I hate buying expenses.

1:23:20

Yeah. If it asks, if it

1:23:22

asks, if my phone says, hey,

1:23:24

where's the receipt for that? And

1:23:26

I take a picture. I'm happy.

1:23:28

Yeah. Yes. Right. Yes. I need

1:23:30

someone to nag me on certain

1:23:32

things in life and I'll gladly

1:23:34

take that. Oh my gosh. Let's

1:23:36

take a break. I gotta take

1:23:38

a break. We have more. You're

1:23:40

going to have plenty of time

1:23:42

to smash AI. We got lots

1:23:44

more AI coming up, including. Well.

1:23:46

I'll let you stay tuned. We

1:23:48

still haven't talked about TikTok. We've

1:23:50

got a lot of talking to

1:23:52

do. We've got talking to do...

1:23:55

Cybersecurity. Senator Josh Hawley has proposed

1:23:57

jail time for people who download

1:23:59

Deep Seek. Not an insignificant amount

1:24:01

of day time. It's the... Decoupling

1:24:03

America's artificial intelligence capabilities

1:24:05

from China Act. D-A-A-A-I, no, you

1:24:07

could have done better, Josh. You

1:24:10

could have done better, there's so

1:24:12

many ways. So many, so many

1:24:15

ways, you could be fined not

1:24:17

more than a million dollars

1:24:19

and imprisoned for not more

1:24:21

than 20 years. Oh, well, how merciful

1:24:24

of him. No more than 20

1:24:26

years, okay. Just for downloading Deep

1:24:28

Seek, which I have done now

1:24:30

on many of my devices. That's

1:24:32

the Chinese AI that apparently Josh

1:24:35

is very worried about. Great to

1:24:37

have all. I wish this is

1:24:39

this should be an eight-hour show.

1:24:41

There is so much to talk

1:24:43

about. Last night I said, oh, do

1:24:45

I have enough stories? And then I

1:24:47

looked and I went, oh my God.

1:24:50

And I want to get out of

1:24:52

here in an hour because there's some.

1:24:54

I don't know, some football game going

1:24:57

to happen a little later on. You're

1:24:59

watching this week in tech with Lisa

1:25:01

Schmeiser of No jitter.com, Daniel Rubino, Windows

1:25:04

Central, and our own personal attorney,

1:25:06

Kathy Gellis, from Techtert. Good to have

1:25:08

all of you. By the way, not

1:25:10

just Techtert, R street.org, right? You wrote

1:25:13

a nice piece for them. That

1:25:15

was really good. I wrote an erstwhile fellow

1:25:17

with them, and I wrote a bait paper.

1:25:19

Yeah, I didn't realize you were a fellow

1:25:21

there. Yeah, partially. I don't know

1:25:23

if I have a permanent title, but

1:25:26

I did fell under the auspices of

1:25:28

a fellowship. I wrote a white paper

1:25:30

about jaw boning and the DMCA. By

1:25:32

the way, apparently jaw boning,

1:25:34

whatever that is, has a little

1:25:36

bit to do with the TikTok

1:25:38

case as well. I read your

1:25:40

article. We'll talk about it when

1:25:42

we return. But first, a word

1:25:44

from our sponsor. A sponsor, I'm

1:25:47

very proud to say I not

1:25:49

only use. Steve Gibson uses and

1:25:51

we recommend heartily bit warden the

1:25:53

best password manager well more

1:25:55

than a password manager there are

1:25:57

trusted leader and passwords secrets and

1:25:59

pass key management. By the way,

1:26:02

I love pass keys. I wish

1:26:04

more sites used it. When it

1:26:06

first came out, you had pass

1:26:08

keys on your phone, you had

1:26:10

pass keys on your computer. Now

1:26:12

that bit warden supports pass keys,

1:26:14

wherever you have bit warden. You

1:26:16

have your passkeys and that has

1:26:18

been a big improvement in how

1:26:20

passkeys works for me. Bitward now

1:26:22

has 10 million users in 180

1:26:24

countries. This is an open source

1:26:26

success story. Over 50,000 business customers

1:26:28

too. In fact, they've entered 2025

1:26:30

as the essential security solution for

1:26:32

businesses of all sizes, consistently ranked

1:26:34

number one and user satisfaction by

1:26:36

G2. recognized as a leader by

1:26:38

Software Reviews Data Quadrant. Bitwarden continues

1:26:41

to protect businesses worldwide. One of

1:26:43

the reasons I love Bitwarden, and

1:26:45

I think it's because they're open

1:26:47

source. They're always expanding their capabilities.

1:26:49

They're always improving. Recently they announced

1:26:51

the general availability of their native

1:26:53

mobile apps for Iowa and Android.

1:26:55

I did not realize this. I'd

1:26:57

been using Bitwarden for years, but

1:26:59

it wasn't a native app on

1:27:01

iOS. I had no idea. And

1:27:03

it's made a huge difference, you

1:27:05

know? I guess it was an

1:27:07

electron app or a view, a

1:27:09

webview app. Key benefits of the

1:27:11

native mobile app include on iOS

1:27:13

and Android, faster load times, improved

1:27:15

overall app functionality. iOS and Android

1:27:17

platform-specific design, so it really, you

1:27:20

know, it looks like it belongs

1:27:22

on the platform. Much better hardware

1:27:24

integration, in fact, that includes biometric

1:27:26

authentication and multi- device support. So

1:27:28

that's really enhanced usability usability. do

1:27:30

the face ID and my bit

1:27:32

warden and it's unlocked. They have

1:27:34

strengthened their password manager now with

1:27:36

SSH. This is really important. So

1:27:38

if you use SSH as I

1:27:40

do it and you use a

1:27:42

key to log into SSH. You

1:27:44

can now store your authorized SSH

1:27:46

keys in bit warden. This is

1:27:48

huge. to 90% of authorized SSH

1:27:50

keys in large organizations just go

1:27:52

unused because it's not convenient. Well

1:27:54

now this new update centralizes cryptographic

1:27:57

key management, not just SSH, but

1:27:59

in general, enabling secure storage, import,

1:28:01

import, import, you can even generate

1:28:03

your SSH keys directly within the

1:28:05

bitward and vault, which is a

1:28:07

huge improvement for developers and IT

1:28:09

professionals. I'm thrilled. I mean, I

1:28:11

used to have to do a

1:28:13

whole bunch of command line food.

1:28:15

What set Spidwarden apart is that

1:28:17

they prioritize simplicity because they understand

1:28:19

if a security tool isn't easy

1:28:21

to use, isn't simple, people aren't

1:28:23

going to use it, right? Bitwarden

1:28:25

setup only takes a few minutes.

1:28:27

You can easily import from most

1:28:29

password management solutions. And of course,

1:28:31

because it's open source, that means

1:28:33

Spidwarden source code can be inspected

1:28:36

by anyone. They undergo regular or

1:28:38

third party audits and, very importantly,

1:28:40

they publish the full report. So

1:28:42

you can see everything going on

1:28:44

within Bitwarden and be reassured that

1:28:46

it is absolutely secure. Look, your

1:28:48

business deserves a cost-effective solution for

1:28:50

enhanced online security. See for yourself.

1:28:52

Get started today with Bitwarden's free

1:28:54

trial of a teams or enterprise

1:28:56

plan or get started for free

1:28:58

across all devices as an individual

1:29:00

user. bitwarden.com. Let me underscore this,

1:29:02

because they're open source, if you're

1:29:04

an individual user, it's free, and

1:29:06

you get all the capabilities, unlimited

1:29:08

passwords, unlimited devices, iOS, Android, Mac,

1:29:10

Windows, Linux, you also get pass

1:29:12

keys, you also get hardware security

1:29:15

keys, all of that in a

1:29:17

free version, free, forever, for life.

1:29:19

And that's nice. bitwarden.com slash, Twitter.

1:29:21

Since that's still an ongoing thing

1:29:23

right? I mean we never resolved

1:29:25

it. We just did something really

1:29:27

stupid and And here we are.

1:29:29

So just to fill you in,

1:29:31

this is the best. Previously, this

1:29:33

week in tech. Kathy is a

1:29:35

adamant supporter of the First Amendment,

1:29:37

as we all should be, and

1:29:39

says that TikTok is protected by

1:29:41

the First Amendment. Not just TikTok

1:29:43

itself, but every user of TikTok

1:29:45

is exercising their right to free

1:29:47

speech. There are many who say

1:29:49

that or believe that TikTok is

1:29:52

a security problem because it is

1:29:54

a Chinese company with it gives

1:29:56

it presumably gives access effect we

1:29:58

had a open AI guy on

1:30:00

Wednesday on the show said nobody

1:30:02

should have I don't have TikTok

1:30:04

on my phone you'd be crazy

1:30:06

have TikTok on your phone or

1:30:08

Deep Seek or any of these

1:30:10

apps. And I'm not disputing that.

1:30:12

I don't have it either and

1:30:14

for similar reasons. So let's just,

1:30:16

you know, stipulate that Tiktok is

1:30:18

as much of a privacy menace

1:30:20

as it's been accused. That notwithstanding,

1:30:22

the question is, what does the

1:30:24

government get to do about that?

1:30:26

And I think the First Amendment

1:30:28

says, not this, but the Supreme

1:30:31

Court disagreed, although they disagreed very

1:30:33

narrowly. But it was nine, nothing,

1:30:35

right, were you surprised that it

1:30:37

was unanimous? We should mention Kathy

1:30:39

was in the court during the

1:30:41

oral arguments. We had you on

1:30:43

shortly after that. She's admitted to

1:30:45

the Supreme Court to the board.

1:30:47

So I wrote an amicus brief

1:30:49

in the case arguing this is

1:30:51

not how anything works. And I

1:30:53

attended the oral argument listening to

1:30:55

it. I was cautiously optimistic because

1:30:57

it's sort of, I think they

1:30:59

kind of understood what the First

1:31:01

Amendment issues were. But the whole

1:31:03

thing was a mess. Oh, it's

1:31:05

even messier now. It's even messier

1:31:07

now. And so you're asking about

1:31:10

the nine nothing. And I think

1:31:12

the nine nothing is kind of

1:31:14

a byproduct of they just wanted

1:31:16

to get it off its docket

1:31:18

for whatever reason. And one of

1:31:20

the things that concerns me is

1:31:22

I don't know why they thought

1:31:24

they were. unable to stop the

1:31:26

clock on this and provide and

1:31:28

enjoying things and then have things

1:31:30

being briefed at a proper pace

1:31:32

when they could read all the

1:31:34

briefs and get all the full

1:31:36

argument to get all the amici

1:31:38

to have time not over their

1:31:40

Christmas break to be able to

1:31:42

weigh in. So they ended up,

1:31:44

because they didn't do that, they

1:31:47

ended up with everything on a

1:31:49

very accelerated schedule, and then we're

1:31:51

kind of like, well, we heard

1:31:53

the case, we got to do

1:31:55

something. And I have a sense

1:31:57

that the 9-0 decision that they

1:31:59

came out with was the only

1:32:01

thing they could kind of get

1:32:03

everybody to agree on to get

1:32:05

something out the door. So it's

1:32:07

got a lot of framing to

1:32:09

say, be really careful, we're only

1:32:11

speaking to this particular situation as

1:32:13

it applies to Tik-talk. you know

1:32:15

don't read too widely we want

1:32:17

to be cautious about not you

1:32:19

know hurting technology this and the

1:32:21

other thing that kind of acknowledge

1:32:23

that they might be wrong but

1:32:26

they were wrong and they made

1:32:28

a mess and the only thing

1:32:30

that i think is somewhat good

1:32:32

about it is i think they