0:00

Coming up on tech news weekly,

0:02

Jason Al and I have a great show

0:04

planned for you. First, we stop

0:06

in with iFix. It's Sharam Maktari

0:08

to talk about the brand new

0:10

non pro iPhones. These

0:13

iPhones have quite the teardown

0:15

surprise, then Becca Ricks

0:17

from the Mozilla Foundation, stop spy to talk

0:19

about how those YouTube buttons that

0:21

are supposed to tell YouTube that you don't

0:24

like certain content and don't wanna see it

0:26

anymore, they may not work

0:28

as we expect them to and may not

0:30

be all that effective at all. Then

0:32

we talk about a new headset

0:34

to rival the Mediquest two

0:37

called the Pico headset and

0:39

round things out with an AI

0:42

chat bot. that uses humans

0:45

to provide better responses. Stay

0:47

tuned.

0:50

Podcasts you love. from

0:52

people you trust. This

0:55

is quiet. This

0:59

is Technus Weekly episode two hundred fifty

1:02

three recorded Thursday, September twenty

1:04

second twenty twenty two. This

1:06

episode of Technus Weekly is brought to you by

1:09

click up the productivity platform

1:11

that'll save you one day a week on work

1:13

guaranteed. Use code

1:15

t SW to get fifteen

1:17

percent off QuickBooks massive unlimited

1:19

plan for a year, meaning you

1:21

can start reclaiming your time for under

1:23

five bucks a month. Sign up today

1:25

at click up dot com, but hurry,

1:27

this offer ends up. And

1:29

by infra scale, infra scale delivers

1:32

industry leading data protection through backup

1:34

and disaster recovery. Visit

1:36

empress scale dot com slash tweet to

1:38

sign up for a free demo and see how Impress

1:40

scale protects your business today. and

1:43

by ZipRecruiter, Hiring.

1:46

Look, it's challenging. But there's one place

1:48

where finding great candidates is easy.

1:50

ZipRecruiter. ZipRecruiter finds

1:53

and matches the right candidates for you, and

1:55

our listeners can try it free today

1:58

at zip recruiter dot com slash

2:00

TNW ZipRecruiter,

2:03

the smartest way to hire. Hello,

2:07

and welcome to Tech News

2:09

Weekly, the show where every week we talk to

2:11

and about the people making and breaking

2:13

the tech news. I am one of your hosts,

2:15

Micah Sargent. And I'm the other

2:18

guy, Jason Howell, and I gotta tell

2:20

you sometimes it's it's

2:22

a delight for me on this show Micah,

2:24

you focus so much on Apple hardware.

2:26

Right? That's obviously got iOS

2:28

today. you're you're a iOS

2:31

and iPhone user. I

2:34

have all that Android. So

2:36

I'm always you know, I feel like I up

2:38

in the Android bucket a

2:40

lot of the time. So I actually like when we kinda

2:42

change roles -- Mix it up. -- to change it up.

2:44

Yeah. Gotta mix it up. So that's what

2:46

we're doing today. We're gonna start this show

2:48

with exactly that. A brand new batch

2:51

of iPhone hardware is out

2:53

of course comes with the need to know

2:55

exactly what might be hiding because

2:57

you know how it rolls. Right? We have new

2:59

hardware, especially from a company like Apple,

3:01

very soon after. someone tears

3:04

it open and tells you what's hiding

3:06

inside for a number of reasons, repairability,

3:08

just curiosity, whatever it may be.

3:10

And thankfully, got some pretty awesome

3:12

people who seem to do this like clockwork of

3:14

the folks that I fix it, and joining us

3:16

to share what he and the team

3:19

discovered lurking inside the new iPhone

3:21

four team devices as I fix it.

3:23

Charam Macquarie. Welcome to the show.

3:26

Hey, guys. Thank you for having me. Good to

3:28

be here. Yeah. It's good to get you

3:30

here. We got a lot to talk about because there are

3:32

some interesting discoveries here when it

3:34

comes to the iPhone four

3:36

team lineup. So thanks

3:39

for hopping on. Let's before we

3:41

get into the pro and the max, let's kinda

3:43

go through the kind of the low

3:45

I don't know what you call them. The lower end or

3:47

the the more vanilla versions,

3:50

the iPhone fourteen's, scored

3:52

pretty well when it comes to repairability

3:55

at least by comparison, which we'll get to

3:57

that in a minute in the pro and max. But

3:59

it seems due to a redesign

4:02

in how these devices put together.

4:04

Tell us a little bit about those changes. So

4:08

spot on base model, vanilla iPhone

4:10

fourteen, whatever you wanna call it. The

4:13

the base model has been redesigned from

4:15

the inside out. It's really quite interesting

4:17

and impressive what they've done. What

4:20

the engineers essentially did was they took

4:22

the back plate that sits right to the back of the

4:24

phone and they've processed a few millimeters

4:26

into the case itself. Now,

4:29

there are implications for this,

4:31

but one of the key takeaways is that the phone

4:33

is now more repairable because the backplate

4:35

is removable. as is the

4:37

front screen. It's a huge win

4:40

for repair advocates. Does

4:42

that when you say it takes it in

4:44

a little bit, does that mean is that to say

4:46

that there's more space on the inside

4:49

as a result of this? Or everything is

4:51

pretty tightly compact regardless it's

4:53

just easier to access? Oh,

4:55

it's it's tightly compact. There should

4:57

be less space in there. And

4:59

that's what's interesting. Because

5:02

because they've taken the backplate. That's where the

5:04

rear glass is adhered to. Because

5:06

they've taken that backplate and brought this in

5:08

some. There should be less space because you've also got

5:10

a backplate port the rear glass now.

5:12

Mhmm. But the form

5:14

factor is the same. You've got on the outside,

5:16

this looks like any other iPhone fourteen we've

5:18

ever had. and that's what's truly impressive.

5:20

They've done an excellent job

5:23

with with this design. Have

5:25

you seen a butterfly approach like

5:27

this? because, I mean, I've I've seen a million, you know, tear

5:29

downs and everything. And usually, it's

5:31

held together by hot glue accessible from

5:34

one side, obviously, makes you

5:36

know, if you've got a simple well, I say

5:38

simple with air quotes, but if you've

5:40

got a simple, you know, replacement of, like, a battery

5:42

or display, you have to go through all these layers

5:44

in order to get there. Is this a first

5:46

as far as this butterfly approach is

5:48

concerned based on what you've discovered that I

5:50

fix it? Personally, I've not

5:52

seen anything like this before. It

5:54

doesn't mean that there aren't any out there, but I haven't

5:56

seen any. It is definitely

5:58

among the major brands. It's not something we seen

6:01

I've got I'm an android man myself. I've got

6:03

a note nine here. I've done the screen replacement on

6:05

that. And on the Android, they are

6:09

you you would remove the backup the

6:11

phone to get into the internals. But

6:13

the problem with the Android phones are as

6:15

as with the iPhone four, the last iPhone

6:17

to open from the back was the iPhone four.

6:20

The screen does not separate

6:22

from the frame because only one side of this

6:24

phone separates. It it is an it is a huge

6:26

engineering challenge to make both sides open

6:28

up. What happens there,

6:30

as with my iPhone, my

6:32

Note nine, you

6:34

waste the mid frame when you replace the screen

6:36

because the screen replacement getting the screen

6:38

off and getting a new screen onto the mid frame

6:40

is extremely difficult, especially with the

6:42

curved edges. So

6:44

you throw away the mid frame, you buy the whole

6:46

screen and the mid frame as a unit and

6:48

then you put the internals inside the new mid

6:50

frame and then you put the back on that's how you do a

6:52

screen replacement on an Android and on

6:54

most other phones. So this

6:57

this is huge that they've done this.

7:00

Nice. Nice. Alright. So then

7:02

before we kind of turn

7:04

our attention on how the

7:06

pro and the max are actually incredibly different

7:08

from what you're talking about. Any

7:11

other kind of discoveries pop up that

7:13

you feel are pretty notable about

7:15

what was hiding inside the the

7:17

the baseline iPhone fourteen models?

7:20

In the baseline model, I I think

7:22

it was just this I say

7:24

just this. It's such a

7:26

such a challenge to to do

7:28

what they've done here. I I

7:30

think it's just become incredibly

7:33

modular now. There is very little

7:35

to waste when you perform a repair here.

7:38

And that is that is the

7:41

key aspect of the iPhone fourteen, the

7:43

base model. the key

7:45

improvement and it's the key improvement we

7:47

expect to see. I would love to see

7:49

in every phone

7:51

in their lineup next And there's there's really

7:53

no reason why they shouldn't implement this. It's been

7:55

so well received. Yeah. I'd be

7:57

really surprised. Well, and what occurs to me also

7:59

right now is that Apple very recently,

8:01

you know, launched their own kind

8:03

of self repair kits, which

8:06

as Micah would attest, you know,

8:08

is Apple promising that, hey, you know, you

8:10

can get this you can do it all yourself. That doesn't

8:12

necessarily mean any of it's easy

8:14

or convenient necessarily. And

8:17

so I don't know. It just kind of

8:19

occurs to me right now that some of

8:21

these changes kind of play

8:23

into potentially making

8:25

that service, making what you guys do more

8:28

viable for more people as a result.

8:30

Right? Potentially,

8:32

yeah, potentially, the as

8:35

you as you said, rightly

8:37

yourself using Apple service

8:39

tools is not I wouldn't

8:41

say it's not easy, but it's not practical.

8:43

The the just the stereo. size of the

8:45

equipment is not it's not a

8:47

reasonable way to repair

8:49

a phone. But it's good to

8:51

have the option. I mean, the more

8:53

choices you have in your your

8:55

means of preparing your devices is only a

8:57

good thing. And you can choose our stuff. You can choose their

8:59

stuff. Whatever works for you. Absolutely.

9:03

Totally agree. Okay. So then

9:05

let's kind of turn our attention on

9:07

the the the bigger devices, let's

9:09

say, the iPhone fourteen Pro

9:11

the pro max. And

9:13

when I think of pro

9:15

or max or plus devices or

9:17

whatever, I think of everything

9:19

you got in the vanilla and

9:22

more. But in this case

9:24

right. But in this case, it really seems like

9:26

they're not cut from the same cloth. They're

9:28

actually very different. And actually, the pro and

9:30

the pro max are a little bit more

9:32

traditional design. So

9:34

what exactly, you know, did

9:36

you did your team find here?

9:38

And were you surprised at what you found? I

9:41

I think the pro and

9:43

pro max were run

9:45

of the mill Apple architecture. It

9:47

was not surprising for us. It's

9:50

it's only surprising in that it's it's

9:53

not different. from past generations of phones.

9:55

It's what we expected to see. Had

9:57

it not been for the base model iPhone

9:59

fourteen. It's

10:01

It's also understandable why they did

10:03

this. You have to understand the

10:05

iPhone is a huge product globally.

10:07

It represents something like twenty five

10:09

percent global cell phone sales.

10:12

In the U. S. it's a

10:14

huge seller. It

10:16

represents a large investment It

10:19

affects any changes to these phones

10:21

affects Apple's bottom line, and it makes

10:24

investors nervous. So it only makes

10:26

sense to make incremental changes. and

10:28

it only makes sense to make the

10:30

the big changes in your

10:32

least least popular

10:34

device, the device that is least

10:36

likely to affect your bottom line

10:39

if things should go wrong.

10:42

And from that perspective, it

10:44

makes perfect business sense for them to

10:46

make the major change in the iPhone

10:49

fourteen base model and

10:51

perform the incremental changes in

10:53

in the other models. I I think for

10:55

the most part, people have been interested who

10:57

have have been enamored with this

10:59

dynamic island feature, which is

11:01

mostly a software feature. There is a hardware

11:03

change there. They did In order to make

11:05

the the pill

11:08

shaped camera sets up in the screen, they

11:10

had to move some of the sensors a

11:12

little lower. And there's a

11:14

we've seen the patterns for this. There's some sort

11:16

of projector mechanism that allows

11:19

for the infrared, and I

11:21

believe it's the

11:24

lidar sensor that has

11:26

been moved slightly below the

11:28

rest of the assembly in that pill. to

11:31

allow for this to be compacted as much as

11:33

it has been. So that is a change. That is a hardware

11:35

change. And obviously, we've got the

11:37

satellite and SOS messaging in there as

11:39

well. Yeah. Indeed. And

11:41

there is this kind of

11:45

all in dedication to the eSIM,

11:47

which essentially, at least here in the

11:49

US, eliminates the need for the

11:51

Sim tray. I guess, I was

11:53

a little surprised if I'm that because

11:55

when I think of that, I think of Oh,

11:57

just think about all of the, you know, all the

11:59

other things Apple can do with that space or

12:01

with that space inside of the design, and

12:03

not just Apple, anyone that, you know, like this

12:05

happened in Android, as well as well

12:07

as Apple devices removing

12:09

the headphone port. And a lot of the times, they

12:11

would say, well, you know, if we remove that,

12:13

then that gives us more flexibility to do

12:16

other things, blah blah. And it was hard to tell

12:18

whether that was just, like, passive

12:20

justification to, like, you know,

12:22

get you on board with it or if that

12:24

actually was the case. here, the

12:26

Sim tray was filled with, like,

12:28

a little plastic spacer. So there was really

12:30

nothing of any importance kind of

12:32

put into that place in its

12:34

in its swap out.

12:37

It is. Absolutely.

12:39

It was surprising for us too. This

12:41

is the plastic spacer. It weighs about

12:43

a gram. And this is

12:45

the Promax Logic board. That's

12:47

where the SIM module should

12:49

be should be sitting. And it's just

12:51

an empty space. I I don't think it's something

12:54

It it depends on who

12:56

who you ask. It's it's not something we can

12:58

really say is a very bad thing.

13:00

It's not good. It's not bad. The

13:03

one thing you will see on the Internet, a lot

13:05

of people are commenting. They're saying, well, it's

13:07

only the US version. Everybody else gets

13:09

SIM card. But they're missing the

13:11

fact that the US represents

13:13

a market larger than

13:16

Russia, the CSTO countries, and

13:18

Europe combined. It is a huge, huge

13:20

market to to ignore such a large

13:22

market is unusual,

13:24

especially when you consider that they make four

13:26

different types of this board. There's one

13:28

for the Chinese market. Run

13:30

one for the Russian CSTO markets,

13:32

a global version and a US version. There's

13:34

four different versions. They could have put something

13:36

else like micro SD card

13:38

reader, for example. So so you can so you

13:40

can make better use of that eight k camera

13:43

eight k camera

13:45

function because that's gonna chew up

13:47

some space. Yeah. I just wow. I just

13:49

can't see Apple putting a micro

13:51

SD slot on the

13:53

phone. That would be that would be

13:55

neat. I mean, I would applaud that, but I

13:57

just scared the apple too, though.

13:59

That wasn't too much money.

14:01

Yeah. Yeah. Exactly. Right. Well and

14:03

what what does that do? That kind of puts

14:05

the consumer into the position when they

14:08

buy the phone to decide.

14:10

Actually, I'm gonna pony up and

14:12

get larger storage space because I don't

14:14

have any other choice. Right? Yeah.

14:16

They're they're cloud storage for, you

14:18

know, so Apple gets the services money

14:20

there as well. If if people

14:22

wanna learn more about their right

14:24

to repair and all of this information,

14:26

you told told me in an

14:29

email about about a site. Where where do

14:31

you wanna point people to so that they can find

14:33

information about all this stuff? Well, we

14:35

talk a lot about repair and how

14:37

people can get involved on our blog.

14:39

So if if you check out our website

14:41

check out our blog in particular. You can find a lot of information

14:43

there, but also repair dot org. If you want to

14:45

get involved, if you want to

14:47

influence the

14:49

the kind of changes that we're advocating

14:51

for. Repairedosh org is the place to

14:53

go. Right on. Shiran

14:56

Maktari from I Fixit. Thank you so

14:58

much for hopping on today and telling

15:00

us all about what I think a

15:02

lot of people who watch and listen to

15:04

this show think would be actually a pretty fun job

15:06

tearing apart technology and

15:08

gadgets like you do. So thanks for doing what

15:10

you do, and thanks for talking all about it on the

15:12

show today. My genuine pleasure.

15:14

Thank you for having me. We'll talk to you

15:16

soon.

15:16

Thanks

15:18

so much. So

15:20

we're we're gonna take a quick

15:23

break before we come back to

15:25

talk about a

15:28

tool version, a digital version, an online

15:30

version of that elevator close

15:32

button. Does it actually work?

15:35

I don't know. I don't know. We'll find

15:37

out. No. There's a little

15:39

tease while we take a quick

15:41

break. So I can tell you about click

15:43

up. We're bringing you this episode

15:45

of Tech News Weekly.

15:47

So just spend a second here

15:49

thinking about having one

15:51

extra day a week.

15:54

More time to cook healthy

15:56

meals, perhaps work on a

15:58

novel you've been writing or maybe

16:00

just binging some good reality TV.

16:02

For me, I would spend that one extra

16:05

day a week sleeping as much

16:07

as humanly possible. Love to

16:09

sleep. And if I had extra day a week, that's

16:11

what I'd do with it. it's all possible

16:13

now with ClickUp, the

16:15

productivity platform that will save

16:17

you one day a week on work.

16:18

and it's guaranteed. Clickup

16:21

began with a premise that productivity was

16:23

broken. All these different apps

16:25

that have all of these different tools

16:27

and features that you're all trying to

16:29

keep track of, and that starts to add up

16:31

as you're going from one ecosystem

16:33

to another, signing into

16:35

one service, signing into another service,

16:37

etcetera, etcetera. It just it takes up so

16:39

much of your time. So what

16:41

would happen if we could

16:43

improve that

16:45

productivity and make it so that

16:47

everything is where it needs to be.

16:49

Clickup is the one tool to house all your

16:51

tasks, all your projects, all your docs,

16:53

all your goals, all your spreadsheets and

16:55

more. clickup is built for teams from one,

16:57

yes, as small as one person

16:59

to a thousand or more. Very

17:02

great that you can, you know,

17:04

size it based on what you

17:06

need. It's packed with features and

17:08

customization options that no other

17:10

productivity tool has so you

17:12

can work the way you work best.

17:14

whether you're in project management, engineering, sales,

17:16

marketing, HR, click up has easy

17:18

to use solutions that create a

17:20

more efficient work environment.

17:23

So

17:23

Why

17:23

not join more than eight

17:26

hundred thousand highly

17:28

productive teams using ClickUp

17:30

today? All you have to do is use

17:32

code TNW to get fifth nineteen

17:34

percent off QuickBooks massive

17:36

unlimited plan for a year,

17:38

meaning you can start reclaiming your

17:40

time for under five bucks

17:42

a month. Sign up today at click up dot com and

17:44

use the code TNW

17:46

But hurry, because

17:47

this offer

17:48

ends soon. Thank you, click up.

17:51

for sponsoring this week's episode of TechNews Weekly.

17:54

Alright. We are back from the break

17:56

and just recently I saw

17:58

this piece fly by. from

18:00

the Mozilla Foundation about

18:03

and as I said before the break, an

18:05

online version of what I consider the

18:07

elevator door close button.

18:09

Where we go, I'm pressing it, but

18:11

I don't know if it's actually listening to what

18:13

I'm saying. And it turns out that the

18:15

same thing may apply to

18:18

YouTube's feedback buttons.

18:21

Joining us today to talk about the

18:23

research regarding the

18:25

YouTube feedback buttons and their efficacy is

18:29

BeccaRicks senior researcher at

18:31

Mozilla. Who is here with us? Thank

18:33

you for joining us Becca.

18:35

Thanks for

18:35

having me. Yeah. It's such a pleasure

18:37

to get you here. I would like

18:39

to start out by having kind

18:41

of a quick definition and explanation, if you will,

18:43

for folks who might not understand. Because

18:46

I think we make at times a lot of

18:48

jokes about, oh, the algo is failing me

18:50

or the the tick talk

18:52

algorithm or this or that. When someone was

18:54

referring to the YouTube algorithm, what

18:56

does that mean exactly?

18:58

Yeah. So when we're talking

19:00

about the YouTube algorithm, often

19:02

what we're referring to is actually YouTube's

19:05

recommendation system. So, you know, when you go

19:07

to the YouTube homepage and you see a panel

19:09

of videos recommended to you that might be

19:11

based on previous things you've

19:13

searched for, interests you've had.

19:15

That's sort of the recommendation algorithm

19:17

at work. And you can also see it when

19:19

you watch a video. There are often

19:21

recommendations on the side. also

19:22

based on stuff that you've seen before. And we

19:24

we know

19:25

a little bit about how that recommender

19:27

system works because YouTube has

19:30

some information about what signals

19:32

go into it, so we know that things

19:34

like clicks, how long you're

19:36

watching a video, how

19:38

often you're sharing it, whether you like it, are all

19:40

signals that YouTube is looking at to

19:42

personalize your recommendations.

19:44

Awesome. Thank you for that. Now what now

19:47

we've got that explained. I'd

19:49

love to kind of dig into this research because

19:51

now we can talk about, like,

19:53

in in reading about this, I was so curious,

19:56

what led to kind

19:58

of looking into this? Because, you know, by

20:00

default, you might not consider that

20:02

these buttons may not be working as

20:04

you expect or the feedback you're giving is

20:06

as expected. But when you start

20:08

asking around, then it turns out

20:10

oh, we're all having similar experiences, and

20:13

suddenly it's a conspiracy. So

20:15

tell us what inspired the Mozilla Foundation

20:17

to look into the effectiveness of

20:19

YouTube's feedback buttons?

20:20

Yeah. So we started looking into

20:23

YouTube a couple of years ago because we

20:25

were hearing from some of our supporters that

20:27

they were having really strange experiences

20:29

with the recommender system. And

20:31

we actually developed a tool

20:33

called regrets reporter, which is a browser

20:36

extension that allows people to actually donate

20:38

their data about you to us that we can

20:40

analyze so that we can actually look at how

20:42

real people are experiencing the

20:44

platform. And a previous report

20:46

that we had we had

20:48

platform, looked at what kinds

20:50

of content were being recommended to

20:52

people that they considered harmful.

20:54

So last year, we put out a report where we

20:56

were really looking at some of the videos

20:58

people quote unquote regretted. So

21:01

that was things like harmful

21:03

misinformation, scans,

21:06

spam, you know, all the clickbait that I think a lot

21:08

of us who use YouTube are are quite

21:10

familiar with. And

21:12

in response to that, you know, YouTube

21:14

has been a little

21:16

bit more transparent about

21:18

how people can take steps to try

21:20

to control their recommendations. So they

21:23

released this blog post that

21:25

basically walks you through a couple of

21:27

the controls you can use. And

21:29

we've also heard from them over the past

21:31

couple of years that they're moving

21:33

away from looking at signals like

21:36

watch time or engagement. So

21:38

basically measures of, like, are you just, like, watching

21:40

video after video after video and more

21:42

towards something called user satisfaction

21:44

or user well-being. And

21:46

so we really just wanted to see

21:48

if that was true, if they were waiting

21:50

some of those signals around

21:52

people's health and satisfaction more than

21:54

some of those other signals around engagement,

21:56

which actually align really well with their

21:59

business model. And so

22:01

what we decided to do is that

22:03

we worked with almost twenty three

22:05

thousand people who donated

22:07

their data to us through our regrets.

22:09

report a browser extension to really

22:12

look at how they were interacting with some of

22:14

these feedback tools. So those are

22:16

things like the dislike button

22:18

or the do not recommend channel

22:20

button to really look at what happened to your

22:22

recommendations after you use

22:24

those buttons. Understood.

22:25

And now we can talk

22:28

about the outcome of that

22:30

research in looking into

22:33

the data

22:35

that you got that was donated through

22:37

that regrets reporter extension. What

22:39

did the Mozilla Foundation find?

22:41

Yeah.

22:42

So in the research,

22:44

we decided we were going to do a little bit

22:46

of qualitative research as well as this

22:48

big quantitative study. And so

22:51

first, what we really looked at

22:53

was people's attitudes and

22:55

their feelings of control around

22:57

their recommendations. So

22:59

we ran a survey. We conducted

23:02

interviews where we really dug into some

23:05

of the tactics that people were

23:07

using in an attempt to try to control

23:08

their recommendations. And

23:10

so we really learned that in

23:13

general, people use kind of like a

23:15

trial and error approach to

23:17

their recommendations. So,

23:19

you know, people were using

23:21

the do not recommend this channel button, but they

23:23

were also using, like, privacy VPNs

23:26

or, you know, a number of people

23:28

said they would log out of their account

23:31

watch a video that they knew would mess up

23:33

their recommendations and then log back in. So people

23:35

kind of engaged in all different kinds

23:38

of behaviors. because they felt like

23:40

these controls didn't always work or they

23:42

didn't really understand how they

23:44

worked. So what that really showed us was

23:46

that people are

23:48

just not totally confident that using these controls

23:50

work, and we actually found that almost

23:52

forty percent of our

23:54

respondents said that they felt that their

23:56

actions had no effect at all

23:58

on what they were being

23:59

recommended, which we

24:01

thought was really interesting. And

24:03

then what was great is we could

24:05

then run this quantitative study where

24:07

we could actually look at whether those feelings

24:10

and impressions were true, like actually born

24:12

out in the data. And

24:15

so What we did find is that we

24:17

tested four different buttons. We

24:19

used the dislike button,

24:21

the not interested button, the

24:24

not recommend channel and then remove from

24:26

watch history. We tested those

24:28

four to look at how it affected

24:30

your recommendations. and across the board, we found

24:32

that they were not effective tools

24:34

for preventing unwanted

24:36

recommendations. The

24:39

most effective of all of those was do not

24:41

recommend channel, which

24:43

prevented forty three percent of unwanted

24:45

recommendations, and the lowest was the dislike

24:47

which only prevented eleven percent.

24:49

So what that indicated to us

24:51

is that these tools are somewhat effective

24:54

at preventing people from seeing stuff they don't wanna

24:56

see, but by and large, it's probably

24:58

the case that YouTube is waiting

25:01

other signals like engagement.

25:03

or the popularity of the video, maybe more heavily

25:05

than other signals like user feedback.

25:08

Mhmm. I wanna say, I

25:12

I feel I've seen I've observed that

25:14

a lot of times qualitative

25:17

qualitative, excuse me, measurements

25:19

are often given a

25:22

lower rung of importance to

25:24

quantitative measurements, particularly

25:26

in rigorous scientific studies and

25:28

in research. And when

25:30

it comes to something like

25:32

this, I'm so glad that

25:34

you all made this part of

25:36

the study that you did because

25:38

only were you then able to compare the

25:41

two and look at what people thought

25:43

versus what was actually happening, but just

25:45

knowing what people thought. if

25:48

I can sort of paraphrase one

25:50

of my favorite reality TV

25:52

shows in survivor, they talk

25:55

about how reality is

25:58

perception. And so it it regardless

26:00

of what is actually happening,

26:02

the way that something is being

26:05

perceived is what

26:07

is kind of the truth in that

26:09

moment. And so it

26:11

was interesting to look through this

26:13

study and see that folks genuinely

26:16

felt like those buttons didn't really

26:18

do anything for them. And

26:20

especially when it came to the

26:22

thumbs down button, that seemed

26:25

to be the case in many

26:27

cases. And I

26:29

wanna ask you too. It didn't

26:31

originally include this in the questions, but I wanted to

26:33

know a little bit more about the

26:35

regrets reporter because when I had

26:37

first read through this the press

26:39

release and through the study, I didn't realize

26:41

that this was an extension that is kind of a

26:43

a tool that the folks can use.

26:45

It's not just a

26:47

research gathering tool that

26:50

folks who wanted to be part of this

26:52

research campaign we're we're

26:54

using. Can you talk actually just a little

26:56

bit about the regrets reporter?

26:58

And then which browsers

27:00

support it and where folks could go to get that?

27:02

Howard Bauchner:

27:03

Yeah, for sure. So regrets reporter

27:05

was developed about two years

27:08

ago as both a research tool

27:10

and a tool that people can use to just

27:12

more quickly send a negative feedback

27:14

signal to YouTube. So it's a

27:16

way for people to sort of

27:18

use some of those tools that YouTube

27:20

offers by offering a layer on top of it,

27:22

but then also people can opt into some of

27:24

these studies and You know,

27:26

it we're in a really unique

27:28

situation that I think a lot of other

27:30

researchers aren't in because we

27:32

have access to this massive, engaged

27:35

grassroots community of people who are

27:37

really excited to be able to share their

27:39

data with us obviously

27:41

in a privacy preserving way

27:43

in order to be able to run these kinds of studies.

27:45

So you can download the browser extension

27:47

in both Firefox and Chrome

27:51

and you can find that by searching in

27:53

the Chrome and Firefox extension

27:56

stores. Yeah.

27:59

Cool.

28:00

And then one of the, you

28:02

know, of course, biggest things that I'm curious about

28:04

is anytime you do one of these studies,

28:06

there are some big kind of

28:09

highlights or trends or

28:12

outcomes that inevitably show

28:14

up, anything that surprised you,

28:16

anything that particularly stuck out, stuck

28:18

out, excuse me, to you that you

28:20

wanna share with our listeners. Yeah.

28:22

I think I was particularly

28:24

struck by some of the ways that the

28:27

qualitative and quantitative research, which were

28:29

kind of separate studies to some

28:31

degree. Actually, some of the findings

28:33

from that research really reinforced one

28:35

another, which I think to your earlier

28:37

point shows how why

28:39

doing this kind of mixed methods

28:41

research is really important.

28:43

So for instance, we learned in

28:45

our survey data that A lot of people

28:47

said, for example, when they used the

28:49

don't recommend channel button, they felt

28:51

like it was relatively useful

28:54

for prevent for blocking

28:56

recommendations from a specific channel, but

28:58

that they felt they continued to get

29:00

recommended similar videos from

29:03

other channels. And when we looked at our regrets

29:05

reporter data, we saw the exact same

29:07

dynamic in terms of

29:09

the effectiveness of that button. And I think

29:11

what that indicates to me is that people

29:13

aren't totally clear on whether, you know, when I use

29:15

this, is this like a vague

29:18

expression of preference where I'm like, I sort of don't

29:20

like this, maybe don't recommend it,

29:22

or whether instruction, like,

29:24

absolutely block this, block this, block this, block this, don't

29:26

use this. So to me, that indicates

29:28

that there's a lot that YouTube could be

29:30

doing around education, but

29:32

also just developing better tools that put people back in the

29:35

driver's

29:35

seat. So I think that

29:36

was really interesting. And

29:39

then I think the other big insight we

29:42

had was just about how

29:44

YouTube approaches,

29:47

how it infers your preferences.

29:50

So right now, the model it operates

29:52

off of is essentially a lot

29:54

of passive data collection.

29:56

So it's basically watching what you do when

29:58

it really could be asking you what you want

30:00

to watch

30:01

on the platform. And

30:03

so much of the platform is really powered by that

30:05

passive data collection. And we think, you know, there

30:07

could be opportunities for people

30:10

to more proactively state what

30:12

their preferences are. Mhmm. Yeah.

30:14

And I guess that that's kind of how

30:16

I wanted to round things out is

30:18

particularly for Google

30:20

slash YouTube. Just in

30:23

general, especially given what

30:25

you said earlier about how the

30:27

company indicated that they were shifting

30:30

the way that they looked at the

30:32

the ranking system and the

30:34

recommendation system given

30:36

that it would appear

30:38

perhaps based on the data behind the

30:40

scenes, there's not much of a change.

30:42

do we think do you think that that is a change

30:45

that's still in the works? And is

30:47

there anything that YouTube and

30:49

Google can do to

30:51

improve upon this? And perhaps even

30:54

if the quantitative data doesn't

30:56

bear out, the fact that

30:58

so many people feel like those buttons

31:00

don't do anything is the

31:02

damage just done, or is there a way to

31:04

kind of return

31:06

to to, you know, a trust

31:08

in the in the way that this works?

31:10

Yeah. So

31:10

something we know about these tools is that

31:12

it's really important that they work well

31:14

and they're not just sort of placebos because

31:18

if you have them there just to make people feel like they're in control,

31:20

but then they don't actually work, people tend

31:22

to lose a lot of trust in the platform. And

31:24

I think we've kind of reached that point with

31:26

YouTube where people are like, I

31:28

don't really know if I actually have control. We

31:32

definitely, at a minimum, recommend that there's

31:34

a lot more YouTube could be

31:36

doing around like, making those

31:38

controls easier to understand and access.

31:40

So like I mentioned, we learned that, like,

31:42

people are frustrated. They don't know

31:44

if they're using do not recommend channel, is it going

31:46

to prevent similar recommendations? Like, what is

31:48

it actually doing? So I think there could be a

31:50

lot of clarity around that, and they could really

31:53

think about the actual user

31:55

experience here. But then I

31:57

think on a higher level, as I mentioned before,

31:59

there's an opportunity

31:59

for YouTube to really reimagine

32:02

how it

32:02

thinks about user feedback so that it's not

32:04

just sort of

32:05

one signal weighted, like, maybe not

32:07

as heavily as other signals. but is

32:09

treated as the most important signal in

32:12

folks determining what they want

32:14

to watch. And I think they also have an

32:16

opportunity to really think about what

32:18

would it what would it look like

32:20

for people to actively

32:22

state what their preferences are and

32:24

actively curate versus passive

32:26

data? collection. because I think the

32:28

model right now is somewhat paternalistic

32:31

where you have a platform that's

32:33

kind of passively inferring

32:35

what it thinks you might wanna watch

32:37

versus you telling the platform. This is

32:39

what I want to be able to do on

32:41

the platform. And I think it actually would

32:43

be a win win for both users

32:45

and the platform itself because

32:47

people would wanna use the platform or if it

32:49

actually worked effectively. Yeah.

32:51

Well, thank

32:52

you so much for joining us today to talk about

32:54

this study. Of course, folks can head

32:56

over to the MOSLA Foundation

32:59

website even just by

33:01

going to mozilla dot org. But are

33:03

you available online if folks wanna follow the work

33:05

that you're doing?

33:07

Yes. So you can find me at

33:10

TwitterBARicks, where

33:12

I have been posting about some of this

33:14

work. Awesome. Thank

33:15

you so much for your time today. We really

33:17

appreciate it. Thank you so

33:18

much.

33:20

Alright. Coming up next, the

33:22

big tech companies going toe to

33:24

toe with high end VR

33:27

headsets. And actually, are you really surprised that

33:29

they're all battling each other on this

33:31

battle front? This episode of Technus Weekly

33:33

has brought you by Impress Scale.

33:35

The statistics for ransomware attacks

33:37

are alarming, of course, cybersecurity

33:40

and cyber criminals. can actually penetrate up to ninety

33:42

three percent of company networks. That's

33:45

according to beta news dot com.

33:47

And it's not just large

33:50

organizations either forty six percent of

33:52

small to medium sized businesses have actually

33:54

been victims of ransomware

33:56

attacks. Now the Imprimarily scaled

33:58

cloud backup solution provides the

34:00

security that you actually need to

34:02

manage backups and secures

34:04

them from hackers or adverse

34:07

events. That way, you can, you know, sleep easier

34:09

at night knowing that your company will

34:11

never have to pay a ransom. Wouldn't that

34:13

be nice? You can back

34:15

up and protect SaaS applications, endpoints, and

34:18

servers as well as being able to

34:20

execute disaster

34:22

recovery on-site or in the

34:24

cloud. Now every company

34:26

needs a secure endpoint data

34:28

protection solution that's easy

34:30

to install and easy to manage and

34:32

infra scale actually offers

34:34

this. They integrate with

34:36

Hyper V and VMware. That

34:38

enables site to site failover

34:40

with orchestration

34:42

in the infascale cloud. And you can keep

34:44

packers away. Immutability really is

34:46

at the core of how their product

34:49

is designed. Your data is,

34:51

of course, in encrypted even in storage so no one can

34:53

alter it with infra scale backup and

34:56

disaster recovery. You're gonna be ready for

34:58

whatever comes your way, whether it's

35:00

a server

35:02

crash, Maybe it's human error, malicious activity, and natural

35:04

disaster, whatever it may be. According to

35:06

Gartner, the average cost of downtime caused

35:10

by a local or site wide incident across all industry

35:12

sectors is actually five thousand

35:14

six hundred dollars per minute That

35:18

comes out to three hundred and thirty

35:20

thousand dollars per hour.

35:22

Yikes. Clients and partners

35:24

alike, including MSPs

35:26

and VARs, trust

35:28

infra scale backup and recovery

35:30

solutions to help keep their

35:32

business running to eliminate downtime

35:34

and day the loss with the most

35:36

cost effective enterprise grade data protection solutions. More than

35:38

sixty five thousand companies in fact are

35:40

protected by inflow scale. If

35:42

disaster strikes to

35:44

your applications, Data, your

35:46

systems are recovered and available

35:48

in record time. Their

35:50

award winning world class support

35:52

is there to ensure your business is

35:54

protected twenty four seven now

35:56

also with SSD at

35:58

no extra

35:59

charge. So whatever your

36:01

data, whatever your environment, infrastructure

36:04

provides continuity and resiliency

36:06

for your business. All you gotta do is visit

36:08

infra scale dot com slash wit.

36:11

Do that, go there and you can sign

36:13

up for a free demo and see how

36:15

InfraScale protects your business today.

36:17

That's infraScale dot com. slash

36:20

twitch, and we thank him for scale for their

36:22

support of TechNews weekly.

36:24

Alright. So

36:26

I saw this news a little bit ago and it kind

36:28

of peaked my interest. The, you know, Micah, you have a

36:31

Mediquest to I have a

36:33

Mediquest to right now

36:35

when it comes to that price category. And by

36:37

the way, meta very recently raised

36:40

the price of the meta quest to to

36:42

four hundred dollars. So making it less

36:44

affordable, but

36:46

still is still kind of in the range of affordability, I

36:48

suppose, when you're talking about, you know, the

36:50

next generation kind of of

36:54

VR hardware without going all in on,

36:56

like, a major kind of VR computer

36:58

system to back it

37:02

up. So Meta has this meta quest too. It's really kind of

37:04

the go to for that price category. It's

37:06

become the easy thing for people to recommend as

37:08

long as you can get over the hooks that Meta has

37:10

into it.

37:12

But this news actually kind

37:14

of shows that, you know, big tech

37:16

big tech kind of follows in

37:20

its own small and exclusive club. They're all chasing the same

37:23

things. Right? ByteDance, who's the

37:25

parent company of TikTok, just

37:28

revealed this morning the PICO four headset.

37:31

And it really seems to be

37:33

a direct competitor to the

37:36

MediQuest to when you're talking about

37:38

this kind of level of

37:40

VR, the self contained VR

37:42

headset level of VR.

37:44

It has inside out cameras, of course,

37:46

It's powered by the Qualcomm Snapdragon XR two, eight

37:49

gigs of RAM. It has twenty

37:51

one sixty by twenty one sixty

37:53

pixels per eye. So

37:56

that's not bad. A hundred twenty eight to hundred fifty six gig storage options.

37:59

Oh, I mean, basically, most of what

38:01

I'm reading here sounds very, very similar to what

38:03

we have with the MediQuest too.

38:06

Pricing for this ranges from four

38:08

hundred twenty nine euros if you're getting

38:10

the hundred twenty eight gig storage model.

38:14

to four hundred ninety nine euros if you're getting the larger two hundred fifty

38:16

six gauge gig storage model.

38:19

What you do get

38:21

here that you don't give it the

38:23

Metiquis right out of the boxes. If you can see that headset there, the strap

38:25

that goes around your head on the

38:27

back of it is

38:30

is the battery. I don't know if that's the only battery, but that's

38:32

that's a battery pack. And ultimately,

38:34

what that does, obviously, that allows

38:37

to extend the life of this

38:39

without having to plug it in and charge

38:41

it as often. But it also distributes

38:44

the weight. And that's

38:46

that's definitely a complaint that I've had about the MediQuest

38:48

two is, you know, there it's

38:50

an accessory that you can get to get that

38:52

added battery, and I think the

38:54

battery does place itself on

38:56

the back if you buy that, but that's an added cost.

38:58

Here, it's kind of part of the

39:00

hardware. So and then you can see the

39:02

controllers there. all these

39:04

controllers end up seeming pretty

39:06

familiar but different enough to have

39:08

their own kind of style and player. Now the

39:10

thing about the P04 is it's

39:12

not set to ship in the US. I don't

39:14

know if that's like for

39:16

good or just for now. But right now,

39:18

it's launching in several European and

39:20

Asian countries. preorders in

39:22

October, shipping October eighteenth.

39:24

So the hardware itself when you

39:26

compare apples to apples looks very

39:29

similar with the meta quest

39:32

two. But this is in

39:34

light or in comparison to

39:36

what we're expecting

39:38

next month month from

39:40

meta. And we've known

39:42

about a device coming from meta that

39:44

they're working on called the Cambria.

39:46

I think I think the Cambria might be the kind of

39:48

the whatever the the

39:50

in development name. I don't know if that's the

39:52

official name of our codes

39:54

and launches. Yeah. The code

39:56

name exactly. But the

39:58

Cambria is gonna be a really big step

40:00

up. I'm I'm super curious about

40:02

this. So the the the PICO

40:04

four headset isn't meant to compete with this because

40:06

it's actually the the Cambria

40:08

device offers

40:10

more. it will and actually, we're we're not gonna wait very long, by the

40:12

way, to see this. MetaConnect is on October

40:14

eleventh. And from what

40:17

I understand, it's almost a

40:19

certainty that we're gonna see this.

40:21

Mark Zuckerberg has already teased in

40:23

the past that the newer kind

40:25

of upgraded Quest hardware will be seen

40:27

in October, and then you got this event

40:29

in October where they usually introduce

40:32

hardware. So it's almost a guarantee that we're gonna

40:34

see it. But what does it

40:36

offer? It offers eye

40:38

tracking. It offers facial tracking.

40:40

So these are these are things, you

40:42

know, that are a step up from what we're used to

40:44

on this consumer level v r right now, if you can imagine

40:46

being in AVR experience and

40:48

having eye tracking with the

40:50

person that you're talking to. Instead of that, just

40:52

being kind

40:54

of like kinda goes like very

40:56

generic, you know, eye

40:58

eyes fixed forward facing wherever

41:00

the face happens to

41:02

be facing now you might

41:04

have the possibility of having a more

41:06

realistic eye contact experience,

41:08

that sort of thing. Facial tracking

41:12

could potentially I don't know how they do that effectively, but

41:14

obviously they're, you know, they've figured out

41:16

a system so that avatars

41:18

and VR aren't just, you know,

41:20

randomized between smile and,

41:22

oh, you're talking. So we'll do

41:24

a mouth open animation, you

41:26

know, that sort of -- Right. --

41:29

maybe Maybe these things can be a little bit

41:31

more realistic and and really kind

41:33

of increase the realism of the whole

41:35

experience. And that's not to mention the

41:37

higher resolution screen, which personally, I

41:39

feel like is a really important part. I know

41:41

some people feel like that's less important than other

41:44

things. I so disagree.

41:46

I think it's incredibly important.

41:48

I'm I'm with you. because

41:50

that is the difference, I

41:52

think, for like, there are two

41:54

things that are the difference for people who

41:57

get sick in VR. One

41:59

is the frame rate or refresh rate and the other is the resolution

42:01

of the screen. The higher the

42:03

resolution you can get and

42:05

the better the fresh

42:07

rate you can get, the less likely you are to

42:10

fill to feel ill while you are

42:12

using the

42:14

device. So a thousand percent,

42:16

I think that's so incredibly important.

42:18

And, like, I would even

42:20

say, stop trying to do all of these other

42:22

features until you get

42:24

this perfect. because that is the thing that's going to keep people

42:26

from wanting to be part of,

42:28

you know, a longer term VR

42:30

experience is if every time they use it,

42:32

they get

42:34

ill, That's not -- Yeah. -- that's -- Oh, for sure. -- not great.

42:36

And it really only takes a couple of those

42:38

experiences for you to be like, yeah, I've tried

42:40

VR. I don't like And they've

42:43

talked to her forever. Yeah. If ever. So

42:45

yeah. I totally agree.

42:48

Now this

42:50

Cambria device, from my understanding, more geared

42:52

towards to

42:54

pro users, also kind of

42:56

meant maybe more for business because

43:00

the thing is apparently gonna be a lot more expensive

43:02

than the four hundred dollar price point that we're

43:04

used to right now. It's also

43:07

gonna have a mix VR or,

43:09

sorry, mixed reality capabilities. So it's not just

43:11

gonna be all entirely within

43:13

the goggle. There will

43:15

be a little a little bit of

43:17

hand off and interplay between what's in the in the room

43:20

versus in the goggle. And so

43:22

super curious to see how that all

43:24

develops, but

43:26

you know, again, going back to PECO, which

43:29

they have a pro version

43:31

of the PECO four that is

43:34

rumored and has been known about for a while now. Actually,

43:37

when Pico announced earlier

43:40

this week that they would be making an announcement

43:42

about new

43:44

hardware, people who knew, you know, were following this closely,

43:46

kind of expected that there would be the

43:48

pico four and the pico four pro

43:51

announced. But with

43:54

Oculus having their new hardware next month, you

43:56

know, maybe PICO has a a different plan to

43:58

reveal their kind of competitor to that.

44:01

the I know if that does so many favors or maybe there's a

44:04

technological reason. But essentially,

44:06

this pro version of the P04

44:08

would I mean, at

44:10

least with the limited information that we have right

44:12

now, go somewhat toe to toe on features.

44:14

It will have eye tracking,

44:17

It'll have facial tracking. I don't know about

44:19

the mixed reality kind of use case

44:21

of it, but it'll definitely be

44:24

much more expensive focusing on prize users

44:26

focusing on those pro users.

44:28

So I think what what kind of struck

44:30

me about this is just kind of

44:32

yet again realizing

44:34

that, like, these big tech companies are actually playing

44:36

in a very limited

44:38

kind of pool together. It's

44:40

like they're all, you know, so

44:43

often we're talking about social media

44:46

and this one, you know, this company does this

44:48

thing and then, you know, like, b

44:50

roll for example and then suddenly everybody has

44:52

their b roll or club house. And then suddenly,

44:54

everybody has their club houses. They're very but but

44:56

it's all kind of interplayed within

44:58

this small group that

45:00

that just kinda tosses the ball back and forth. And I think we're

45:02

starting to see that in VR. Of course VR is a

45:04

little bit different because there aren't a

45:06

whole lot

45:08

of players creating for,

45:10

you know, creating hardware for VR

45:12

of this scale, of this potential scale

45:14

anyways. Meta really takes that blows that

45:16

out of the water ByteDance, you

45:19

know, powering TikTok, I

45:21

could see and TikTok

45:23

just on having so much

45:25

momentum right now could be really interesting.

45:27

And that's not even considering which is somewhere right around the corner. I

45:30

I don't know. Do you

45:32

know, like, are you aware

45:34

of the latest machinations there? Is this,

45:36

like, next year that we might see an

45:38

Apple VR or no one knows probably

45:40

for sure? Yeah. We we don't know for

45:42

sure. It could be next

45:44

year. I think the

45:46

first introduction

45:48

could could possibly be at

45:52

WWDC which

45:54

happens in the summer -- Yeah. -- of

45:56

twenty twenty

45:58

three. But Who knows?

45:59

Because Apple is reportedly

46:02

working on both a VR

46:04

and an AR headset and

46:06

potentially a Mixtures of Honey device as

46:09

well. And so what gets

46:11

released when and what is the

46:13

final product or final products is

46:16

all up in the air. And yeah.

46:18

I I don't know. We'll have to we'll have to

46:20

wait and see. But eventually, Apple will

46:23

have a v slash a

46:25

r headset as part of

46:27

the the market. and that's gonna be a

46:29

pricey affair as well. Of

46:32

course, Apple doing what it's done so many

46:34

times where

46:36

it kind of it comes in later than, let's competitors

46:38

do it on doing something.

46:40

But in, you know, in the

46:42

the magical way that Apple is able

46:44

to do it, they're able to,

46:46

you know, use that time to create a project product that

46:49

when they come out really

46:51

seems like, okay, I know

46:53

you're coming late but you're really

46:56

bringing it when you come. And, you know, it it

46:58

can often be kind of like

47:00

a a defining moment

47:02

for that particular hardware category. So

47:04

I can totally see Apple doing that as well. But I think this of paves the way

47:07

and, you know, that's gonna be incredibly expensive

47:09

from what we understand. These are gonna be pretty

47:11

pricey. So it kinda seems

47:14

like these these these different

47:16

levels of VR are

47:18

starting to become a little clearer

47:20

as far as, like, okay, four three

47:23

hundred fifty, four hundred dollars has been the

47:25

norm for most people. And may may you know,

47:27

I'm sure that will continue to be a price

47:30

point that people can buy into. but the really

47:32

exciting stuff for this

47:34

because this is, you know,

47:36

it's a new technology and all new in

47:38

air quotes, but it's

47:40

a new technology with momentum right now, that

47:42

really stands to benefit from that

47:44

premium experience in really

47:46

interesting ways. And

47:48

so I'm I'm curious to

47:50

kinda see how this, like, upper

47:52

tier, upper Echelon of

47:54

VR actually defines the

47:56

category and if

47:58

this can kinda move that momentum along and I, you

48:00

know, love agree

48:02

or disagree, but Apple is

48:06

really good. creating that momentum when they finally decide to to enter in,

48:08

so I see that happening too. But

48:10

but yeah. So there

48:12

you go. PECO hardware. I don't

48:14

know if we're gonna see I mean, PECO has

48:16

released some VR hardware in the

48:18

US. They've actually been around since twenty

48:20

fifteen, so they've been doing VR longer than

48:22

Meta has. when you consider that. They were

48:24

acquired by ByteDance last

48:26

year. So with

48:28

ByteDance backing,

48:30

super curious to see kind of where

48:33

PECO resides in this realm. And if if

48:35

some of this hardware, you know,

48:37

comes in in a larger

48:39

scale to the US, which is something that

48:42

Meta has, you know, has got totally

48:44

buttoned down being able to bring its

48:46

hardware pretty much anywhere. Can Pico do

48:48

the same with the by Dan? remains to be

48:50

seen. But There we go,

48:52

Shay. Alright. Up

48:55

next is my

48:57

story of the week. going back into

48:59

AI and large language

49:02

models. That that

49:04

may not be the term now that I'm thinking about

49:06

it. In any case, they are

49:08

language models that are gigantic.

49:11

Before we do that though, I

49:13

wanna take a quick break so I can

49:15

tell you about recruiter who are bringing

49:17

you this episode of Tech

49:20

News Weekly. You know

49:22

it. We know it. Hiring is challenging.

49:25

Luckily, there's one place we're finding qualified candidates

49:28

is easy. That's zip

49:30

recruiter dot com slash TNW

49:34

In fact, Just recently, we were able to hire someone

49:36

from ZipRecruiter as we had

49:38

a spot open and needed it

49:40

filled. And our

49:42

new member of the continuity team is

49:45

fantastic. And ZipRecruiter,

49:47

thank you for helping

49:49

make that happen. Hiring business

49:52

while also running a business is

49:54

a difficult thing to do. You

49:56

already are trying to focus on the day

49:58

to day, but now you need somebody. And to

50:01

try and sort of split attention like that can be

50:04

very complicated. So let

50:06

ZipRecruiter handle it for you. It does the work

50:08

for you. It's powerful

50:10

technology finds and matches the right

50:12

candidates with your job. You can easily

50:14

review recommended candidates

50:16

and invite your top choices to apply. That often leads

50:18

to a quick response

50:20

because they go, oh, they they they

50:22

want me to apply to the

50:24

job. That's That's great.

50:26

I'm hopping right on that. Four out of five

50:28

employers who post on ZipRecruiter get

50:30

a quality candidate within the first day

50:32

And ZipRecruiter is the number one rated hiring site based

50:35

on the g two satisfaction ratings

50:37

as of January first,

50:39

twenty twenty two. Try

50:42

zip recruiter for free today

50:44

at zip recruiter dot com slash

50:47

TNW That's

50:49

zip recruiter dot com slash

50:52

TNW ZipRecruiter,

50:54

the smartest way to

50:57

hire. Alright. I am

50:59

super into AI

51:02

stuff as of late. And there

51:04

was a a new report out

51:06

about it's an MIT

51:08

technology review. about Google's deep

51:10

mind and a chatbot that they're working

51:12

on. So

51:14

in typical chatbot

51:17

speak. These systems are

51:19

backed by language it

51:21

is large language models. Yay. Okay. I got

51:23

the term correct. I

51:26

was doubting myself and I shouldn't have. They use large

51:28

language models which essentially pull

51:30

in a bunch of of

51:33

text from all over

51:36

and use it to kind of inform the system

51:39

how humans talk,

51:41

how they interact, how

51:44

they respond. And so

51:46

I showed off in Google's

51:48

AI test kitchen. You know,

51:50

I think it was a couple weeks ago.

51:53

an example of one of

51:55

Google's systems, Lambda, at

51:57

work, where I could give it a prompt and

51:59

it would talk back to me. And

52:01

of course, there was the whole situation where an

52:04

engineer claimed that

52:06

the chatbot was sentient

52:08

and it was not. But

52:11

that led to a whole history of stories.

52:13

And so with all of that

52:15

comes this kind of ongoing conversation

52:17

that we have

52:20

about planning

52:22

and preparing for the impact

52:24

of these technologies because

52:28

as I've said before on

52:30

the show, you know, it seems like a lot of the time there's

52:32

this rush to make

52:34

cool stuff happen and it's only

52:36

after that that we see the impact that the

52:38

technology has and see the

52:40

negative potential negative impact that the

52:42

technology has, and then we try to fix it

52:44

in post. And this

52:46

is not one of the situations where fixing

52:49

it in post is just a a

52:51

reasonable action. So

52:55

Google's deep mind is

52:58

working with the large language

53:00

model called Chinchilla. I love all

53:02

of these different names. We've got deep

53:04

mind chinchilla. and

53:06

what the new tool is called

53:08

is Sparrow, that is the

53:10

chatbot. And

53:12

Sparrow its purpose is to answer questions

53:14

that people ask and

53:18

use Google

53:20

Search to

53:22

provide more information that the

53:24

person would need and also to use

53:26

Google Search to actually answer the question.

53:29

So think about right now, if I

53:31

were to go on to google

53:33

dot com and type

53:36

in how to make a

53:39

bundt cake. Likely at the top,

53:41

there will either be a YouTube video

53:43

where a portion of the

53:45

video is highlighted and it

53:48

shows this is how you make a bundt cake

53:50

or in this case it shows you recipes

53:52

for bundt cake and you can click on one

53:54

of those and pop it open real easy

53:56

and see how to make a bundt What

53:58

what Sparrow is aimed at

54:00

doing is kind of being the

54:03

answer to those questions in

54:06

a way that feels a little bit more personable. So I

54:08

could ask that. And instead of sort of

54:10

seeing a bunch of different links

54:12

and finding one myself, then it

54:16

in theory, do a better job of

54:18

saying, oh,

54:20

this recipe for a bundt cake seems

54:22

to be getting a lot of attention.

54:25

here's how you go about create

54:27

making a bundt cake.

54:29

And what is unique about

54:32

this system

54:34

is that Chinchilla,

54:36

as the the large

54:38

language model behind Sparrow, of

54:40

course, has lots of data that may

54:43

include some questionable responses, or

54:46

including, you know, if you were to ask it, if it's a

54:48

person, it could respond. Yes. I'm a person.

54:50

Sparrow is aimed at not doing

54:52

that. First, there are

54:54

twenty three rules that

54:56

the researchers and developers

54:58

working on Sparrow have sort

55:00

of built into this tool can't

55:03

offer financial advice. It cannot make threatening

55:06

statements. It can't claim to be a person,

55:08

and then many others. Those

55:10

are just three, so twenty

55:12

other rules. And

55:14

when it answers, it

55:16

answers with multiple responses

55:18

and lets a the

55:20

human being who is using the system

55:24

Choose which answer they

55:26

feel is the best one.

55:28

And so

55:30

by combining the AI and large language

55:32

model that is, of course, based on

55:34

human response with

55:36

actual human

55:38

rating and response to what the system is

55:42

saying, the hope, the

55:44

theory, the hypothesis

55:46

is that this

55:49

will upon responses that the

55:51

artificial intelligence chatbot

55:55

provides. Of course,

55:58

with

55:58

this, you are working with a group

56:00

of people that you hope you can trust to

56:03

answer and rate appropriately.

56:06

and that's always something that has to be considered. And

56:08

of course, this is not the be

56:10

all end all, but

56:14

one researcher at

56:16

hugging face, which is an AI startup said,

56:18

about Sparrow. It's, quote, a

56:20

nice next step that follows a

56:23

general trend in AI where we are more seriously

56:25

trying to improve the safety aspects of

56:27

large language model deployments. Of

56:29

course, the system still

56:31

does make mistakes. including going

56:33

off topic, making up random answers that aren't actual answers

56:36

to questions. And

56:38

in in the MIT technology

56:42

review, piece says determined participants were also able to make the

56:44

model break rules eight percent of

56:46

the time. Older

56:49

models broke rules three

56:51

times more often than Sparrow has.

56:54

So it is quite an improvement to see it

56:56

not break the rules all that often, but

56:58

it is still able to break

57:00

the rules And so that is,

57:02

of course, of concern. I just find all of this fascinating

57:04

from the perspective of

57:08

It's it's it's it's so interesting how

57:10

humans create the AI

57:12

and then the AI

57:14

does its thing and it

57:18

answers to a human being and

57:20

the human being

57:22

then improves upon the

57:24

AI by saying this was

57:27

good, this was bad, but the AI

57:29

the whole time is mimicking human

57:31

behavior. And so there's so much humanity

57:34

involved in artificial intelligence,

57:36

particularly whenever it comes

57:38

to language learning models. and I find

57:40

that so fascinating the way that that all kind of works together. I

57:43

think I think what I get from what

57:45

you just said and whether this is what

57:47

you meant or not this is

57:49

kind of what it what it what it pops into my brain

57:51

is that when AI misbehaves, it's

57:54

actually kinda

57:57

a reflection of how humans misbehave. Right? Yeah.

58:00

Absolutely. Because that's it's pulling from that

58:02

language model. And so

58:04

we we get some surprise

58:07

when AI does that, but it's actually not surprising at all. It was

58:09

created by humans. It's based

58:11

on data fed to it by humans,

58:13

chosen by humans,

58:16

that data is created by humans. So, of course, like, if

58:18

if an AI ends up, you

58:20

know, going down a bad path,

58:23

It's because we get we've set it down

58:26

that path whether we realized it or

58:28

not. Yeah. It's it's a reflection of

58:30

humanity. It's a What is it looking

58:32

to us? Yeah.

58:34

Absolutely. We stare into the

58:36

AI void and it stares back at us

58:38

and says, I am you. I

58:40

am you. Look, don't don't get down on

58:43

me, man. Yeah. Maybe on my

58:46

fault. And so yeah.

58:48

I I think that this this

58:50

specific kind of language model and

58:53

and chatbot is

58:56

incredibly compelling to me because I

58:58

can so clearly see

59:00

how this can be implemented

59:02

as a future technology

59:05

and on improving our voice assistance

59:07

and on improving our even our searches that we do

59:10

online, being able

59:12

to just

59:14

just have a system understand

59:18

what we mean when we ask a question where we don't

59:20

have to meet it more on its playing

59:22

field. like you have to do with

59:26

with the art

59:28

AI stuff that we have right

59:30

now. You you do have to meet those

59:32

systems. You have to meet OpenAI's dolly

59:34

two on its playing field in

59:36

terms of using the proper terms

59:40

and knowing what some keywords will do, the same thing goes

59:42

for deep wait. Deep mind?

59:44

Does that what it's called? No.

59:46

the Oh, I can't think of

59:48

the discord one. Oh, why

59:51

am I playing? Mid journey.

59:53

Mid journey. There you

59:56

go. Yeah. all of those -- Yeah. -- there's a specific way to talk to those

59:58

systems -- They ask you to talk to

59:59

them. Yeah. And so if

1:00:02

you can break out of

1:00:04

that when it comes to our virtual

1:00:06

assistants, which we also have to talk to in kind of

1:00:08

a specific way. I think that's

1:00:10

gonna be great. and there will be improvement there. So it's about

1:00:12

I'm glad that there's this balance

1:00:15

that is being more

1:00:18

attention to than I've seen in the

1:00:20

past in terms of making

1:00:22

sure that

1:00:24

to the best of our ability, they

1:00:27

don't reflect the the bad

1:00:29

parts of humanity and that -- Yeah. --

1:00:31

they they they do their job and

1:00:33

they do it well and they do

1:00:35

it kindly and don't,

1:00:38

you know, result in horrible responses.

1:00:41

make break our hearts. I think the

1:00:44

the one thing that's popping into my

1:00:46

mind right now

1:00:48

and, actually, I don't know if

1:00:50

if you would remember or

1:00:52

even know about this ad campaign. I think

1:00:54

it was from well, yeah. So it was from

1:00:56

the late eighties It was

1:00:58

a it was a drug free America

1:01:00

campaign, and

1:01:02

it has the quote, who taught you

1:01:04

how to do this stuff? You alright?

1:01:06

I learned it from watching you, and it was the you know, walks in

1:01:09

and discovers that the the sun has,

1:01:11

like, a, you know, has some marijuana

1:01:13

or something like that. came

1:01:16

from. I learned it from watching you came from that. Yes. I know

1:01:18

that I know that quote. I learned it from watching you,

1:01:20

but I know it. I learned it from watching you.

1:01:24

Yes, it was an anti it was, you know, an drug

1:01:26

kind of campaign that,

1:01:28

you know, at that time, was

1:01:30

everywhere, and it was comical, you

1:01:34

know, and that's why it kind of has stayed in the stayed

1:01:36

stayed around and and been named and

1:01:39

everything. But it just kinda AI

1:01:41

kinda reminds me of that. Like, when we're

1:01:43

really surprised that AI is going down this bad

1:01:45

road, it's like, where did you learn this

1:01:48

stuff? AI? You alright? I learned it

1:01:50

by watch in you. That's exactly in.

1:01:52

Wow. It's all ears back around. There

1:01:54

you go. I'll send you the ad after we're done,

1:01:56

and you can watch it. Thank you. get a

1:02:00

good laugh. But we have reached the end of this episode

1:02:02

of Tech News Weekly. You

1:02:04

don't wanna miss it every Thursday. We do this

1:02:06

show. If you go to Twitter

1:02:08

TV slash T and W,

1:02:10

you can find all the ways to subscribe,

1:02:12

so you don't miss it, and we appreciate

1:02:14

that. Yeah. You know what else?

1:02:16

We appreciate. Those of you who

1:02:18

joined ClubTwit, because when you do

1:02:20

that, you are supporting us

1:02:22

directly. Club Twitter is an awesome

1:02:24

place with awesome friends

1:02:26

and awesome stuff you get as part of

1:02:28

the club. For seven dollars

1:02:30

a month or eighty four dollars a year, you

1:02:32

can join club Twitter. twitch dot tv slash

1:02:34

club tweet. When you

1:02:36

do, you get access

1:02:38

to all of our shows with no ads,

1:02:40

so it's just the content without those

1:02:42

interruptions. You get access

1:02:44

to the to the Twitch

1:02:46

plus bonus feed that has extra content you

1:02:48

won't find anywhere else before the

1:02:50

show, after the show behind the scenes,

1:02:52

etcetera. And access to the

1:02:54

ClubTwist Discord Server, a

1:02:56

fun place to go to chat with your fellow Club TWIT

1:02:58

members, and those of us here

1:03:00

at Twit. Along with that, you will also gain access

1:03:02

to my show hands on

1:03:04

Mac, which is a short

1:03:06

format show

1:03:08

all about making the

1:03:10

most of the devices you have on

1:03:12

hand, the Apple devices you have on hand.

1:03:14

So that's the Mac, that's iPhone,

1:03:16

iPad, etcetera, as well

1:03:18

as poll the rots,

1:03:20

hands on windows. That is

1:03:22

the windows tips and tricks

1:03:24

show with lots of great

1:03:27

stuff. for making the most of that Windows

1:03:29

PC you're rocking. Yeah.

1:03:32

So check that out to it dot tv slash club to

1:03:34

it again seven bucks a year bucks a month,

1:03:36

eighty four dollars a year, and

1:03:38

we appreciate your support. Thank you.

1:03:40

If you would like to follow

1:03:43

me online, I'm my cosargent on many of social media network where you can

1:03:45

head to chihuahua dot coffee. That's ch

1:03:48

IHUAHUA dot coffee, where I've got

1:03:50

links to the places I'm most

1:03:52

active online. me

1:03:54

out later today if you're a subscriber

1:03:56

to for hands on

1:03:58

Mac or later this week for

1:04:01

the TechAI Radio Show on

1:04:04

Saturdays, which I host with Leo

1:04:06

Leport. The Radio Show heard around the

1:04:08

world where we take your questions and do our best to

1:04:10

answer them, or on Tuesdays for

1:04:12

iOS today, which I record with

1:04:15

Macao Rose Mary Orchard, where

1:04:17

we talk all things, iOS, TVOS, home,

1:04:20

podcast, etcetera. Jason, how

1:04:22

about you? Well, you can

1:04:25

find me on Twitter at jason Howell. You can find

1:04:27

me talking all about Android, on all

1:04:29

about Android. That's why we called it all

1:04:31

about Android. We got t u slash

1:04:33

AAAA We do that every Tuesday evening and

1:04:35

have a lot of fun with

1:04:37

that. And, yeah, so much more behind

1:04:39

the scenes doing a lot of Leo shows as you

1:04:41

are as well, Micah. So if

1:04:44

you see in his shows, sometimes you're seeing our

1:04:46

work as well. I mean, producing

1:04:48

from behind the scenes, that's what we're doing

1:04:50

there. Anyways, Thank you for watching. Thank you to John studio

1:04:52

for pushing the buttons and making the

1:04:54

the show happen there. Thank you

1:04:58

to Burke for all the help that he provides behind the

1:05:00

scenes, making sure our guests are ready for the

1:05:02

show and all that. And

1:05:04

thank you

1:05:06

to you for watching and listening, we would not have a show for one for you,

1:05:08

so thank you for doing that. We'll see you next

1:05:10

time on tech news weekly. Hi,

1:05:12

everybody. Goodbye.

1:05:14

Hey. I'm Rod Pyle, editor in chief bad actor magazine. And

1:05:16

each week, I joined with my cohost to

1:05:18

bring you this week in space, the

1:05:20

latest and greatest news. the

1:05:23

final frontier. We talk to NASA chief space scientists, engineers, educators,

1:05:25

and artists, and sometimes we just shoot

1:05:27

the breeze or what's hot and what's

1:05:29

not in Facebook and

1:05:32

TV, and we do it all for you are fellow true believers.

1:05:34

So whether you're an armchair adventurer or

1:05:36

waiting for your turn to grab a slot in Elon's

1:05:38

Mars rocket, join us on this weekend's space

1:05:41

to be part of greatest adventure of

1:05:44

all time.