0:00

It's time for Twig this week in Google Jeff

0:02

Jarvis is here. Paris Martin, no is

0:04

here. We're going to talk about, of course, our

0:06

top story. The New York Times suing Microsoft

0:09

and open AI saying they're stealing

0:11

our content, but are they really,

0:13

I'll make a case for keeping

0:15

your hands off of AI. No

0:17

regulation at all. And

0:19

Neil dash makes a case for the internet getting

0:22

weird again in 2024. And

0:25

why you might want to pay attention to how much

0:27

you're making on those app-based

0:29

sites like Airbnb, you're

0:31

now going to be reported to the

0:34

IRS. It's all coming up next. So

0:36

this week at Google podcasts,

0:38

you love from

0:40

people you trust. This

0:44

is Twig. This

0:50

is Twig this week in Google episode 749

0:52

recorded Wednesday, January 3rd, 2024,

0:58

a well-regulated chart team. This

1:01

episode of this week in Google brought to

1:04

you by babble, the language learning app that

1:06

actually works. Babble uses quick 10 minute

1:08

lessons designed by over 150 language experts. So

1:12

you can start speaking a new language in

1:14

as little as three weeks. Babble's built

1:16

with science backed cognitive tools like

1:19

spaced repetition and interactive

1:21

lessons. And it's created to help

1:23

you build real world conversations. So

1:26

you can learn how to order food, ask

1:28

for directions and speak to

1:30

merchants. It's easy. Babble

1:32

also teaches and I love this context

1:34

traditions, the culture of the language that

1:36

you're learning. I'm learning Spanish right now.

1:39

I studied French earlier. It's

1:41

just fantastic from self-study app lessons

1:43

to podcasts, even live classes,

1:46

which is great for conversational speaking.

1:48

Babble has a wide range of learning

1:50

experiences for learners of all ages based

1:52

on level and your time commitment. That's

1:55

one of the things I love about babble. It's just a few

1:57

minutes a day. If I want 15 hours

1:59

with. BABL is equal to one

2:01

university semester. Plus all of BABL's

2:03

14 language courses are backed

2:06

by their 20-day money back guarantee.

2:09

Now I'm a lifetime member of BABL,

2:11

because I love it so much, but

2:13

we've got a special limited time deal.

2:15

If you're new to BABL, right now

2:17

get 55% off your BABL subscription. This

2:19

is only for listeners. You have to

2:21

go to babl.com/twig. 55%

2:25

off at

2:27

babl.com/twig. babbel.com/

2:30

twig. Rules and restrictions apply. It's

2:33

time for Twig. This week in Google,

2:35

the show where we wear funny glasses

2:37

and talk about Google. Hello Paris Martineau.

2:39

You see I wore my specs. Yours

2:42

are handmade. I was gonna say, I don't

2:44

know what you're talking about. These glasses are

2:47

decidedly serious. Nothing funny about them. Nothing funny

2:49

about those. Jeff Jarvis is

2:51

not wearing funny glasses either. They didn't send me

2:53

the memo. I'm so hurt. Yeah, he's just wearing

2:55

your John Lennon, your usual

2:58

John Lennon glasses. Jeff is the Leonard Tao Professor

3:00

for Journalistic Innovation at the Craig Newmark Graduate

3:04

School of Journalism at the

3:06

City University of New York, Emeritus. We have to keep

3:08

playing that through August because

3:10

we pay for the singers. So

3:14

we'll keep playing that.

3:16

And then Paris is of course at

3:18

the Information where she does great

3:21

work. Still working on that big story that you can't tell us about? I

3:24

am. Okay, good. Watch

3:26

the internet, guys. When it breaks, you'll let

3:28

us know, right? I will. You'll

3:30

be the first to tell us. So I

3:33

don't normally, I usually wear contacts. I wear glasses,

3:35

but I usually wear contacts on the show. But

3:38

I thought, you know, I got some fun glasses and

3:40

I thought, I can, Paris and

3:42

I now can talk about our glasses. These,

3:45

these glasses which really make me

3:47

look like Elliot Gould in Ocean's Eleven or

3:50

maybe Larry Budd Melman, I'm not sure which.

3:53

You do seem like you're gonna do a heist. For

3:56

those of you listening, big. black

4:00

glasses kind of like Michael Caine might wear.

4:03

But these are from a company called Vinyl

4:05

Eyes. They're made out of old records. Oh,

4:09

that makes sense. I was about to say

4:11

they have some sort of reflection that you're

4:13

seeing. That's the groove. It's the life that

4:15

is like a record. It's the grooves of

4:17

the record as I move

4:20

my face around. Yeah, there they are. You

4:22

need to put a needle on it to see what

4:24

the little bits are. Oh, you do. Okay, so each

4:27

model has a different name. This is called Fleetwood. I'm figuring

4:29

it's Fleetwood Mac. It could be the

4:31

Rumors album. It could be Tusk. I

4:33

don't know. It's true. They

4:36

also have an AC, S-E-E-D-C version, which

4:38

I figure is old AC-DC records. So

4:41

isn't that a good idea? So I'm doing

4:43

my part to keep the earth green. That's

4:46

super cool. By wearing an old Fleetwood Mac

4:48

album on my face. No

4:51

one would ever have used for that

4:53

Fleetwood Mac vinyl. There's nowhere else could go.

4:55

We have two, not one, but two

4:58

vinyl record stories in Petaluma alone.

5:01

Oh, you would? Oh,

5:03

it's the hippest thing. You know, I have a... I

5:06

don't know. All right. Do

5:08

either of you guys have record players? No. Do

5:11

you? I think you would. I do, yeah.

5:14

You're a Brooklyn hipster. I mean, that goes with that

5:16

saying. I got my way for one for Christmas two

5:18

years ago. She has all her albums downstairs, yeah. I

5:21

thought she's played with one. She has.

5:23

I just don't want to ever have to move albums

5:26

again. That is such

5:28

a pain. Yeah,

5:30

I don't have that many. I

5:32

have more of a concern of moving all

5:34

of my books. Books are bad too. It's

5:37

just unnecessary. I stuffed everything, my albums, my

5:39

books and everything into an iPhone. When

5:42

I move, I put it in my pocket and I'm gone. No

5:45

boxes. You put your 27 phones

5:47

in your pocket and there you go. Well,

5:49

there's a disadvantage. Yes, yeah, having so many

5:51

phones is not a good thing. How

5:55

many phone numbers do you have? Well,

5:57

you know, it's funny. We did a little... Clean

6:00

out during the holiday season. We went

6:02

down of writing and reading your phone

6:05

number. Cn went out of rise know

6:07

I'm mesmerized stock to bow yes it's

6:09

a I want to close out my

6:12

account. Believe. Lows:

6:14

it's closer. I. Am

6:16

an app. The were through all of

6:18

them so subtly assists the sea ice

6:20

storm and eighteen seek out with no

6:22

eighteen T Sims or numbers or anything

6:24

but I have it accounts and I'm

6:26

paying money so senseless there must be

6:28

some gum. So

6:31

that the the next I would see the honestly

6:34

they make it so hard against against the have

6:36

to go nowhere else and we get the rest

6:38

are the chances are you can't do it here

6:40

as rewards we do a suggest going one on

6:43

the phone and columns but ah well as the

6:45

torture The member manager was good he said but

6:47

I will walk you through and I will expedite

6:49

so it so he says a kid dallas number.says

6:51

okay immediately press one I said okay. This

6:54

is Susan my starts talking press one a Catholic,

6:56

some acting. As

6:58

it went right through the tech support and

7:01

it was much so credit to the manager,

7:03

the register Bedlam. Was. Very very

7:05

nice of them. He could have we

7:07

explained it's it's something went wrong the

7:09

signals apparently they the must have lost

7:11

a tower in Petaluma. Is used

7:14

a very strong five G signals and now

7:16

not. So. What

7:18

Do you mean? How many phone numbers did you have

7:20

on your for As and plan? Let.

7:23

Me Put it this way: one hundred. And

7:25

one you can get. it seemed minute sure I

7:27

get these him for I went against and said

7:29

no you can't have too many men it's. I

7:35

don't know the limit is, but there is actually a limit.

7:39

On Nazis you have a huge family.

7:41

Does you have Zoom in your corner?

7:43

These other only take it another. Is

7:47

lost out to a stretch limo full of

7:49

skills and. My family.

7:51

What are you doing here Road. or

7:55

right ah did you see had a nice

7:57

a holiday were ten members is a new

7:59

year's eve party you actually New Year's Eve

8:01

Eve Eve Eve Eve very important yeah wonderful

8:03

Eve it's the day before the day before

8:05

the new year and it's a great holiday

8:08

to celebrate with your friends nice although

8:10

I guess not for the next couple

8:12

of years because it's been good weekend

8:14

recently yeah it's been the weekend but

8:17

starting that year's Eve Eve will fall

8:19

on a workday which is not as

8:21

fun as it really is fun and

8:24

Jeff did you did you do anything for the holidays so I

8:27

was angry when the neighbors

8:30

fireworks woke me up at night on New

8:33

Year's Eve I'm an

8:35

old guy yeah did you go out

8:38

and shake your fist step along nothing

8:40

thanks too much energy just

8:43

first he just thought that thought so

8:45

the big story we talked a

8:50

little bit about it on Windows Weekly this week the

8:53

New York Times is

8:57

suing Microsoft and open AI

9:00

saying you bad people you

9:02

are your

9:05

AI systems are widespread

9:07

copying copying

9:09

our stories and that's copyright

9:12

infringement telling

9:14

Lee the folks at the Times have

9:16

asked for a jury they've demanded in

9:18

the terms of the plea a

9:20

jury trial is

9:23

it there right alone to do that yeah

9:26

I think I also have yeah I mean the defendant

9:28

I think would have more of a right to say

9:30

now I think that you well I

9:32

guess the judge will rule a judge

9:34

will rule but I think

9:36

that is typically the plaintiff

9:39

who demands I don't know really

9:41

no because remember Donald

9:44

Trump was complaining about not having a trial in the

9:46

New York case and the

9:48

judge pointed out that his attorneys had

9:50

specifically not asked for a jury so

9:56

yeah I think either side probably can ask for it anyway I

9:58

don't know I'm not a lawyer we should get Matthew Yellison.

10:02

Mike Masnick wrote about this and said,

10:04

the New York Times really should think

10:06

about what they're

10:10

asking for because this could bite them on the beehives.

10:16

The New York Times, he points

10:18

out, very commonly will

10:21

notice a story in another journal

10:24

and then launch their

10:26

own investigation from

10:30

it, sometimes without credit. And he says,

10:33

if you can sue OpenAI for taking

10:35

your material and training an

10:37

AI on it, what's to stop those

10:39

other journals from suing you? The

10:42

New York Times lawsuit, he says, against OpenAI would

10:44

open up the New York Times to all sorts

10:46

of lawsuits should it win. But

10:49

to me, honestly, that isn't

10:51

even the issue. I

10:54

really think that from my point

10:56

of view, and by the way, I got

10:58

in a little fight with Paul Tharott who said, as

11:01

representing creators, the

11:05

New York Times fighting the good fight. Creators

11:07

ought to sue AI for using

11:10

their content. So

11:13

I went on the CBC to talk about

11:16

this. And

11:18

my argument, and Paris made fun of me saying,

11:20

well, you can't stay away from a camera for

11:22

more than a week, can you? Okay, it was

11:24

very funny because it was a Wednesday right when

11:26

we record this show that you were like, I'm

11:28

going to be live. I've got to have my

11:30

airtime. What am I

11:32

doing? So my

11:35

contention is that the machine has a

11:37

right to learn and

11:39

that it doesn't record

11:42

it. And we'll go

11:44

into, I think, the details about some of the Times

11:46

allegations about specific segments in a minute. But

11:49

if the machine can't learn, if the

11:52

machine can't do what we do, Then

11:55

it's a problem. And Also on line

11:57

68, it is the heritage of our

11:59

industry. That we go back to

12:01

renting a measure the show before the newspapers

12:03

had scissors editors. Who. Was

12:05

their job to cut up newspapers and

12:07

put them in copyright did not covered

12:09

his papers at all at first or

12:11

magazines even when it did it didn't

12:13

cover news per se it only covered

12:15

com a special things of authors of

12:17

and so it is a bit i

12:20

might write as it is it is

12:22

you're right leone stuff for issue on

12:24

but i think it's is disingenuous of

12:26

the are tons. To. Act as

12:28

if are they don't do this every damn day.

12:30

I love this paragraph from Mike's article. In the

12:32

end though, the crux of this lawsuit. Is.

12:34

The same as all the others.

12:37

And he's talking about Sarah Silverman

12:39

lawsuit, George Rr. Martin's lawsuit against

12:41

an open A Ice. It's a

12:43

false belief that reading something. Whether.

12:46

By a human aura, machine somehow

12:48

implicates copyright. This is false. Is

12:51

the courts or the legislature decide

12:53

otherwise, it would upset pretty much

12:55

all of the history of copyright.

12:58

And create some significant real

13:00

world problems. Or. Out though

13:03

Paul said that in the New York

13:05

Times complaint which go run sixty four

13:07

pages that they. Provided.

13:09

Ample examples of chatty busy quoting them

13:12

for being I saw I thought they

13:14

were they got him to, but Mazda

13:16

Katzenberg appointed as another one I put

13:19

up online. Or. Sixty

13:21

Six. Which. Has it demonstrations.

13:23

One of the examples is a

13:26

cool a quote from a review

13:28

of Guy Fieri New York restaurants

13:30

and as I'm. Whoever.

13:33

Does who wrote this? Ah sad that

13:35

Kevin A. are I am. A.

13:37

To see I'm right in. The him

13:39

on Twitter that I'm quote was was

13:41

all over the internet was quoted again

13:44

and again and again and again. A

13:46

so wasn't hard to get of one

13:48

one place and then be what cat

13:50

what what Mazda points out. A.

13:53

lot of inside impersonal what lights

13:55

points out is that the way

13:57

the queries were done if my

14:00

OpenAI down the path so that basically the only response

14:03

could have been the words that were next to the

14:05

words that were next to the words next words in

14:07

that case. Yeah that's

14:09

what I without reading the entire plea that

14:11

was my belief was that

14:14

the New York Times had carefully crafted

14:16

the prompts. Oh yes. And my issue

14:18

and what I told Paul is no

14:20

one is you're not asserting I

14:22

hope and the time wise the Times is

14:24

that somebody would read or query chat GPT

14:26

in lieu of reading the New York Times

14:28

that they would say oh I don't have

14:30

to buy the Times because I can get

14:32

everything I want from chat GPT because

14:35

they're just gonna quote the time. Well they are

14:37

arguing that in so far as there's the

14:40

same argument that Google was used against Google.

14:42

Exactly made it against Google. Well and Mike

14:44

points out that I don't know if it

14:46

was in the in the complaint per se

14:48

but they they whine about about well Wirecutter.

14:50

Mike's example is I go to Wirecutter and

14:52

I ask what's the best bike I get

14:54

the answer that's all I need is the

14:56

brand and the model number from search.

14:58

Yes that's true I don't go to the New York Times

15:00

to read the whole thing and yes

15:02

the New York Times doesn't then get the affiliate money

15:05

but God didn't give it to them. Sorry

15:07

guys you're you're involved in an

15:10

information ecosystem that you the New

15:12

York Times take advantage of every

15:14

single day where information once known

15:16

is free to use. The hot

15:19

news doctrine the Associated Press years ago tried

15:21

to have a hot new I remember this

15:24

which was to say that there were two things

15:26

that went on one was that they said that

15:28

that that there was a period of time was

15:30

never established but about 12 hours or

15:32

day where if you broke the story when

15:34

Paris has her big story out it's

15:37

hers for a day nobody can even

15:39

mention it because it's Paris's right and

15:42

that the courts didn't go for that they did at first they didn't

15:44

go for that and then when radio

15:46

came along newspapers tried to do the same

15:48

thing and they told radio that they weren't

15:50

allowed to discuss any story until 12 hours

15:52

after it happened. That

15:55

also went by the way and

15:57

but the same kind of sacred language

15:59

that's in the New York Times complaint,

16:02

which Mike makes fun of a great

16:04

life is, we're so special. We're the

16:06

New York Times. We're saving democracy. We

16:08

put so much money into it. Yeah,

16:10

yeah, yeah. So, I want to hear

16:12

what you have to say about it, Peristo. Yes. I

16:14

have other thoughts, though. I

16:16

mean, I'm curious about your thoughts. I

16:20

think that the arguments that you guys have outlined are

16:22

correct. I think that

16:24

at first, you know, you see the argument

16:27

the Times is presenting as, oh, people shouldn't

16:30

be able to scrape our news and use

16:32

it to train these systems. But I think,

16:34

as we've just discussed, the

16:37

actual issue at hand is a lot more complicated

16:39

than that. And also, the idea that

16:41

anyone is going to be using chat

16:43

GPT to get a line-by-line read

16:46

of a Guy Fieri review, and

16:48

that is going to undermine democracy.

16:50

A 12-year-old Guy Fieri review, which

16:52

was talked about like crazy because

16:54

it was so unfair to Fieri.

16:56

People screamed about it and quoted

16:59

it at length all over. Sorry,

17:01

guys. Go ahead. No. I mean, I think

17:03

that, I think, what

17:05

do you guys think is going to happen in the courts

17:08

with this? Here's my issue. First of all, I

17:10

think it's telling that the New York Times demanded

17:13

jury truck because they know a judge will,

17:15

as judges already have, throw this out on

17:18

the face of it. It's fair use.

17:20

Judges have already ruled again and again

17:22

that AI has the right to scrape

17:24

the Internet and generate its large-language models

17:26

from that content that's publicly available. And

17:29

it's not a violation of copyright.

17:31

I think that that's going to always be the

17:33

case with a judge. Jury might be different. And

17:36

I think it's one of the reasons this bleeding is so

17:38

emotional is they're playing to a prospective

17:41

jury saying, oh, you know, you don't

17:43

want the New York Times to fail, do you? But

17:48

here's my big thing now

17:50

that I am an AI

17:52

accelerationist. I

18:01

think Paul Therat said, and I think

18:03

I agree, that really this is a negotiation ploy,

18:05

as it usually is. Yes. At

18:07

the Times just wants to sue them so

18:09

that OpenAI will give them some money. And

18:12

unfortunately, both Google and OpenAI have already

18:14

done this with other journals,

18:18

and as a result have created this

18:20

slippery slope. I don't

18:22

think they should ever do this, because

18:25

I think really this

18:28

is the worst kind of regulatory capture.

18:31

Yes. Yes. Because the future

18:33

of AI really, in my opinion, and we have a

18:35

Jan Lekoon article we can talk about in Wired, an

18:37

interview in Wired, but I agree with

18:40

him where he says really the real future, the future

18:42

you want with AI is open. That's

18:44

what OpenAI was supposed to be. You want open

18:47

source AI. You want everybody to be able to

18:49

develop with it, use it, create stuff with it.

18:51

You don't want the big tech companies to be

18:53

the gatekeepers. You don't want Microsoft, Google,

18:56

Amazon, Apple, anybody to

18:58

own AI. You don't want them to be the

19:00

gatekeepers. You want AI to be everywhere. And

19:03

I truly believe that. I think that's

19:05

really important. So no matter what

19:07

the upshot of this case is, if the

19:10

New York Times wins, if

19:12

OpenAI ends up

19:15

paying them, it means that it

19:17

makes it harder for the open source

19:20

AI to build and succeed.

19:23

So it's going to have a bad outcome.

19:25

The only possible good outcome is

19:27

if the jury, and I don't

19:29

think so, but if the jury says no, this is fair

19:32

use, go away New York Times,

19:35

then that would open it up for everybody. I

19:37

mean, do you think though that companies like OpenAI

19:39

should be able to use anyone's

19:41

materials or anything? If it's publicly on the

19:44

internet, they can learn from it. They can

19:46

train from it. Not quote it, not

19:48

steal it. And this is the

19:50

thing. If it's behind a paywall and

19:53

it's not metered, should they be

19:55

able to scrape that?

19:57

Actually, that's an interesting one. If

20:00

OpenAI pays for one subscription to the New

20:02

York Times and then scrapes all the content.

20:04

Right, exactly. That's an interesting question. I

20:07

had a conversation, if I could add to this, I

20:09

had a conversation this week with Rich Sprinta, who started

20:11

Topics years ago and I didn't know it. Rich is

20:13

now the Executive Director of the Common Crawl

20:17

Foundation. Which

20:19

Mike mentions in this, by the way. Mike mentions in

20:21

this. Yeah. And what Mike

20:23

also says in his story is that the New York Times, and I

20:25

just find this, there's a

20:27

larger issue for the ethics and morals of journalism

20:29

and society here. The New York Times, Mike

20:32

says, has demanded that Common Crawl take

20:34

off all of the content that it

20:37

got from the New York Times. Now

20:40

OpenAI and company have done

20:42

a, you know, a robots.txt for this

20:44

case, which is fine. But

20:47

the New York Times is talking, respectively,

20:49

in the past, to take

20:51

it off. What is Common Crawl? Is

20:54

it like archive.org? Is it like the Internet

20:56

Archive? It's different to this extent. All it

20:58

does is it scrapes huge amounts

21:00

of data from

21:03

the open web, open, open, open web.

21:06

Mainly it was intended for academics, so

21:09

that there was a source of study, which

21:11

is invaluable. And it's open

21:13

source and it's free and it's gigantic. He told

21:16

me how long it would take to download it.

21:18

It's like, you know, forever. Huge,

21:21

huge thing. Whereas Internet

21:23

Archive saves the actual pages and the images and

21:25

all that. This is text only. It

21:28

was for the purposes of academic

21:30

research. It's a foundation.

21:33

Well, so along come

21:36

LLMs and, hello, Glorioski.

21:38

Look at that. What a great resource,

21:40

right? So they're using it and they're all happy to use it. But

21:44

now you have all these organizations like Reddit, as

21:47

very publicly said, well, we don't want anybody to scrape us

21:49

because we think there's a gold pot of gold

21:51

there. The problem in the case

21:53

of Reddit is, well, you didn't make the stuff,

21:55

Reddit. Your users did. Right. And

21:58

so... So everybody thinks

22:00

that there's some pot of gold here and they all

22:02

own something. And I

22:04

think that we've got to have a

22:06

conversation in journalism and media about our

22:09

obligation to the public information ecosystem. If

22:11

everything ends up behind the paywall, which is their

22:14

right to do, that's fine. If everything remains unscrapable

22:16

or you're doomed at a court case. And the

22:18

only thing that's left out there for people and

22:20

machines to learn from without going bankrupt and

22:23

for open source efforts and for small

22:25

journalists and so on is crap and

22:28

propaganda and lies. I

22:30

mean, how are those original

22:32

creators of the news content

22:35

supposed to survive in a media

22:37

ecosystem where their work has

22:40

to be essentially fair use

22:42

and publicly available for free? No,

22:45

they can put it behind the paywall. It

22:47

just has to also be accessible by OpenAI.

22:50

No, no, I would say

22:52

the good compromise here is if it is behind a

22:54

paywall, OpenAI can't get it. Books

22:56

should only be publicly accessible material. For

22:59

instance, your articles are behind the paywall and

23:01

the information. They shouldn't be able to scrape

23:03

the information. However, I

23:05

do think that OpenAI is different from

23:07

Books3. Books3 did not

23:09

obviously buy every book in Books3. Many

23:12

of those are pirated. And that's what Sarah

23:14

Silverman et al's objection

23:16

was. Well, you're reading our

23:18

books by pirating them. But

23:21

in this case, you really can subscribe to the New

23:23

York Times. The New York Times can put in terms

23:25

of use. That

23:27

would be a solution. Yes.

23:30

Yes. The bigger question is

23:32

that the business

23:34

model of news is badly

23:36

broken. And

23:38

I think that when we see ourselves in a

23:40

position where we think all of our value is

23:42

resident in this thing we call content, we're screwed.

23:45

Because content is a commodity. Because machines can now

23:47

make it. Only in

23:49

content is special. It's not where the value is. The

23:52

information what's special is that you do reporting

23:54

others don't do. And

23:56

people who find value in that in

23:58

their jobs in many cases. and me

24:00

and my podcast, choose

24:03

to subscribe for a lot of money. I was glad

24:05

there was a sale on it. I got it cheaper.

24:07

Thank you very much, Jessica.

24:10

But, so, and if

24:12

you're really special and if you're really good, you can do that.

24:15

So much of journalism, look at the Guy Fieri

24:17

story. So much of journalism is about us copying

24:19

each other. You get our own pages and

24:21

our own page views and our own likes and our own

24:23

clicks and our own and pennies. And

24:25

the amount of unique original

24:28

journalism that occurs, like the

24:30

information, is a lot rarer than

24:32

we admit. So we're trying

24:34

to support a whole infrastructure of copying

24:36

each other to get our own SEO

24:39

and clicks. So I think we've got

24:41

to have an honest audit of

24:43

where the real value is in journalism and where it isn't.

24:46

I almost feel like a lot of the New York Times, I

24:48

love the New York Times. I pay for the New York Times.

24:50

I criticize the New York Times because it's the best and I

24:52

want it to be better. But a

24:54

lot of the New York Times has nothing

24:57

to do with news, has nothing to

24:59

do with improving the democracy. It's fluffed

25:01

to attract attention, which is an old

25:04

business model that's gone away. Sorry. I

25:08

also think that there is a society,

25:12

and we've talked about this before, just as

25:15

in the early days of the internet, we didn't

25:17

tax it. We were reluctant to put

25:19

a lot of government regulation on it because we didn't know

25:21

where it was going. We wanted it to grow, and that

25:24

worked out. Admittedly, there have

25:26

been problems, but we know what those are. But

25:29

I think it was the right choice. I think we needed to do the

25:31

same thing with AI. I

25:34

really believe that AI has a lot of potential,

25:36

but we don't know exactly what it's

25:38

going to be. And I think

25:41

that it would be problematic if it's

25:44

old media trying to hold back new media.

25:47

And I think that would be problematic. And

25:50

Eric, do you

25:52

see any difference in learning

25:56

from materials to just

25:58

teach the machine to speak? as

26:00

a skill versus

26:05

giving back answers

26:08

from current content that

26:11

are reliable because their

26:13

current content. Can you see what

26:15

they're kind of do? I think I

26:19

see a difference there but I think

26:21

that ultimately the hesitation

26:23

I have around this generally comes from

26:25

the fact that we're ultimately talking about

26:29

for-profit companies that are building

26:31

models to then sell use

26:34

of that model to people.

26:37

I think that there should be a robust

26:40

public debate and I guess

26:43

legal debate around

26:45

whether or not those

26:48

companies can benefit

26:50

from other people's work. I

26:56

love that idea because and that's

26:58

by the way why OpenAI originally

27:00

wasn't a non-profit, it wasn't tenable

27:02

but if you wrote an exemption

27:04

for open source AI

27:06

work and said but if you're

27:08

going to charge for it, well then you

27:11

got a license. I'm okay with that. In

27:13

fact, I think that's the best possibility because

27:16

I don't want Microsoft and Google and the

27:18

others to dominate AI. I think

27:20

that would be a horrible mistake. So

27:23

let me ask a related question. The

27:26

information report, I thought it was the information. Oh

27:29

I guess it wasn't but somebody reported, I think I

27:31

thought it was the information, that OpenAI's annual

27:34

revenue reportedly... Well it's the information. You're talking

27:36

about another story that actually aggregated our work

27:38

which is the very phenomenon we're talking about.

27:40

You're like putting it the wrong way. Right.

27:43

Here it is. OpenAI's annualized revenue

27:46

tops $1.6 billion and then the rest of the headline is as customers shrug

27:57

off CEO drama. That's

28:00

a great story. Maria Heater, Amira Frade,

28:03

and Stephanie Palazzolo wrote that. So your

28:05

colleagues. So that went up at what

28:07

time, Leo? That went

28:09

up 7 a.m. on New Year's Eve

28:11

Eve. Right. So

28:13

on New Year's Day at minus 3

28:16

is 11 a.m. Silicon

28:21

Angle put up the story, which is the one

28:23

I accidentally put in the rundown, saying

28:26

the information reported. Yeah. Well,

28:28

we do that too, though. So I got to

28:30

point out. I mean, that's the whole show, basically.

28:32

The whole show. I, because of

28:34

that, I'm cognizant of that. I always, you notice,

28:36

I just read their names and

28:39

I quoted it. I

28:41

try to give credit back. But honestly,

28:43

that's another reason I have some concern

28:45

about this lawsuit, because we don't do

28:47

any reporting. I don't do any reporting. You

28:50

guys do. Certainly, you do, Perris. Perris does

28:52

it. I don't know what you do, Jeff. No, I don't.

28:55

But I don't do any. I

28:57

do zero reporting. Just out there.

28:59

Zero reporting. The entire extent of

29:01

my work is to

29:03

read all this stuff, digest it, and

29:05

editorialize on the talk about it. That's

29:08

what we do. AI. So

29:10

I think I'm in the same boat as

29:12

OpenAI. You are. But here's my question,

29:15

though. Who's

29:17

paying OpenAI $1.6 billion? I

29:20

am. I give them $20 a month. But

29:23

geez, at $20 a month, how many people are doing this

29:25

a lot? Yeah, it's not

29:27

given honest answers. How

29:30

are they getting that much revenue? I

29:33

don't get it. I mean, they're getting it from, I guess,

29:36

I would assume, obviously, this is not

29:39

reporting based, but I would assume that

29:41

actual users like Leo are probably

29:43

a small percentage of that. Yes.

29:46

Corporate clients are a large percentage. What are they

29:49

getting? Well, I can tell you one thing that

29:51

I get that is worth it. Actually, because by

29:53

the way, Microsoft has put ChatGPG

29:55

4 out for free on

29:58

iPhone and Android. You can get that. of

30:00

being chat app and do

30:02

everything. But what I like and

30:04

what I use and what is worth $20 a

30:06

month for me is the expert systems I've created

30:08

as GPTs. This

30:11

is by the way, I think in the long run, this is

30:13

what's going to happen is, you know, you

30:15

had this app store revolution with Apple's iPhone. I think

30:17

you're about to have an app

30:19

revolution with AI where people, because these

30:21

all have open APIs, you can license

30:23

it. I think that's where the majority

30:26

of money, by the way, comes from

30:28

is licensing the API. And

30:30

so I have a couple

30:33

of really useful expert

30:35

systems. And this is more just

30:37

for me. In that case, to learn this content

30:39

that you have uploaded, open AI didn't provide it,

30:41

you provided it. So this is how this works. And I'll

30:43

show you in the configuration. It's both. So

30:46

without an LLM, there's

30:50

nothing here to do, right? Right.

30:53

I'll go into my common lisper, which is the

30:55

name of this little expert

30:57

system I created. The

31:00

LLM, I could upload as I have all this

31:02

stuff, but it wouldn't be able to put it

31:04

together into a expert. So

31:09

there's both the corpus of

31:11

data, but it's on top

31:13

of a large model that

31:16

is generated by reading the New York

31:18

Times and the information and whatever else it can

31:20

to create an LLM. But

31:22

what I've told the LLM is don't hallucinate.

31:25

Only give me answers that come from it. I now have 10,

31:28

11 books in here. I'm

31:31

doing this with Common Lisp, and one of the advantages

31:34

of doing it with Common Lisp is it's so old

31:36

that there's a lot of public domain PDFs

31:38

of classic books by Paul Graham

31:40

and Peter Norvig. The

31:43

entire Common Lisp spec

31:45

is available as a PDF online, a

31:48

number of great books. So I just put all those things. I

31:50

already have them as PDFs. I just put them all into this

31:54

GPT. And it's really been useful. I mean,

31:56

it's incredible. I can ask it

31:58

a question that I would normally go... and search the web for,

32:00

I would search, you know, Common Lisp

32:03

for loop and then have

32:06

to sift through the Google results. Now

32:09

I can actually ask it to explain

32:11

it to me. Here, I'll show you. Explain

32:16

loops. Now it

32:18

knows it's Common Lisp, so I

32:20

don't even have to say in Common Lisp, it knows that. And

32:23

so it's gonna give me some sample code,

32:25

it's gonna describe this. This is like having

32:28

a teacher and it's all expert information

32:30

that comes directly from these books. There's

32:33

no who's there. Did you buy all those books? Some

32:35

of our public domain, Peter Narvig's book is public

32:37

domain, some I bought, yeah. But all of

32:39

them are in the public domain at

32:42

this point. I mean,

32:44

look at... Is there anything that stops you from offering

32:46

this to others? Well,

32:49

yes, interestingly, lately,

32:52

this is a

32:54

change in chat GPT. Public

32:56

actions required valid privacy policy

32:58

URLs. So I have to

33:00

create a... What does that mean? I don't know, but I have

33:02

to create a... I stopped, I make it only mean now because

33:06

it's so it's only for me. But in theory,

33:08

you could, if you wanted to have a little

33:10

business, create expert system

33:13

GP. Let's say, let's

33:15

say I am a BMW repair

33:17

shop and I have every manual

33:19

for every BMW ever made. If I can get

33:22

that to an expert system, that's

33:24

a huge value. Now here, this is the question, this

33:26

is why the New York Times is worried. Can

33:29

I then sell that expert system? That's

33:31

what I'm asking, right? Yeah. I

33:34

don't know. What do you think the answer is,

33:36

Leo? Well, I'll get...

33:39

I don't know the answer is I have

33:41

an Emacs expert, which I

33:43

don't distribute because

33:45

one of the corpus,

33:48

one of the things in the corpus here is a

33:51

is a book by a great guy named

33:53

Mickey Peterson that I bought. I have a

33:55

PDF of. I was able to

33:57

upload it mastering Emacs, but it's not

33:59

public... domain, he sells it. So

34:02

I would feel funny about selling

34:07

this expert system. I

34:09

don't have a problem with using it for

34:11

myself. By the way, this is why open

34:13

source is also so important, is I should

34:15

be able to do this all on my

34:17

own system with my own LLM, with

34:20

my own database, my

34:23

own corpus of knowledge, not

34:25

sell it, but just for myself. That's

34:27

where you're going to get some really interesting things happening. Yeah,

34:30

somebody's saying in our Discord, take all

34:32

of Steve Gibson's notes and put them

34:34

into an expert system. In

34:36

fact, I started doing that. I put all

34:38

the transcripts of Steve's shows, all

34:41

his show notes into an expert system. Now

34:43

you have an expert Steve Gibson. Now here's

34:45

a really interesting question. That's

34:48

all content Steve has made as part of

34:50

our podcast. He's published

34:52

it publicly. It's

34:56

like everything Steve ever said or knows, does

34:59

Steve have rights to that LLM,

35:02

to that GPT? What's the difference between that?

35:05

It's that. All right, so let's just play with it for

35:07

a second. So I can go

35:09

to the web and I can read all that

35:11

stuff laboriously myself, which would take a very long

35:13

time. I could hire someone,

35:16

a librarian, to go do that for me.

35:18

Or I could hire a machine. Well, to

35:20

some degree, a search engine also. A search

35:22

engine, right? Exactly what I'm saying. Yeah.

35:25

Search for the word. Search through all of the

35:27

transcripts, which people do all the time. For a

35:29

term, I want to know more about ransomware and

35:31

get all of those references. How

35:35

that's not so different from a

35:37

chat GPT saying, oh, summarizing it

35:39

in a paragraph. That's

35:41

the only real difference is same content. I

35:45

don't know. I mean, I don't know. This is

35:47

a very interesting. This is the first time your analogy

35:49

between this and the accelerationist potential

35:52

of the internet has ever really clicked for me.

35:54

So I'll give you that. Say more. Say

35:56

more. I mean, I think just in the sense that what

35:58

you're describing. a tool

36:01

in that can make

36:05

existing processes happen

36:08

at a much faster rate. More

36:11

than that. The technology already to, you know,

36:13

like go through all of Steve's

36:17

podcast episodes, read the

36:19

notes of what he said and get up

36:21

to date on certain topics already exists, it

36:23

would be laborious. But putting

36:25

all of that into a GPT, having

36:27

it summarize it for me, isn't

36:30

inherently all

36:32

that different. And here's the

36:34

real point, it's a society, a societal

36:37

good. Yes, it

36:39

is. If we can take the knowledge

36:41

of the world and make

36:43

it available, this is Google's mission

36:46

statement, right? To put

36:48

the information of the world at your fingertips.

36:50

If we could do it even better in

36:52

such a way that it's useful so that

36:54

you can query a GPT

36:57

about security and get

36:59

good responses from it, yes,

37:02

I understand the New York Times or Steve

37:04

Gibson or the information or Paris Merno might

37:06

say, but wait, that's

37:09

all, that's my knowledge. But

37:11

from a societal point of

37:13

view, that's a huge societal

37:15

benefit. And this is the

37:18

point of copyright in a nutshell, is to

37:20

give both sides a benefit, to give the

37:22

creator the right to

37:24

make money off of it, patent

37:27

same thing, but ultimately to make it into

37:29

the public domain at some point, right now

37:31

it's life plus 70, which is nuts, but

37:33

ultimately make it into the public domain so

37:35

that we can all benefit from it. And

37:38

I think that that's the thing I really want to

37:40

focus on. Even if you're

37:42

a creator, as part of that. Yes, you're a

37:44

creator. Once I've learned it, that is my

37:46

knowledge. You can't take it away from me. You can't

37:48

play the men in black pen against

37:50

my head. But here's another thing,

37:53

when I met Stephen Johnson originally about notebook

37:55

LM, what we speculated about is

37:57

why shouldn't the New York Times be offering that?

38:00

as a service on their own. They

38:02

should make a New York Times GPC.

38:04

Exactly. A publisher

38:07

of a book should make, this is something

38:09

that I said in what Google

38:11

do, we have to update the book and then in good

38:13

parenthesis I recanted that and said no, let the book be

38:15

alone. But I'll go back again. A

38:17

book publisher should be able to put up a

38:19

book in such a way that you can query

38:22

it. There you go. And you can ask the

38:24

questions. But they don't do that because they

38:26

said they don't know this is ours, you have to buy it in the format we

38:28

give it to you. Use it that way. There's

38:32

going to be no way to protect the book. There's going to be a little bit of

38:34

a devil's advocate here. This is a good question

38:36

in the chat from Andrew saying the thing is if

38:39

we are out there building all of

38:41

our Steve GPTs, someday who

38:44

will be Steve? Oh, somebody always has to

38:46

be. If we're just letting AI do

38:49

X and Y and Z and everything for us, are

38:51

we not making any experts in these key fields

38:53

anymore? No, of course not. There's always going to be a

38:56

Paris Martineau and a Steve Gibson. You're

38:58

going to still have value because you'll still be able

39:00

to get value out of what you do because you're

39:02

creating the original content. People will still

39:04

subscribe to the information. Well,

39:07

Paris is worried. Well,

39:09

I understand and maybe I should be too.

39:11

I understand it's a little bit of, I

39:14

think they're less vulnerable. Frankly, the class of

39:16

commentary as a class of

39:20

journalism which is very important and valid, it's what we're

39:22

doing right here, is more of

39:24

what should be concerned. I mean,

39:27

who is going to be listening to this?

39:31

You have to create. Yes, you know what? It

39:34

would be trivial frankly and it

39:36

will happen in a couple of years to do this show

39:39

without any of

39:41

us. It won't

39:44

have our jokes about your weird

39:46

glasses. Probably will. That would be the easiest part.

39:51

How do you keep your glasses from sliding down

39:53

your nose? I don't. They slide down my face

39:55

all the time and it annoys me. The first

39:57

show Paris, I thought that was an affectation. I

40:00

thought this was the look. I

40:02

thought that was the parents' knowledge. It's just

40:05

I really can't. I need to get them tightened. It's a

40:07

problem. No, I got them tightened. I think it's just the

40:09

nature. Partly is I don't have a lump in my nose.

40:11

You see, I have a perfect profile.

40:14

Oh, poor you. There's nothing

40:16

to hold it up, you see? You see

40:18

what I'm saying? You should get one of

40:20

those lumps installed. Yeah, lump in the phone.

40:22

But when I wear glasses like you, Jeff,

40:24

with the pads, right, to press in on

40:26

the nose, that hurts. But

40:29

it keeps it from sliding down. Well, also,

40:31

these are incredibly, incredibly light. These are titanium.

40:34

Oh, the minor light. My

40:37

other ones are light, too. And

40:39

these because I don't know, because they are heavy, I guess.

40:42

So in the Gutenberg, for instance, this part of me, I've got

40:44

to plug in once an hour. I

40:46

think that's twice now. I quote

40:49

the fight between Kevin Kelly and John Updike

40:51

in 2006. So

40:54

Kevin Kelly said just what you just said, Leo. Oh, you're

40:56

going to make it new ways. You're going to do new

40:58

things. It's going to be open.

41:00

It's going to be wonderful. And every book is linkable.

41:02

And everything is out of that, right? And Updike just

41:04

went crazy at Book Expo. And he

41:06

called that a pretty grisly scenario. He

41:09

said books traditionally have edges. The

41:11

electronic anthill. Where are the edges?

41:15

So booksellers, he said, defend your lowly

41:17

forts. And he said, I don't want to perform

41:20

for my lunch. I want to write for my lunch. And

41:23

I get that. I get it, too. That's

41:25

why we don't ask people

41:28

from the previous generation what

41:31

they think of the next big thing. Because

41:33

they never really understand what it's going to

41:35

be. And they have their natural aversion to

41:38

it. Because it's a change is

41:40

going to get in the way of it actually happening. And

41:43

you know what? There may be

41:45

disruptions. There almost certainly will be.

41:47

The internet costs huge disruptions in

41:49

businesses, including the New York Times

41:52

business. And they're desperately

41:54

scrambling to find a way

41:56

to survive. Many newspapers have gone out of

41:58

business. But I wouldn't want to. stop

42:01

progress to save newspapers any more than

42:03

I wanted to stop TV to save

42:06

radio. It's just or

42:10

stop cars to save buggy whips. Not

42:13

all progress is good. Exactly

42:15

the same way. Yes. They're going against open AI

42:17

now. Yes. Not all progress is

42:19

good but it is in the nature

42:21

of life and we have to move

42:23

forward. We cannot freeze this in Aspic

42:26

and so to ask John Updike who I

42:28

deeply respect, by the way, guess what?

42:31

There's people still read his novels. They

42:34

love his novels. His novels

42:36

have not gone away. Books still have

42:38

edges. Absolutely. So the

42:40

problem wasn't that you

42:43

know I understand his concern but he was from

42:45

another generation he didn't know

42:47

any more than I did at the time

42:49

what was going to happen and so to

42:52

ask him is to freeze progress in the

42:54

form that somebody from an earlier generation thinks

42:56

it should be frozen and that's a clear

42:58

mistake and that's my exact point with open

43:01

AI or AI in general is we don't

43:03

know and so it would be we see

43:05

it with music it would be wrong for us at

43:08

any point in this to say oh no

43:10

no no you can't let that happen because

43:12

we just don't know open

43:14

AI is the nap many have said this it's the

43:16

Napster of content

43:19

and Napster opened the door and yes all the

43:22

law came out what happened in the end is

43:24

the is the album got decommissioned

43:28

de-orbited right is that is that what the member used and

43:31

but there are far more

43:33

creators being able to be heard right now in

43:35

far more than always that were the case and

43:37

I want to point out Peter Gabriel

43:40

just released an album that

43:42

is an album that is in sequence and the

43:44

way he did it is he released

43:46

a cut at the full moon of every month in the

43:49

year 2023 in album order

43:52

and on December 1st released the full

43:54

album he preserved the album he found

43:56

a way to do it I mean

44:00

And by the way, I was really pissed because I

44:02

was having a hard time buying the album. I wanted to give him my

44:04

$17 on iTunes. And

44:07

Apple doesn't want you to buy albums. They want you

44:09

to subscribe forever to Apple Music. So they make it

44:11

very difficult to find a way to buy them. But

44:14

Jason Snell, who is a friend, the host

44:17

of MacBreak Weekly, and a fan of Peter

44:19

Gabriel, said, oh no, Peter sells all his

44:21

stuff on Bandcamp. Yeah,

44:25

Bandcamp is where they're at. You could go to

44:27

Bandcamp and you could still buy an album. So

44:31

here's a guy, he's 70-something. He's

44:34

an earlier generation musician who

44:36

believes in the form of albums and

44:39

has preserved it. I

44:41

think John Updike didn't go away, Peter Gabriel didn't

44:43

go away, but they should not be allowed to

44:45

say. There's a lot more John Updike and Peter

44:47

Gabriel's nothing to be had who couldn't make it

44:50

through the gauntlet before. That's true, too.

44:52

Okay, I agree, but I will

44:55

also say the current state of

44:57

the music industry, which is streaming

44:59

based, is not very hospitable for

45:02

new artists who are trying to make a living. Getting

45:05

pennies from Spotify is not going to produce the

45:07

next case. Well, we agree to see what they're

45:10

doing in the podcast. You're right. I

45:12

agree. We should go to Bandcamp and sell your album.

45:14

Bandcamp is still messing up. And so in

45:16

the book that's coming out next year, I'm not

45:18

sure we're not plugging yet, but

45:21

I argued that the mistake that

45:23

I made was that I gave too much

45:25

to companies. I gave too much to Twitter,

45:27

I gave too much to Facebook, and

45:30

then this is the lesson Leo taught me and it beat

45:32

into my head when it came to Macedon, is

45:34

we've got to honor the open structures

45:36

of the Internet. And

45:40

it's not always going to be

45:42

the easiest way or the obvious

45:44

way or maybe even the most

45:46

profitable way. But

45:48

there is also to be considered the interests

45:51

of society. And

45:53

I think it is in society's interest that we

45:55

allow this stuff to develop.

45:57

We keep an eye on it. there

46:00

will be bad things coming out of AI. We

46:02

know that. But there could

46:04

be so many good things. It's very much like

46:07

the internet. There's so many good things possible

46:09

as well. So I

46:13

really want a smash cut of all of your

46:15

different take things I know last year. I think

46:19

that would be pretty good. In the words, we could get AI

46:21

on that. In the words of Henry

46:23

Davis and Paris, foolish

46:26

consistency is the

46:28

hobgoblin of small minds. Am

46:31

I wrong? Actually was

46:37

Ralph Waldo Emerson. But other than that,

46:42

you hallucinated.

46:45

Yeah, humans hallucinate a lot more than

46:47

machines. Hobgoblin. Yeah, that's a

46:49

good word, isn't it? It's really good.

46:51

I think there's more hobgoblin references. That

46:54

was fun. That was

46:56

really fun. It's a really

46:58

exciting area. And it's one of the reasons, as

47:00

I've said many times, we

47:03

have a job to do. And so this

47:05

is why I'm not worried about AI. Because

47:07

you and I have

47:09

a job to do, Paris. And

47:12

Jeff, I guess, have a job.

47:14

Until August? No, no. We have

47:16

a job to do, just

47:19

to understand this and explain it. And

47:22

also advocate. I mean, that's why I'm not

47:25

a pure journalist. We advocate. We

47:28

do our best. And I love

47:30

I've done this my whole career since

47:32

the early 90s. Because I love technology.

47:34

But I also feel very

47:36

strongly that we have that there are certain

47:38

paths we should avoid. So Beth, I've always

47:40

been a big believer in open versus proprietary.

47:42

And I've always flogged that. Now, that's my

47:44

point of view. Maybe people

47:47

don't agree with me. There's plenty of them. But

47:50

that's part of our job. And I think we'll have

47:52

a job to do. And AI cannot advocate. And

47:55

AI doesn't really, can only give

47:57

you information. And then it's not well

47:59

and not often not well, but

48:01

then me too. Emerson

48:04

Thoreau. It's all, you know, the

48:07

same. Poor Michael Cohen. Michael Cohen. It

48:09

all is the same. Yeah, that's right.

48:11

Michael Cohen used Bard. He did it

48:14

too. To write his pleading. Well, you

48:16

know what? He's a doofus. He's a

48:18

doofus. But I also

48:20

could see how in Bard, Bard now appears

48:22

right next to the Google search bar, and

48:25

it takes the place of the old Wikipedia

48:27

information up there. And so you ask

48:29

a question in the search, and there is information right

48:31

there in Bard. And if you don't, if you're not

48:33

paying attention to the news, which he doesn't accept about

48:35

himself, I could see

48:37

the confusion. He's not his lawyer,

48:39

is another case. Is that a

48:41

good one? Yeah. The context is

48:43

Michael Cohen says that he unwittingly

48:45

passed along to his attorney bogus

48:47

artificial intelligence-generated legal case citations that

48:50

he got online before they were

48:52

submitted to a judge, which

48:54

is pretty funny, honestly. It is. It

48:57

is. Okay.

49:00

It's crazy, though, that they ended

49:02

up cited as part of written

49:05

arguments by his attorney. His attorney

49:07

didn't do his job, and then the attorney, there was all, it

49:09

was right before anybody else was listening. Well, I mean, if you're

49:11

the attorney for Michael Cohen, you've got a lot of things in

49:13

your plate. You're a busy guy.

49:16

Honestly, I mean, I don't know enough about

49:18

the job that these people do, but I

49:21

bet you there is some time pressure, and

49:24

you're working really hard to pull these

49:26

things together, and maybe you

49:28

should have checked. Well,

49:30

Michael doesn't have access to Westlaw anymore because he's not

49:32

a lawyer anymore. Oh, that's right. He isn't a lawyer,

49:34

is he? He's been to Spar. All right. Right.

49:37

Well, we'll talk about Bard, actually, when we come back. This is

49:39

this week in Google. The show where

49:41

we cover some... And Bard is a Google

49:43

product. Bard is a Google product. Yeah, here

49:45

we are. We're getting into Google before the

49:47

ad break. That's huge. How about that? We're

49:49

going to talk about Bard. It's getting close.

49:52

So if you want to use it in your trial,

49:55

stay tuned. But first, a

49:58

word from our sponsor. I

50:00

love these guys. Collide. And why

50:03

do I love Collide? Because they

50:05

love end users. When

50:08

you go through airport security, you know that

50:10

line where the TSA agents check your ID?

50:13

And the other line where a machine scans your bag?

50:17

Try to picture this, because this is a great

50:19

analogy for what happens in enterprise security. It's not

50:22

passengers and luggage. It's end

50:24

users, and it's their devices. Right

50:26

now, these days, most companies are pretty good at the

50:28

first part, where they check user identity.

50:31

Perhaps you're using Okta to do that.

50:34

That's great. It's strong authentication. The problem

50:36

is, there's the assumption you authenticate the

50:38

user, oh, their bags are fine. But

50:41

that's not necessarily the case. User

50:43

devices can roll right through that

50:45

authentication. They're not getting inspected. In

50:48

fact, right now, 47% of

50:51

companies allow unmanaged, untrusted

50:53

devices right in to access

50:55

their data. That

50:58

means an employee can log in from a laptop

51:00

that has, let's say, its firewall turned off, hasn't

51:02

been updated in six months. Or

51:05

worse, a laptop might belong to a

51:07

bad actor using employee credentials. They

51:10

may be able to get in, but the laptop got in

51:12

with them, and you are in trouble. Collide

51:14

solves this device trust problem. Ensures

51:16

that no device can log in

51:19

to your Okta-protected apps unless the

51:21

device passes your security checks. Plus,

51:24

and this is great, too. You

51:26

can use Collide on devices without MDM. Your

51:29

Linux fleet, contractor devices, you can't force

51:32

them to put MDM on their devices.

51:35

Every BYOD phone and laptop in the company,

51:37

you know the executive's going to bring that

51:39

phone in, but it still works.

51:43

collide.com/twig. Watch

51:45

a demo, see how it works. k-o-l-i-d-e.com/

51:50

twig. You need this. Collide. We

51:53

thank Collide for supporting this

51:56

week in Google. You

51:58

want to see the BART? is

52:00

inching towards launch. I

52:02

think that's an interesting choice of words from 9 to 5

52:05

Google, inching. Apparently,

52:08

where it's imminent, right? This

52:11

is from APK Insight, where they take applications,

52:15

files, the APKs, and decompile them and

52:17

analyze them. And inside, they found a

52:20

little pop-up that's

52:23

an Assistant with Bard tab right there in

52:25

the Google app. This

52:28

is designed to fully replace the existing

52:30

Google Assistant on

52:32

supported Android devices. So for right now,

52:34

you've got your Pixel. And

52:37

you go, hey, you know who? It's

52:41

going to be Bard's going to be in there in some

52:43

way. The

52:45

company is also planning to prominently place the

52:47

Assistant with Bard experience on the Discover page

52:49

of the Google Search app. Here

52:52

is from 9 to 5 Google

52:55

and APK Insight

52:58

the video. What

53:01

appears to be ways to quickly switch between

53:03

performing a normal Google search and

53:05

getting help from AI? See, Michael Cohen's going to love

53:07

this. Here's a

53:09

here. That's going to be huge for him. This is going to be huge.

53:13

So there's the Assistant. Hi,

53:15

I'm Assistant with Bard. Oh,

53:17

like that? Or here's like you're in,

53:19

what would you like to do today? Would you like to see

53:21

me type? Or you talk? Hello there. Can

53:23

you tell me what's Santa? I'm

53:26

sorry, what's on the image, please? I

53:28

don't know where Santa came in. It said Santa.

53:30

It did. It did say Santa. Yeah. And

53:32

then it's looking at the screen. And

53:35

this is Bard. APK

53:37

Insight says for now, it's not clear whether this

53:40

is intended to be a permanent fixture of Discover

53:42

tab. Or maybe it's just

53:44

a one-time only little ad for checking

53:47

out Assistant with Bard. Then

53:51

there's a guy named Dylan Roussel who managed

53:53

to enable the actual pop-up window. I love

53:55

people. It's

53:58

hysterical. and

54:00

they dig and shows us

54:02

how that's going to look to use

54:05

Bard to submit questions. So we're getting

54:07

closer and closer. I have to say

54:09

I pay $20 bucks as I mentioned

54:11

a month to have chat GPT on

54:14

my phone and on a

54:16

new iPhone I'm able to use their action

54:19

button to launch it and talk to

54:21

chat GPT and have it respond back. I showed you that

54:23

a few weeks ago. Now

54:26

we're getting, I think

54:28

it's really interesting that Google looks

54:30

like it might want to add Bard to the

54:32

regular Google Assistant. So

54:38

do you also pay for an anthropic to do the

54:40

work of the guy? I think we must, right? We

54:44

use Anthropics AI Claude, which

54:46

by the way, by informant deep

54:49

within the industry said,

54:51

oh those guys at Anthropic they're jerks. He

54:54

said, oh you're a long walk guy? My

54:57

long walk guy. He said, you're a long

54:59

walk with Sam Altman? Yeah, yeah. He

55:01

said, those are the guys who are

55:04

so concerned about safety that they left

55:08

Google because they were so worried about safety.

55:10

He said, so they've created, Anthropics created this

55:13

safe AI. He

55:15

was very dismissive of it. But I have

55:17

to say it's been very useful for us. But

55:19

here's my question. Yeah. I'm

55:21

sure we pay for it. Anthony, do we pay

55:23

for that? Yeah, we must. What do you use

55:26

it for? We

55:29

use it to do show notes. We

55:32

also, I believe we use it to chop up

55:35

the show into little TikTok-like

55:38

vertical videos. That's a different

55:40

tool. What tool is that? That's podium. I'm

55:42

sorry, podium is what we use for the transcript and stuff like

55:44

that. These

55:47

are all tools you pay for. All AI tools we

55:49

pay for. But

55:52

here's my question. We use a lot of them. We use blog posts

55:54

out of cloud as well. Cloud makes blog

55:56

posts. Yeah, but you write them. Some point.

55:59

Well, we edit them as humans. But

56:01

yeah, it'll take a show, it'll listen

56:03

to the whole show, it'll read

56:06

the transcript of the show and then like give us

56:08

like a summary. Let Claude do it. I don't want

56:10

Claude to do it. No, I'll tell you, this is

56:12

historically a huge problem for podcasts because

56:15

Google doesn't search the audio. They always said

56:18

they were gunno and they never really did.

56:20

So there's no discoverability in a podcast. So

56:22

we've always said what we really should do

56:24

is a blog post for every episode. But

56:26

I'm not going to do that.

56:29

None of our hosts wanted to do that. That's

56:31

just, you know, that's adding insult to injury. So

56:34

the AIs now take this show, they

56:36

transcribe it and they generate an article

56:38

from it and then the

56:40

producers look at it and clean it up.

56:43

Where is that? Is that in the blog on the Twitter

56:45

TV? Yep.

56:48

So there's actually a lot of stuff in here

56:50

that we got to make this more discoverable. These

56:53

are transcripts. There's the newsletter there.

56:56

Micah Sargent's Holiday Gift Ideas. Micah probably

56:58

wrote that. The

57:01

transcripts are created by Podium. Oh,

57:04

here you go. Look at this. Remember

57:07

when we had Steven Johnson on the show?

57:10

So we fed that show to an

57:12

AI. To

57:14

his AI? Different AI. No,

57:16

different AI to who do we use for

57:18

this? Claude? To Claude? Yeah,

57:21

Claude. Claude wrote this and

57:24

then Anthony Nielsen, our AI guru, but

57:26

it's sometimes the producer, sometimes Anthony, went

57:29

through it, made sure it was legit. How much do

57:31

we pay for this feature? What

57:34

am I looking at? This

57:36

is Claude reading the blog post. We

57:40

don't pay for that. Claude just does it

57:42

for free? That's my question. That's free.

57:44

Pretty soon, everything you've mentioned is going to be available

57:46

for free because they're all going to be competing and

57:48

you're not going to be paying for it. The internet was

57:50

like that too. Look, it's even got a quote. It's

57:53

free for like three to five years. Yeah, and then we'll have to all pay for it. $25

57:55

a month or probably like $100 a month. Yeah,

57:59

and we'll be back. pissed. This is

58:01

actually pretty good. This is

58:03

a summary of what Stephen

58:05

said. It's got a quote from Stephen. But

58:08

Anthony, how much cleanup did you have to do

58:10

on this, Anthony? It's

58:17

almost there. He just touches it up a little bit

58:20

and makes sure there's no lies. It works out

58:22

by names and by quotes. I want to

58:24

take everything from Jeff. That's his instruction. And then

58:28

we were going to use it for show notes, right, with

58:30

the emoji bullet points. Have we started doing that yet or

58:32

no? Not emoji bullet

58:34

points. I like the emoji bullet points. The metric

58:36

said it would break things. Oh, it would break

58:38

things. Oh, well, that's not good. So

58:41

we don't use it, but it does make the show notes

58:43

here. It depends on the producer.

58:46

Okay. And then what

58:48

do we use? What's the tool we... Do we pay for

58:50

that feature? Okay. Oh,

58:53

no, we just run that through Claude. Claude

58:56

is free and free, free, free?

58:58

No, no. No. Well, you have

59:00

eliminated usage in the like... Okay.

59:02

So if we use it more than a certain amount,

59:04

we would have to pay for it.

59:07

Oh, so you're paying for it. Yeah, I'm

59:09

paying $20 for chat GPT. So, so

59:11

20 bucks. And then, podium

59:13

is much more expensive, I'm sure. That's

59:15

our... does transcriptions. And then,

59:18

and then what is the... What

59:23

is the thing we use to make the

59:26

TikToks? TikToks. Opus.pro.

59:29

And that we pay for. Yeah. And

59:31

that's like it's, you know,

59:33

for an episode, it's like, you know, five bucks and you

59:35

get... It'll like generate, you

59:38

know, 30 clips that you could kind of

59:40

pull from and edit and... It does a

59:42

pretty... It also adds the text,

59:45

right? And it'll automatically like,

59:47

you know, deal with the... There's one of

59:49

our shows. There's you, Paris. Hey,

59:52

there's me. That's

59:54

cool. Never noticed that. So,

59:57

yeah, that is cool. There's Scott

59:59

Galloway, I think. Yeah. Ugh. Ugh.

1:00:03

There's young and profiting. This

1:00:08

is great because honestly we don't have the manpower.

1:00:11

You know my son Henry, he looks

1:00:13

at our TikTok and says, Dad, would you just

1:00:16

fire whoever's doing that? I said,

1:00:18

we can't, it's an AI. He said, Dad, that's terrible. I said,

1:00:20

it's not terrible. She said, look, let me do it. I said,

1:00:22

no, because I can't pay you. What you would need to get

1:00:24

paid to do this. But his contention, and of

1:00:26

course he's been very successful in TikTok and it's not

1:00:28

with millions

1:00:31

of followers, but his

1:00:33

contention is it's not really creating really

1:00:35

truly viral stuff.

1:00:38

I will say, I mean. It's better than not

1:00:40

being on the platform at all, but it's not probably going to

1:00:42

get people to do it. I mean, you know, it's not going

1:00:44

to be the same. But I think it's going to be a

1:00:46

very, very, very, very, very, very, very, very, very, going

1:00:52

to get people to check

1:00:54

out to it from TikTok. But nothing

1:00:56

we would do would do that, right?

1:00:58

That's the problem. I think it would.

1:01:01

I think it's a really popular

1:01:03

format on TikTok to kind of

1:01:05

have cuts from podcasts

1:01:09

by kind of going back and forth. Yeah. And

1:01:12

it does drive subscribers. I mean,

1:01:15

that's how Dropout, that a

1:01:17

like a streaming service I've talked about

1:01:19

before, gotten the vast majority

1:01:21

of their subscribers of the last couple of

1:01:23

years is from TikTok. But that would also

1:01:25

require hiring someone. All Henry had to

1:01:28

do is say, well, how many views around that? How

1:01:32

many followers do you have? How

1:01:35

many likes do you have total? Yeah.

1:01:40

So, I mean, we like, we

1:01:42

recently just started. You don't have to defend it.

1:01:45

You don't have to defend it. He's

1:01:47

a snob because, but I'll tell you what the

1:01:49

real thing is, and he doesn't admit this, even

1:01:52

those in his heart of hearts, the

1:01:54

real audience is not humans. The

1:01:57

real audience is the TikTok algorithm. Because the

1:01:59

way you you actually get views on TikTok

1:02:01

is by TikTok surfacing it in the For

1:02:04

You tab. So until you

1:02:06

get the TikTok algorithm, not a

1:02:08

human, but the algorithm to notice

1:02:10

you, it doesn't matter. Yeah, but

1:02:12

that's the term human, like... We

1:02:14

don't know how the TikTok algorithm works. I

1:02:16

don't think it's just from human attention. I

1:02:20

think TikTok's looking for certain kinds of

1:02:22

content. Okay, I'm sending

1:02:24

you guys an example on Discord of a

1:02:26

podcast that I can't speak to the quality

1:02:28

of any of these videos. I just know

1:02:31

that I've seen this in my feed

1:02:33

beforehand, and it is kind of talking

1:02:35

heads, people recording a podcast, but they

1:02:37

do frame the shots in a way, and their

1:02:40

videos out? 3.4 million followers. The

1:02:44

basement yard. I

1:02:47

have no idea what it's about, but I... So,

1:02:51

oh, this looks kind of like what we do

1:02:53

where it's a vertical, you've got, you

1:02:55

know, the captions. The problem

1:02:57

is, we have to

1:03:00

say something interesting. I

1:03:02

think we say a lot of interesting

1:03:04

things. I think you just got to think of

1:03:06

what's TikTok-able. You have to

1:03:08

really play to the algorithm. Maybe that

1:03:10

is ultimately in the people, and there's a lot of

1:03:12

competition. Which does take us to line 93. Wow.

1:03:18

Okay, now this is a point, just

1:03:20

a little behind the scenes here for everybody, where

1:03:23

I have to decide before looking if it

1:03:27

is worth completely stopping, even just

1:03:29

talking about, and go to

1:03:32

line 93, which could be

1:03:35

Jeff Jarvis and a

1:03:37

TikTok with a monkey, or we

1:03:39

just don't... No

1:03:41

one knows. Oh wow, this is... It

1:03:43

is a TikTok. Yeah, it is a TikTok,

1:03:45

and it is relevant. You see? It

1:03:47

is relevant. It's not seen from first.

1:03:49

But was it worth... He scolding me people. He's scolding

1:03:52

me people. I don't care. Changing the

1:03:54

entire thrust. I don't care. Here

1:03:56

we go. We're doing it. When guys start the podcast for

1:03:58

no reason, it is really... We're the podcast

1:04:00

boys. Put the mics down! Put the mics

1:04:02

down! Don't even think about it! This episode is over

1:04:04

fellas. We're the podcast boys. Anything

1:04:07

you say can and will be used against you. Subscribe to the Patreon. Alright,

1:04:09

I'm out! It's become a real problem here in Los Angeles. It's

1:04:13

pretty good. Put the mics down! Put

1:04:15

the mics down! God damn it, they fled! I'm

1:04:18

sorry. I'm sorry. I'm sorry.

1:04:21

I'm sorry. I'm sorry. I'm

1:04:23

sorry. I can see that it has 634,000 likes.

1:04:47

Yeah. It's ridiculous. So this

1:04:49

is why I don't, honestly

1:04:51

don't feel like we should try to compete

1:04:53

on TikTok. Now I

1:04:56

thought we did a good job with

1:04:58

your little gale.com thing on TikTok. Did

1:05:01

you see that? What we did with that?

1:05:03

I didn't, but I do follow you on

1:05:06

TikTok. Alright, here is, this is from an

1:05:08

episode of Twig. Oh my God, I'm having

1:05:10

to watch. My pick of the week this

1:05:12

week is one of my colleagues accidentally

1:05:15

mistyped Gmail the other day.

1:05:18

Okay, so Henry would say this was. It took too

1:05:20

long. I've flipped. I've flipped up. I'm

1:05:22

gone. I've flipped and also

1:05:25

we got to get rid of the little,

1:05:27

on the bottom third of the TikTok is

1:05:29

a subscribe to Twit sort of banner thing.

1:05:31

You can't click. You got to get

1:05:33

rid of that. And you can't click it. You can't

1:05:35

click. There's no, it's not clickable. It's just a thing.

1:05:37

This is great for AI. And I think we're doing a really good job.

1:05:39

By the way, as an example, I bet

1:05:42

you anything that the TikTok algorithm sees

1:05:44

that and goes, eh, we're not going to put that on the for

1:05:46

you page. Probably. It's

1:05:48

stuff like that. It has nothing to do with how many

1:05:50

people saw it or whether they liked it or clicked it

1:05:53

or said anything. Henry says it doesn't matter how many people

1:05:55

like you. That is not part

1:05:57

of the algorithm. And I think the algorithm does things

1:05:59

like, yeah. I don't like the way that's laid out.

1:06:07

It should be illegal to make me listen to myself

1:06:09

on this show. I know. It's

1:06:12

not painful. I hate it. I know

1:06:14

what you mean. It does have a different... I

1:06:17

think this was a really good segment and a great thing

1:06:19

for a TikTok. But of course

1:06:21

it's not going to succeed on TikTok. It's

1:06:24

not going to grab you.

1:06:26

You're going to swipe. Yeah,

1:06:29

I think it would probably require some of your

1:06:32

son's fast editing to make it work. He

1:06:34

needs knife work. We need those. We need

1:06:36

the crackling round of bacon. We need knife

1:06:38

work and frying bacon. Oh yeah. Look

1:06:42

at this poor guy. I've got to tell you. He's

1:06:45

doing very well, but he works morning

1:06:47

till night making these stupid

1:06:49

sandwiches nobody eats. I said,

1:06:52

Henry, how come you not be

1:06:54

fat? He says, I don't eat

1:06:56

that stuff. I

1:06:58

said, well, do you... So he's doing it in his mom's

1:07:00

kitchen. Actually, he's moved back down to LA, but he was

1:07:02

doing it in his mom's kitchen. Do you give it to

1:07:04

mom? He said, no. I said, I

1:07:06

don't want to. Jennifer said, yeah, usually I

1:07:08

have to do over eat. I said, I can't get

1:07:10

into the kitchen. And then I said, well, who do

1:07:13

you give it to? You never give it to me.

1:07:15

He said, yeah, I have friends. I

1:07:18

don't understand. But anyway, that's neither

1:07:20

here nor there. It's hard.

1:07:22

Let's put it this way. It's hard to go

1:07:24

viral. There's no magic thing. But

1:07:27

one of the tricks really, at least Henry's

1:07:29

trick, is you've got to... Your audience is

1:07:31

an audience of one that... And it's a

1:07:33

machine. It's a TikTok algorithm. Don't

1:07:37

you think that's true? Speaking of... Yeah,

1:07:40

I mean, absolutely. Aren't you glad

1:07:42

we did that podcast police segment,

1:07:44

Jeff, that really forwarded the conversation?

1:07:48

I thought it did. Line 93. I

1:07:50

think it did. Line 93, guys. Speaking

1:07:52

of which, I think we could talk about Line

1:07:54

98, which we talked about a bit before the show.

1:07:56

Now you've got parents doing it. Listen.

1:08:00

And I'm going to lean in. He hates this. You

1:08:02

remember weeks ago you may have heard of

1:08:05

the TikTok tunnel girl, a

1:08:08

spiritual cousin of the TikTok ill

1:08:10

guy. She was a woman

1:08:12

who was digging a series of tunnels

1:08:14

underneath her Virginia home. She's

1:08:16

been ordered to stop doing that because

1:08:18

it's against the law. How

1:08:22

long did she go and how far

1:08:24

did she get digging these tunnels? Probably

1:08:26

longer than a year because the video I played

1:08:28

for you guys a couple of weeks ago was

1:08:30

a one year recap of her tunnels and

1:08:33

their cavernous. She had to get a

1:08:35

mine car set up in there. Apparently

1:08:37

pools of water. Is

1:08:39

she an engineer? She

1:08:42

is an engineer. She's a software

1:08:44

engineer. Oh, that doesn't count. They

1:08:47

don't teach you in the computer science

1:08:49

program how to keep the roof of

1:08:51

a tunnel from collapsing in on you.

1:08:54

Wait a minute. Wait a minute. Is that

1:08:57

a true? That's not actual Barack Obama

1:08:59

coming to that. He didn't really say

1:09:01

you are my Beyonce? Hero? Yeah,

1:09:04

no. No. Okay. All

1:09:06

right. That's somebody that doesn't even look like

1:09:08

Barack Obama. No,

1:09:11

it's someone on the internet lying about who they

1:09:13

are. But

1:09:15

for some reason the New York parts decided to

1:09:17

include it. Yeah.

1:09:20

So it was an

1:09:23

unpermitted tunnel digging project in a suburban

1:09:25

Washington, D.C. home. And it's now been

1:09:27

slapped with a bunch of potential violations.

1:09:30

What I'm curious is, did

1:09:32

she go under neighbor's property? She

1:09:35

had to. Otherwise it seems bigger than that. I

1:09:37

think she was building, if I recall correctly,

1:09:40

it could be a storm shelter. She

1:09:42

says it was a storm shelter. Yeah. All

1:09:44

right. Now we've moved on to something else.

1:09:47

Yeah, okay. At least you have to scold

1:09:49

parents like you scolded me. No, no.

1:09:52

Yeah. I wanted us. Yeah,

1:09:55

exactly. We're going to get in trouble. Yeah,

1:09:58

exactly. Both kids have to. Right. Oh,

1:10:01

she did post a video of her getting

1:10:03

shut down by the police, which is great

1:10:06

content. I'm sure Sorry,

1:10:08

that is an exercise for

1:10:10

the listener Go

1:10:13

to tick tock. What is your tick tock handle?

1:10:16

Oh, I have no idea. Okay Search

1:10:18

tunnel girl. She'll come out. I'm sure

1:10:20

girl. Yeah The

1:10:24

one plus buds three No,

1:10:27

I don't want to do a story about that. No you put this

1:10:29

on there's the folks You playing

1:10:31

at home. This was the first thing on

1:10:33

the rundown He's

1:10:38

going after us for going down Three

1:10:46

reportedly get price cut That'll

1:10:52

be a real conversation starter Particular

1:10:56

order in fact they are in

1:10:58

chronological order so as you can see

1:11:01

that story was bookmarked yesterday So it's

1:11:03

old too Yeah,

1:11:08

January 2nd You

1:11:11

bust us we bust you well, I'm just

1:11:13

saying I haven't put this in

1:11:15

any order you're assuming Yeah, yeah, yeah,

1:11:17

the first line is the most important. No,

1:11:19

that's not true. In fact, I didn't

1:11:21

lead with this did I led with? line

1:11:26

In our part of the thing. Yeah, yeah,

1:11:28

that should have been in there too. No, it

1:11:30

is It's actually line 51 as well in my

1:11:33

area, but it was in the AI area Anyway,

1:11:36

oh, it's right. It was a little too much

1:11:38

inside baseball. So what do you have to say

1:11:40

about those? one-plus buds

1:11:42

Leo Their price actually

1:11:44

ended up getting cut Deleting

1:11:48

now a

1:11:52

First time we've shamed Leopter a

1:11:54

moving row of the

1:11:56

Google sheet is now a mutual

1:11:58

shaming society. We're We're

1:12:00

all seeing the Howard Sitter Show. Apps

1:12:04

will be reporting your earnings to tax authorities

1:12:06

starting this week. How about that one? If

1:12:09

you make money with an Airbnb

1:12:11

or Etsy or eBay. I'm surprised

1:12:13

they haven't been. I am too.

1:12:16

Don't you have to give a, what do you call it, a $9? $10.99

1:12:18

or something? Yeah. Well, no. That's

1:12:21

what they give you. What do you have to give them? A

1:12:24

W-9. A W-9. Right. Yeah.

1:12:28

It doesn't mean you have to pay income tax

1:12:30

on it. In fact, the 9-5-MAX says whether or

1:12:32

not you have to pay tax on this income

1:12:35

depends on an often complex set of rules which

1:12:37

vary by country. But

1:12:39

this is a global agreement. So it's not just

1:12:41

in the U.S. The 38

1:12:43

members of the Organization for Economic Cooperation and

1:12:46

Development have all agreed, including

1:12:48

the U.S., U.K., many European countries, have

1:12:50

all agreed that these platforms have really

1:12:52

got to tell us if you're making

1:12:54

money. Here's the hard

1:12:56

part of that. So I wrote

1:12:58

a piece for some Spanish magazine. They're going

1:13:00

to pay 300 euros. Fine. I'll

1:13:02

take the 300 euros. But oh my God, I

1:13:05

have to get IRS documents to prove that

1:13:07

I'm an American, do not get

1:13:09

the Spanish tax, take it out. Oish.

1:13:12

That's a lot. A lot. Did

1:13:15

you see Anil Dash? By the way, I've

1:13:17

been attempting to get a hold of Anil, but

1:13:20

he's been on the show many times before in

1:13:22

the past. His Rolling Stone article entitled,

1:13:24

The Internet is About to Get Weird

1:13:26

Again. I think that's

1:13:29

weird as in Austin weird, good

1:13:31

weird. You mean good weird. Yeah.

1:13:33

Yeah. Anil is a

1:13:35

serial entrepreneur. He was Gina

1:13:37

Trapani's boss. I think that's how. And we used

1:13:40

to have him on the show frequently. I've known

1:13:42

Anil for years, probably a decade or more. Oh,

1:13:44

I've known him for probably longer than you have

1:13:46

because we go way back in blogging land. Yeah,

1:13:49

he's a proto blogger. Yeah.

1:13:52

His point, which is I like it, is that

1:13:54

2024 is going to be a watershed.

1:14:00

for the internet that the big tech

1:14:02

companies, the tech giants, thanks

1:14:05

to the EU primarily but somewhat to the

1:14:07

FTC, are being forced to

1:14:09

hold their noses and embrace mandated

1:14:11

changes like I'm reading his

1:14:13

prose here, like opening up their devices to

1:14:16

allow alternate app stores to provide apps to

1:14:18

consumers. Back in the

1:14:20

US, a shocking judgment in Epic Games lawsuit

1:14:23

against Google leaves us with the promise that

1:14:25

Android phones might be opening

1:14:27

up in a similar way. Twitter's

1:14:31

slide into irrelevance and extremism

1:14:34

has hastened the explosive growth of a whole new

1:14:36

host of newer social networks,

1:14:38

including he's a Mastodon

1:14:40

user. And he

1:14:42

mentions Mastodon and Blue Sky

1:14:44

and threads. He

1:14:49

says, and I think this is an interesting point

1:14:51

of view, that he sees it going back somewhat

1:14:53

to its roots. I hope he's right.

1:14:58

We're going to still try to get him on it because I'd love to

1:15:00

hear what he thinks about this. He

1:15:03

says, I'm not a Pollyanna about the fact that

1:15:05

there's still going to be lots of horrible things on

1:15:07

the internet and that too

1:15:09

many of the tycoons who rule the tech

1:15:11

industry are trying to make bad things worse.

1:15:15

There's not going to be some new killer

1:15:17

app that displaces Google or Facebook or Twitter

1:15:19

with a love powered alternative, but that's because

1:15:21

there shouldn't be, and Neil Dash

1:15:25

writes, there should be lots of different human

1:15:27

scale alternative experiences

1:15:29

on the internet that offer up home

1:15:32

cooked, locally grown, ethically sourced code to

1:15:34

table alternatives to the factory farm junk

1:15:36

food of the internet. And

1:15:39

they should be weird. This is what Kevin Marks has

1:15:41

been saying. The indie web. What

1:15:44

you've been saying. That's what I've been saying. You've been

1:15:46

saying it. Honestly, this also applies to our previous conversation.

1:15:48

That's why I don't want big tech to run up

1:15:50

AI. I want AI to be

1:15:52

open and available and weird. AI should be

1:15:55

weird. That's where innovation happens

1:15:57

is at the interface. So

1:15:59

I like this. article in the Rolling Stone it's kind

1:16:01

of an opinion piece I guess but a

1:16:03

Neil Dash yes

1:16:06

labeled commentary yeah so

1:16:09

we'll try to get a Neil on here's a

1:16:11

but to find it in Rolling Stone I was

1:16:13

glad that's interesting well Rolling Stone has become really

1:16:15

a kind of a political opinion journal

1:16:18

more than anything cheaper than reporting

1:16:20

yes Noah Shackman took over

1:16:23

it's yeah become more Daily

1:16:25

Beastie yeah and what is his

1:16:27

background again I'm trying to remember he

1:16:30

was he bought

1:16:32

it right chief of Daily Beast okay

1:16:36

and before that he was at wired

1:16:38

and somewhere else yeah

1:16:40

Rolling Stone which I think

1:16:42

Jan Wenner sold it yeah

1:16:44

but it's not just he didn't own it he

1:16:47

was just installed as the head

1:16:49

of it right it's brought up just all

1:16:51

kinds of entertainment media trade

1:16:55

and retail

1:16:57

which is bizarre cuz Penske is to

1:17:00

me is a motor racing company

1:17:02

well that's the way he's like

1:17:05

the song now Penske is a

1:17:07

media now he is yeah recently

1:17:09

invested a hundred million into Vox

1:17:11

at a time when oh that

1:17:13

was already having issues so

1:17:16

a big change so do you think

1:17:18

the Rolling Stone is more like the

1:17:20

Daily Beast now that Shackman is in

1:17:22

charge I mean editorial wise absolutely yeah

1:17:24

I mean Noah

1:17:26

is very good at what he

1:17:28

does and he smartly decided

1:17:31

that the Daily Beast model

1:17:33

was working and what Rolling

1:17:35

Stone was doing before was not and

1:17:37

brought over a lot of the smartest

1:17:39

political writers and reporters from the Daily

1:17:42

Beast to Rolling Stone in addition to

1:17:44

kind of revamping their cultural coverage yeah

1:17:48

he was a you go to PMC

1:17:50

calm I don't want to really

1:17:52

you're ready to leave Penske if you go

1:17:54

to me and calm there will be a

1:17:56

tick-tock on the website Really?

1:20:01

Remember that? Yeah. Wow. The

1:20:04

Firefly mobile phone for children?

1:20:06

So this

1:20:08

is really interesting. So Roger Penske, who is an

1:20:11

IndyCar driver, became,

1:20:15

took his renown as

1:20:17

a racecar driver to

1:20:19

found the Penske Corporation. Wow.

1:20:23

Very interesting. That

1:20:26

wasn't a bad detour that was an old

1:20:28

change. Oh, it's a weird detour. But I

1:20:31

brought it up because I said, what's happened

1:20:33

to the Rolling Stone? And this is all

1:20:35

from Anil Dash's interesting piece. All

1:20:37

right. Let's take a little break because

1:20:39

I'm breathless. You're

1:20:43

listening to This Week in Google with the

1:20:47

the the Chordling Anderson

1:20:49

Cooper style. All

1:20:54

we need is a cat cafe and we've got a

1:20:56

maid and there's more

1:20:58

to the information. Yeah.

1:21:01

Where's your cat? I was about to

1:21:03

say, my cat is clearly sleeping on

1:21:05

her duties. She's got to be on the screen for this.

1:21:08

I think we're talking to a new

1:21:10

sponsor that makes one

1:21:13

of those kitty layers that rotates.

1:21:15

Ooh. Do you want that? An

1:21:17

automatic cat cleaner? Yeah. Those

1:21:20

are nice. Yeah. Okay. They're

1:21:22

very expensive. They're okay. They're high quality things. All

1:21:24

right. They're a long

1:21:28

time. I

1:21:31

love the hearing off mic. I

1:21:33

want one too. Everybody wants one.

1:21:35

Automatic cat cleaners forever long. Huge.

1:21:37

Huge. Have

1:21:40

you ever heard

1:21:42

of Zulily? No.

1:21:44

What is that? I feel like I know that name.

1:21:46

I think they were B.E. Commerce for a business. They

1:21:49

are going on a business. Thirteen years. They

1:21:52

started in Seattle. They're shutting down their

1:21:54

operations because they say they couldn't compete

1:21:56

with Amazon. If

1:24:00

you settle, not sure what it costs you. Well

1:24:03

Google and the plaintiffs have agreed to terms that will

1:24:05

result in litigation being dismissed. The agreement will

1:24:07

be presented to the court by end of January, the court

1:24:09

giving final approval. This Ars Technica article

1:24:11

does not have a dollar amount, but I think I

1:24:14

did see some with like, was

1:24:16

it $20 billion? I

1:24:18

thought it was hundreds of millions. Hundreds of millions? Let

1:24:20

me see if I can find it. I thought so.

1:24:22

I don't know. I can be wrong. I

1:24:24

can be wrong. I can be hallucinating as they say. Five

1:24:26

billion dollars. Five billion, geez. According

1:24:29

to NPR. I'm not a evil voice. Five

1:24:31

billion dollars. Well, okay, wait a

1:24:33

minute. They agreed to settle a

1:24:35

$5 billion lawsuit. Oh,

1:24:37

yeah. But for how much terms

1:24:40

weren't disclosed, the suit originally sought $5

1:24:42

billion on behalf of users. Yeah, yeah,

1:24:44

yeah. Okay, all right. Yeah. I've

1:24:47

been in that seat before. We

1:24:49

don't know, I guess. I was going

1:24:51

to say, I mean it's rare for settlement amounts to

1:24:53

be made public in these sort of cases. That's

1:24:56

right. That's true. And usually, in fact,

1:24:58

there's a clause that says you can't. Public thing,

1:25:00

isn't it? Yeah. Except

1:25:02

for this. I mean, was it a class action suit? Yes.

1:25:06

In that case, the class... Wait a minute. Was it? Actually,

1:25:09

wait a minute. I'm sorry. I don't

1:25:11

know if it was. It is a class action suit. It is. Yeah,

1:25:14

actually. So, yeah, they'll have... They'll say what

1:25:17

you... We all get the 50 cents

1:25:19

each or whatever. But that's another loss for Google

1:25:21

in the courts. It's not been a good few months

1:25:23

for Google. So, when you used

1:25:25

incognito or you used it still, what

1:25:29

does it protect? Do we know? It's

1:25:31

basically the spouse mode. It

1:25:35

protects somebody with access. It doesn't add

1:25:37

the sites you visit to your history.

1:25:40

It prevents somebody in your home from going to

1:25:43

your browser and looking at what you looked at.

1:25:46

It does not, in fact, in any other

1:25:48

way hide what you do. Ah.

1:25:52

So, I think there was some merit in

1:25:54

this. I mean, by the way, that's true

1:25:56

of all private browsing modes. I

1:25:59

think. Let's

1:26:01

see. Here's

1:26:04

what Firefox says. This is in their, they

1:26:06

don't call it incognito, they call it private

1:26:09

browsing. I'm in a private browsing

1:26:11

window. It says Firefox clears your

1:26:13

search and browsing history when you close

1:26:15

all private windows, but this doesn't make

1:26:17

you anonymous. It

1:26:19

only basically clears your history is what it

1:26:21

really does. So Google says

1:26:24

it won't save your browsing

1:26:26

history, cookies and site data, information entered

1:26:28

in forms. Your activity might

1:26:30

still be visible to websites you visit,

1:26:32

your employer or school, your internet service

1:26:35

provider. Certainly. That's always been

1:26:37

there though, I think. I know Google's always had

1:26:39

this disclaimer just as Firefox had, but the name, it's

1:26:42

like Tesla's

1:26:44

autopilot. The name doesn't

1:26:47

fly. Incognito

1:26:50

and then they have a little spy with sunglasses

1:26:52

on and a hat. It implies

1:26:54

that you're traveling around invisibly. We've

1:26:56

always, I mean I've told people

1:26:58

this for years, but

1:27:01

I think it's a very common misconception just

1:27:03

from the name. And

1:27:07

Google, this will be one of those where they

1:27:09

admit no wrongdoing, but here have some money. Did

1:27:13

you? And that money will ultimately end up being

1:27:15

$7 per person. Oh,

1:27:17

if that is, oh I think the lawyer is the

1:27:19

only people who get rich on this. I

1:27:24

found this out from our local

1:27:26

NBC affiliate in

1:27:28

California. That's when

1:27:30

you were watching the local news. I was watching

1:27:32

the local news during the Rockin'

1:27:35

New Year's Eve. Yeah, let me just get my hair here

1:27:37

because I want to make sure I look good. Leo

1:27:40

is combing his hair while staring intently in

1:27:42

the camera. It's for

1:27:45

the TikTok. California

1:27:50

Law enforcement officials, authorities are told

1:27:52

that if they pull over an

1:27:54

autonomous vehicle without a driver, they

1:27:56

may not write a ticket. You

1:28:02

who had the ticket goes to school.

1:28:05

That's part of the problem, right? When

1:28:07

driverless cars break the rules of the

1:28:09

law road, there's not much the law

1:28:11

enforcement's do about it. Tickets can only

1:28:13

be written in California if there's an

1:28:15

actual driver in the car. That's not

1:28:17

true in all states. By the way,

1:28:20

In some states, you can take the

1:28:22

manufacturer, the operator of the driverless vehicle

1:28:25

with you. You skipped over stuff your.

1:28:27

How does a cop pullover driverless car?

1:28:30

Yeah, we sort of. the car. Noted.

1:28:33

Pullover. Well.

1:28:35

They I'm sure it's. Time

1:28:39

for looks or a sauce, flashing

1:28:41

lights and sirens and over to

1:28:44

get out of the way. you

1:28:46

write with that kind of know

1:28:48

that have been pulled over. Just

1:28:52

like you would. It

1:28:54

sees the flashing lights and it's really no

1:28:56

reason why I then what it is not

1:28:59

be as is that the zebra gotta pull

1:29:01

over. A.

1:29:03

Dry. So in Texas. Ah

1:29:06

credits Texas Transportation Code The owner of

1:29:08

a driverless cars considered the operator whether

1:29:10

they're in the car or not. Read.

1:29:13

In Arizona same thing, the owner of

1:29:16

the vehicle may be issued a traffic

1:29:18

citation. Or other penalty for

1:29:20

vehicle fails to comply with traffic

1:29:22

for motor vehicle was bought in

1:29:24

California. Au Prince is just seems

1:29:27

like a loophole. Perhaps the does.

1:29:30

I never really even thought of, I just

1:29:32

assumed that the car would pullover for is

1:29:34

it turned on the flashers and and. Support

1:29:37

a whole lot of rollover. Everyone should

1:29:39

pull over with a splash of beyond

1:29:42

you to lead to get Fargo bought

1:29:44

or yes or. Well.

1:29:46

But. that's

1:29:49

of as anybody show a hands

1:29:51

been in a way imho and

1:29:54

been pulled over. I

1:29:56

mean the thing is doesn't happen that often to these guys

1:29:58

are pretty good. Them in of those ones. Away

1:30:00

my can do is stop office

1:30:02

very cautious know my grandma. A

1:30:05

way I you know, Cruelly

1:30:07

personally practically out of business because

1:30:10

of that trouble with this San

1:30:12

Francisco. Dmv. And so forth.

1:30:14

so. It's really it's way

1:30:16

more at this point, but between him and who should

1:30:18

be liable? For.

1:30:22

Like you know a chat about the easy

1:30:24

as a software or you have such as

1:30:26

a as I don't know? Well okay so

1:30:28

there's somebody driving. Or something.

1:30:31

Whatever. Is driving his life, their service

1:30:33

and he I who who gets the sickest.

1:30:36

Ah, The company the com line owns the

1:30:39

car to heal on or whoever owns the guts.

1:30:41

I mean sometimes these cars are driven by somebody

1:30:43

back of the Hamas has more often apparently than

1:30:45

than we know. Ah,

1:30:48

But. So that's a driver discuss, not

1:30:50

some in the driver's seat. And yes, Way

1:30:52

Mo is if is nobody else. Way more

1:30:55

because it's their software that violated the law.

1:30:58

You gonna be able to punish him? I'm

1:31:02

sure that's the question the California legislature as

1:31:04

yeah, this is probably why there's no real

1:31:06

lies about my the and a lot worse

1:31:08

were no. I

1:31:11

think it's. Funny. Though

1:31:13

it's got it's gotta be the company's

1:31:15

own car. It's gotta be. But they're

1:31:17

also the ones who put the money

1:31:19

into the politicians. Ah, you think it's

1:31:21

a you think it's a profit motive

1:31:23

for me as isn't everything in America

1:31:25

for him. As a person who's who's

1:31:27

yes, saw the young people are just

1:31:29

so cynical I was still yeah yeah

1:31:31

I mean my sewing libraries seats is

1:31:33

this is. Under

1:31:36

a young person. Take

1:31:38

it to lot to say youngsters imitate

1:31:40

the sale Zero better for all young

1:31:42

person I used to be avoided or

1:31:45

kids to do Believe it I know

1:31:47

I call John as soon Everybody's most

1:31:49

eligible man and San Francisco I well

1:31:51

as one of the hot and on

1:31:53

the card. says. it right here

1:31:56

has formerly one in san francisco's one

1:31:58

hundred most eligible bachelors does it right

1:32:00

there in print to find a

1:32:02

read formerly lived in tomato fields

1:32:06

Oh half a detected tomato field

1:32:08

I don't get no no no we're not

1:32:11

gonna go there this

1:32:13

goes back to a story I was

1:32:15

from years ago talking about can of

1:32:17

tomato soup right okay

1:32:20

I lived in cinnamon city New Jersey

1:32:22

she asked Leo fair enough you can

1:32:25

we know that it was built on

1:32:28

the old Campbell's tomato

1:32:30

fields that's all there

1:32:32

is the story do you grow up

1:32:34

smelling the sickly sweet smell of tomatoes

1:32:36

well it was all it

1:32:38

was a great it was actually a great story about I didn't

1:32:41

realize this but Campbell's made so much tomato

1:32:43

soup they actually they actually had to like

1:32:46

buy up New Jersey and grow tomatoes there's

1:32:48

well the only place they could really do

1:32:50

it and so

1:32:52

they owned a huge farms in New Jersey

1:32:56

cinnamon and Moore's town

1:32:58

yeah so

1:33:01

the miss you the spelled CI double an AMI and

1:33:03

so when it's kind of

1:33:05

article you're reading you know metal sloths or

1:33:09

actually was in there was a

1:33:11

close it was in the Smithsonian magazine

1:33:13

how Campbell's soup turned New Jersey into

1:33:15

a tomato growing state and

1:33:18

they actually made a special kind

1:33:20

of tomato just from modern farmer

1:33:22

from modern farmer a classic magazine

1:33:24

farmer oh seriously

1:33:26

it's wonderful it's reprinted by the Smithsonian I

1:33:28

guess and there's a picture that's what they

1:33:30

give it away that's Harry Hall that's Jeff

1:33:33

check yeah that's Jeff and in

1:33:35

his younger days looking at tomato uncle

1:33:37

Harry this is

1:33:39

in cinnamon and so I

1:33:41

think it is yeah there it

1:33:43

is cinnamon cinnamon and what

1:33:46

was the name of the tomato that

1:33:50

they created there was a special tomato

1:33:52

that they made that was what

1:33:55

they made for Campbell's tomato soup it's really a

1:33:57

great actually a great story

1:33:59

I think it's That's a fascinating story. The

1:34:01

Rutgers Tomato. Probably

1:34:04

done with an extension of Rutgers University. The

1:34:08

JTD Tomato named after John Thompson

1:34:11

Dorrance, later

1:34:13

president of the company released in

1:34:15

1918, medium-sized red tomato

1:34:18

averaging in the 8 to

1:34:20

12 ounce range, uniform

1:34:22

in shape, tasty, and most importantly,

1:34:24

does not crack. The

1:34:28

JTD. Tomatoes

1:34:30

crack? Oh yeah, you've seen that,

1:34:33

haven't you? Where they kind of grow

1:34:35

the skin. Oh,

1:34:37

like an heirloom tomato. Yeah, yeah, they crack. Where

1:34:40

they look like they're weirdly sewn together. Yeah.

1:34:43

Anyway, there you go. The history of tomatoes in

1:34:45

New Jersey. I know, don't you? Now,

1:34:51

what's the accompaniment, the natural

1:34:53

accompaniment for a bowl of Campbell's tomato is?

1:34:56

Grilled cheese. Grilled cheese, yeah. Okay.

1:35:00

Gotta be. We have that most every Friday night. I

1:35:04

have an $1800 appliance that

1:35:06

makes excellent tomato soup. Say

1:35:10

more. You

1:35:12

need your tomatoes to be uncracked? No, no, I

1:35:14

actually can make it with canned

1:35:17

tomatoes, but I prefer to use

1:35:20

the nice Italian Marzano tomatoes.

1:35:23

It's the Thermomix. Do you know about the Thermomix? Stacy

1:35:26

introduced me to this, I think, and

1:35:29

forced me to buy one. I

1:35:33

will not be introducing you to any $1200

1:35:35

tomato specific apartment. I

1:35:38

also have several June ovens, thanks, Dar.

1:35:41

Stacy's by the way, it used to be a

1:35:43

game on the other show, Paris was to get

1:35:45

Leordo Live, but now the things aren't going so

1:35:47

well in podcasts when we stopped doing it. But

1:35:49

selling off all of these things that

1:35:51

he bought. Let's

1:35:54

see. As you were starting to say, Stacy, you

1:35:56

were interrupting. Stacy's going to be on

1:35:58

Twitter. When's that coming up? Benito and be

1:36:00

on Twitter on the 28th 28th of this

1:36:03

month cool and Let's

1:36:05

not forget Stacy's book club which continues

1:36:08

if you are a club twit member You

1:36:10

know that we've been doing this for some months

1:36:14

And even though Stacy has departed

1:36:16

this show. She says I still want to

1:36:18

do the book club So Stacy's book club

1:36:20

is coming up When

1:36:23

is that February 8th? I'm reading the book right

1:36:25

now. It's It's

1:36:27

called the water knife by palog bachigalupi

1:36:29

bachigalupi, and it's really good so far.

1:36:32

I'm really enjoying it so February Huh

1:36:38

What is it about I will be hosting that that's why

1:36:40

I'm reading the book it's about I'm

1:36:43

sorry cross talk it's about

1:36:45

well so far it seems to

1:36:47

be about Las Vegas in the future

1:36:49

Which has become one of those

1:36:51

you know like desert? city

1:36:54

things You know kind

1:36:56

of monoliths and then our college that's

1:36:58

the word days and so

1:37:00

it's in Las Vegas because you can't

1:37:03

live in Las Vegas because it's too hot so that's

1:37:05

become an ecology and it's about the

1:37:07

woman who really kind of is

1:37:09

the runs it just kind of Anyway,

1:37:12

it's very interesting. It's nice. Yeah, and

1:37:15

then I should mention we are gonna do an

1:37:17

insight twit for club members tomorrow One

1:37:20

o'clock Lee Lee Lee said I will do a state

1:37:22

of the nation or a state of the

1:37:24

podcast I guess One

1:37:27

p.m.. Pacific tomorrow For

1:37:30

those of you who are in the club if you're not

1:37:32

in the club And you want to know more about this

1:37:34

kind of stuff and the shows we do in the club

1:37:36

and we're gonna do Samsung Galaxy unpacked in January 17th stuff

1:37:38

like that Join the

1:37:41

club Seven bucks a month you

1:37:43

get ad-free versions of all the shows You

1:37:45

get access to the club twit discord which is

1:37:47

a wonderful community by the way ants in there

1:37:49

right now, which is great. Hi ant we

1:37:53

also He

1:37:55

says he wants an immersion blender

1:37:57

mayo tutorial, okay, ant deal.

1:38:01

I'll show you how I'm never buying

1:38:03

mail again. It's amazing. You

1:38:05

also get special programming. We don't

1:38:07

put out anywhere else like Stacy's Book Club.

1:38:10

And most importantly, that $7 a month really

1:38:12

supports what we do here. We

1:38:16

just cannot survive on ad money alone, I'm

1:38:18

sorry to say. So we

1:38:20

need your help. I'm very pleased to say

1:38:22

people have responded to this. And

1:38:25

it's great. We've got 10,000 members now. We need 35,000. So

1:38:27

which is only 5%.

1:38:31

I think I at

1:38:33

least would like to see that soon. Sooner

1:38:36

the better. I think what is the goal by

1:38:38

the end of the year 35,000? I think so.

1:38:41

twit.tv slash club. Please

1:38:44

give us and and

1:38:47

get gifts. I belong to the

1:38:49

club. I bought a subscription for Son

1:38:51

Jake. Thank you. Your whole family there.

1:38:53

Yeah. Lisa did a blog post on

1:38:55

the 12 ways to support. She's

1:38:58

she's one. In fact, you know

1:39:00

what, it was an example of how to do

1:39:02

tiktok. Because she split it up into a

1:39:04

whole bunch of little bits and then shrug it

1:39:06

out. Yeah. And she did the same thing on the Twitter,

1:39:09

which was great. And

1:39:11

it's on the twit blog at twit.tv. And some

1:39:13

of them don't cost any money. Anything you can

1:39:16

do to help will be much appreciated. Joining

1:39:19

the club is fantastic. I really love the club.

1:39:21

I during the Christmas break, I spent a lot

1:39:23

of time in the in

1:39:25

the discord, getting help from

1:39:28

the advent of code experts in there. That was a

1:39:30

lot of fun.

1:39:32

Moving right along. Should we do our AI? We've

1:39:34

done much of it, but we can do the

1:39:36

rest of it. Our AI segment. Let's this.

1:39:40

Oh, we don't have a there's no AI. Hey,

1:39:44

hey, can't you

1:39:46

see it's AI?

1:39:48

That's the Brooklyn version.

1:39:50

Hey, you should do

1:39:52

that for

1:39:57

the Brooklyn in New Jersey. Hey, yeah. From

1:40:01

Politico, a new kind of AI

1:40:03

copy can fully replicate famous people.

1:40:06

The law is powerless. You

1:40:11

can take photographs of people and steal their souls.

1:40:13

There should not be a law. Actually, the story

1:40:15

is, when you read it, it's

1:40:17

like, well, no, actually Martin Seligman loves this

1:40:19

idea. So Martin Seligman, who

1:40:21

is 81, is a psychologist, very

1:40:24

well-known apparently. He was

1:40:27

pondering, according to Mohar Chatterjee,

1:40:29

who wrote this story, he

1:40:31

was pondering his legacy at a dinner party

1:40:33

in San Francisco one late

1:40:35

February evening. He

1:40:39

got an email from a graduate

1:40:42

student in China, Yukun Zhao. He

1:40:46

had created, Zhao had created without

1:40:48

Seligman's knowledge, a virtual Seligman.

1:40:50

Over two months by feeding every word

1:40:52

Seligman had ever written into

1:40:55

cutting-edge AI software, Zhao

1:40:59

had built an eerily accurate version of

1:41:01

Seligman himself. A talking chatbot is exactly

1:41:04

what we were just talking about, right?

1:41:07

Whose answers drew deeply from Seligman's

1:41:09

ideas, whose prose sounded like a

1:41:11

folksier version of Seligman's own speech,

1:41:13

and whose wisdom anyone could access.

1:41:16

Impressed, Seligman

1:41:18

went, wow! Circulated the chatbot

1:41:20

to his closest friends and family to say,

1:41:23

you know, does this sound like

1:41:25

me? And his wife said,

1:41:27

she was blown away by it. The

1:41:29

bot is cheerfully named Ask Martin.

1:41:33

Cheerful? That's cute. It

1:41:36

was built by researchers in Beijing

1:41:39

and Wuhan without Seligman's permission or

1:41:41

even awareness. Seligman

1:41:44

doesn't mind, he's 81. In fact, this answers the

1:41:46

question, what's going to happen after I go? I

1:41:50

think this is fine. There

1:41:57

are others, Belgian celebrity psychotherapist

1:41:59

Esther Mr. Perrell,

1:42:01

who were not so

1:42:03

happy, in Southern California, a tech

1:42:06

entrepreneur created a chatbot of her

1:42:08

scraping her podcasts off the internet.

1:42:10

No one would ever scrape podcasts

1:42:12

off the internet. He

1:42:15

actually- Belgium, Dr. Phil. I

1:42:18

guess so. What's interesting is the

1:42:20

guy who did this, Alex Fermansky, did it

1:42:22

because he had a recent

1:42:24

heartbreak and he built it to counsel

1:42:26

himself. Well, that's

1:42:28

kind of sad. I want to cry

1:42:30

on it. I know. Here's his Medium

1:42:33

article. Instead of simply

1:42:35

speaking with a therapist, I created an

1:42:37

AI one. Actually, it's sub-set.

1:42:39

It's cheaper. Yeah, probably.

1:42:45

So, we've come a long way from Eliza. I

1:42:48

don't know if Perrell

1:42:50

was unhappy about this. Like

1:42:54

Seligman, the article goes on to say she

1:42:56

was more astonished than angry. She called it artificial

1:42:59

intimacy. I

1:43:03

think this is a good thing. Now

1:43:05

Congress may not. In fact, Congress has good

1:43:09

old Amy Klobuchar has

1:43:12

a bill, she's one of the co-sponsors,

1:43:14

called No Fakes, the No

1:43:16

Fakes Act, which is- I'm

1:43:20

sure it stands. It's of course an initialism. Let me

1:43:22

look it up. I want

1:43:25

to be at the bar when the staffs come up with

1:43:27

these names. Yeah, I would like

1:43:29

that job. I just like the job of making the

1:43:31

little acronyms. They're

1:43:34

really retronyms because they

1:43:36

start with what they want it to say,

1:43:39

right? Of course. Yeah. Yeah.

1:43:42

It would penalize

1:43:44

AI for generating someone's likeness without

1:43:46

their consent. Chris

1:43:51

Coons, Marsha Blackburn, Amy

1:43:53

Klobuchar, and Tom Tillis. Amy,

1:43:56

come on. Knock it off. The

1:43:58

No Fakes Act. Let's

1:44:01

download the text of the No Fakes Act

1:44:04

and see what that stands for. Now you've got me. I really

1:44:07

want to know. Title,

1:44:12

to protect the image, voice, and

1:44:14

visual likeness of individuals and

1:44:16

for other purposes. Oh, get ready. Here

1:44:18

we go. What does No Fakes

1:44:20

stand for? Nurture originals,

1:44:23

foster art, and keep

1:44:26

entertainment safe, act. Oh,

1:44:30

man. Well, that's the No... No

1:44:34

Fakes. That is No Fakes. Yeah,

1:44:37

it spells it. I

1:44:39

mean, yeah, I guess you should

1:44:41

have to give permission. Honestly, if somebody wanted to

1:44:43

create a Leo Laporte, I don't think there's a

1:44:46

problem with that. No, it depends

1:44:48

on how they use it. What if the Leo bot

1:44:50

became more popular than you, Leo? Would you have a

1:44:52

problem with that? Fine. But I'm at the

1:44:54

end of my career. And someone was making money on it?

1:44:56

It might be different for you, right? I'm at the end

1:44:58

of my career, and I would be fine with that. But

1:45:00

you might not. The issue isn't making it. The issue is

1:45:02

how you use it. If you use your image in an

1:45:04

ad and you don't give permission, that's already illegal. That's already

1:45:06

illegal, yeah. If you use it for a commercial purpose rather

1:45:08

than an editorial purpose. By the way, it's not illegal if

1:45:10

I'm dead, is it? And

1:45:13

it's not illegal if it's an editorial purpose. I

1:45:18

don't know. I have a mixed feeling. It's

1:45:20

illegal if you're dead for

1:45:23

a certain period of time. I mean, isn't that how the

1:45:25

estates of famous people work? What, until it's cold? I'd

1:45:28

say you're colder in your grave. The

1:45:31

No-Fakes Act has received support from multiple organizations

1:45:33

across the arts and entertainment industries. The

1:45:36

RIAA, of course, describes

1:45:39

the use of unauthorized AI performances

1:45:41

as theft. The Actors

1:45:43

and Writers Union, SAG-AFTRA, and

1:45:45

WGA. Of

1:45:48

course, one of the reasons they struck

1:45:50

was they were concerned about AI. I don't know. I

1:45:56

think I have the same position, which

1:45:58

is... It's probably already illegal. to

1:46:00

do certain things. And we got to be

1:46:02

careful about too much regulation on this stuff

1:46:04

because we want the innovation to

1:46:06

happen. I think there's a benefit. I don't

1:46:09

know but I mean what innovation is happening

1:46:11

by making an AI chatbot off

1:46:14

based on one person? Well

1:46:16

you're probably right. Publicly scraping

1:46:18

all of their appearances. But

1:46:21

maybe the real benefit is that we learn

1:46:23

something about it, not necessarily that that thing

1:46:25

that they created is of that much value

1:46:27

but... I mean yeah, if you want to

1:46:29

do something in private to

1:46:32

learn something I don't think anyone's stopping you

1:46:35

from doing that. I think that probably

1:46:37

what... obviously I don't know the text of

1:46:39

this bill and whatnot but I assume it

1:46:41

has something to do with commercial use of

1:46:44

these. I think it just prohibits it

1:46:46

outright whether you make money on

1:46:48

it or not. Good article. This

1:46:50

is the one I would point people to.

1:46:53

Jan Lekun who is one of the founding

1:46:55

fathers of LLMs and is very well

1:46:57

known. He's at Facebook. He's

1:47:00

met his chief AI scientist.

1:47:03

A good interview with him in Wired.

1:47:08

He scoffs Wired

1:47:10

writes at his peers dystopian

1:47:12

scenarios of supercharged misinformation

1:47:15

and even eventually human extinction.

1:47:17

He's not a doomer.

1:47:21

When his former collaborators... He's a voice of

1:47:23

sanity in all of these. I think he

1:47:25

is. When his former collaborators Jeffrey Hinton and

1:47:27

Yoshua Bengio put their names at the top

1:47:29

of a statement calling AI a societal scale

1:47:31

risk, Lekun refused to sign

1:47:34

it. He said instead he signed an open

1:47:36

letter to the US President Joe

1:47:38

Biden urging an embrace of

1:47:40

open source AI declaring

1:47:43

it should not be under the control of

1:47:45

a select few corporate entities. Exactly what I've

1:47:47

been saying but I... Here's

1:47:49

a question. Yeah. Here's a question. When

1:47:53

Lambda kind of leaked, the

1:47:56

presumption of many was that all they didn't... Facebook

1:47:59

wouldn't do that. want to do that and

1:48:01

this open source stuff is accidental. I don't think

1:48:03

so. I think Lekun is so religious

1:48:06

on the point. Did

1:48:08

Lambda leak or was it intentional? I

1:48:10

think it initially leaked and then Facebook

1:48:14

released to the open source community a

1:48:17

public version of it called

1:48:19

Llama 2. So I think

1:48:21

yeah Lambda's Googles.

1:48:23

So they're all plays

1:48:25

on the album. That's the one that's alive. Right. Llama

1:48:29

2 is widely used in

1:48:31

open source. I have some open source tools that

1:48:34

use it as a model. Who

1:48:40

wrote this? I only get his name. This

1:48:42

is Stephen Levy, my good friend Stephen. Hi

1:48:44

Stephen. He says,

1:48:46

when I sat down with Lekun in a conference