Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
Ryan Reynolds Here for Mint. With price of
0:03
just about everything going up during inflation, we
0:05
thought we'd bring our prices down. so to
0:07
help us we've run into reverse auctioneer which
0:09
is apparently a thing. Mint Mobile unlimited a
0:11
premium wireless hundred to get thirty thirty. They
0:13
to get thirty but to get twenty twenty
0:16
twenty the to get what you want. Even
0:18
if a team to team could team fifteen.
0:20
Just fifteen bucks a month. So give it
0:22
a try it Mint mobile.com/switch. Forty five
0:24
times are kind of agreements. Prose type of them. Please come on
0:26
it. And customers are limited Time Unlimited. More than forty gigabytes
0:28
per month. Floats Culture and that Mint mobile.com. seasons
0:31
change, why not your tech? Upgrade now
0:33
during the Dell Technologies Summer Sale event
0:35
and save on select PCs, like the
0:37
XPS 16, powered by
0:39
Intel Core processors. You'll be able
0:41
to bring your most intensive projects
0:43
to life with built-in AI, minimalistic
0:45
design, immersive visuals, and cinematic audio.
0:47
When you shop online at dell.com/deals,
0:50
you'll have access to exceptional tech
0:52
and electronics, plus free shipping on
0:54
everything. Amazing prices await you for
0:56
a limited time only at dell.com
0:58
slash deals. That's dell.com slash.
1:10
Amazing prices await you
1:12
for a limited time.
1:15
Only at dell.com/deals That's
1:17
dell.com/deals. Daily.
1:22
Tech News show is made possible
1:24
by it's listeners. thanks to all
1:26
of you including Irwin, Stir Can,
1:29
Hayes, Philip, Shane and welcome in
1:31
our brand new patrons Free Bruce
1:34
and Jeffrey. A
1:36
good. To have new patrons will
1:38
go on in every make him feel at
1:40
home. Or this episode of Dns The Verge
1:43
is Sean Hollister helps us break down Entails:
1:45
attempt to stay in the Ai Pc game
1:47
A I scientists don't want to be silenced
1:49
about risk and Instagram attempt to force you
1:51
to watch as. this
1:56
is the daily tech news for
1:58
tuesday june fourth two thing I'm
2:01
Tom Merritt and I'm in
2:03
Los Angeles. And I'm Sarah Lane
2:05
and I'm at Studio Animal House. They're all sleeping.
2:08
I am the show's producer, Roger Chang. And
2:10
joining us, Senior Editor and founding member of
2:12
the Verge, Sean Hollister, welcome back. Hi.
2:16
It's good to have you, man. Hi.
2:19
That's the sound of someone who's been covering
2:21
Computex for the rest of his
2:23
life. Yeah,
2:25
good to have you with us, Sean. No, it's not
2:27
bad at all. I was nearly going to go and
2:29
then we decided not to. Yeah, that would have been
2:32
fun because of what? It's in Taipei, right? Oh,
2:34
yeah. I went in 2012. It was
2:36
a blast. It's been 12
2:38
years since and I was looking forward to it. Well,
2:41
we are going to talk about the Intel news
2:43
out of that with Sean in just a second.
2:46
Also, we should let you all know, Meta announced
2:48
its Connect 2024 event will take place September 25th
2:50
and 26th. So
2:53
put that on your calendars. Let's see what's
2:56
in the rest of the Q-tip. Samsung
2:59
has added two features to the 32-inch
3:02
Odyssey OLED G8 gaming monitor,
3:04
thanks to the built-in NQ8 AI
3:06
Gen 3 processor. The monitor can
3:08
now use models to
3:10
upscale content and games to near 4K when
3:13
you're using the built-in TV
3:15
and gaming hub apps. There's also
3:17
a pulsating heat pipe. Yep,
3:20
that's what it's called. That was not announced with
3:22
the rest of the monitor's details back in January,
3:24
but this will help prevent burn-in. The
3:26
32-inch Odyssey OLED G8 is available now for
3:29
$1,299. The
3:32
Raspberry Pi is offering
3:34
an AI kit. Indeed,
3:36
it's finally here. Halo's
3:38
Halo 8L M.2 accelerator
3:40
is powering it, adds
3:43
13 tops of power to
3:45
a Raspberry Pi 5 microcomputer, and draws
3:47
less than 2 watts of power and
3:50
can be passively cooled. This would
3:52
let you run some smaller models like coding
3:54
assistants or image editors on your Raspberry Pi.
4:00
held talks at the end of last year
4:02
with China Mobile to bring Apple TV Plus
4:04
to China. No U.S. streaming
4:06
video service is currently available in China.
4:08
If the deal were to be accepted
4:10
China Mobile would offer Apple TV Plus
4:13
for a monthly fee to its customers
4:15
and reportedly split revenue with Apple. The
4:17
deal would only likely be for set-top
4:19
boxes, not phones. There
4:22
was a breach at the banking company Santander
4:24
and another one affecting 650 million accounts
4:27
on Ticketmaster that were both
4:29
linked to a cloud storage
4:31
platform called Snowflake. So
4:34
Snowflake engaged third-party security
4:36
companies, Mandiant and CrowdStrike,
4:38
to investigate. They
4:40
now have reported the results of their
4:43
investigation. They found no evidence of vulnerabilities
4:45
on the Snowflake platform or
4:47
compromised Snowflake personnel accounts. It
4:50
appears that the breach may
4:52
have been achieved by accessing
4:54
Snowflake accounts from clients who
4:56
did not have multi-factor authentication
4:59
on. Microsoft
5:01
365 Education is the target
5:03
of complaints from NOI, the
5:06
N-O-Y-B, that's a privacy watchdog organization
5:08
based in Europe. They're the
5:10
ones who single-handedly stopped there from being a
5:12
user data sharing agreement between the U.S. and
5:14
Europe for more than a decade. Right
5:17
now, it's a request for the
5:19
Austrian Data Protection Authority to investigate
5:21
some vague terms, the
5:23
possibility that Microsoft unfairly pushes
5:25
compliance onto schools, and a
5:27
tracking cookie that appears to
5:29
exist without consent. Quick
5:32
quiz. Anybody remember what NOI-B
5:34
stands for? I
5:37
heard this from Rob Dudwood on Daily Tech
5:39
Headlines today. None of your business. NOI-B.
5:44
I actually looked it up just to
5:46
make sure that I had it spelled right and I
5:48
hadn't thought about what it meant. Intel
5:52
unveiled its Lunar Lake platform at Computex.
5:55
That is its chip that
5:57
is CoPilot Plus compliant. those
6:00
kinds of laptops in this autumn. This
6:02
is just six months after the previous
6:04
gen Meteor Lake. Intel promises up to
6:06
14% faster CPU performance,
6:09
50% better graphics performance, and
6:11
60% better battery life. Sean,
6:14
you've been digging into this directly with the
6:16
folks from Intel. What are the things you
6:18
found most important about Lunar Lake? Well,
6:21
everybody's interested in this being, you know, the
6:23
new AI chip to go up against the
6:26
ones from AMD, Nvidia, Qualcomm, the ones
6:28
that are going to be in these
6:31
co-pilot plus PCs. But at
6:33
next year, nobody's going to care about that
6:35
because it will be table stakes that all
6:37
of these chips will have this
6:39
amount of AI generative processing
6:42
or more thanks to a
6:44
little co-processor in Intel, in
6:46
AMD, in Qualcomm chips in
6:48
your laptop. Everybody's going to
6:50
have this. So that part
6:52
doesn't matter quite as much as the
6:54
fact that Intel has to finally do
6:57
something about Apple. They have to do
6:59
something about ARM chips, ARM
7:01
chips using a different instruction
7:03
set than the x86 chips that
7:06
Intel and AMD have been building
7:08
for, you know, all of our
7:10
lifetime in computing. A
7:13
few years back when Apple introduced those
7:15
M1 processors and said, hey, we're going
7:17
to build our own ARM chips, all
7:20
of a sudden we're like, here
7:22
is a tremendous amount more battery
7:24
life from a laptop. The processing
7:26
is good. It doesn't need as
7:28
much cooling. And it's beyond time
7:30
for Intel to do something about
7:32
that. This is Intel coming out
7:34
swinging there. Yeah, I thought you
7:36
did a great job in the article on the Verge
7:38
sort of breaking down the comparison,
7:40
the simplicity, what's going on
7:42
in these chips. And
7:45
a lot of it sounded very much like an
7:47
M series chip, right? There's
7:49
integrated RAM, there's a simpler die.
7:53
Walk us through a little bit of the highlights of that.
7:56
Yeah, yeah, yeah. So last year Intel
7:59
did for... or what would call its
8:01
biggest architectural shift in 40 years. And
8:04
what that broke down to was we
8:06
are building our chips more like smartphones
8:08
in almost every way. For
8:11
many, many years, smartphones have built
8:13
their chips. They call them system-on-chip
8:15
in that you have all the
8:17
different components of the system, the
8:19
Wi-Fi, the Bluetooth, the cellular connectivity
8:21
of the phone, the processing, the
8:23
graphics. All of them are little
8:25
building blocks on the same die.
8:28
Intel started doing that last year, but
8:30
like you said, six months later, they've
8:32
thrown out a lot of that
8:35
to try doing it again because what they did
8:37
last time didn't quite work, their first step, but
8:39
it didn't quite work. The
8:41
first time they said, we're going to have
8:43
tiny cores, we're going to have medium cores,
8:45
and we're going to have big cores in
8:47
addition to our graphics cores, in addition to
8:50
our AI cores, we're going to put all
8:52
these cores together and we're going
8:54
to put the right loads in the right place at the right
8:56
time. They built some Windows
8:58
scheduling with Microsoft so that they
9:00
could theoretically run most of your
9:02
apps on these tiny, tiny, tiny
9:04
cores so that unless
9:06
you were doing something big with
9:09
your computer, you could save a lot
9:11
of battery life by keeping your tasks
9:13
on this low-power island. The
9:16
low-power island still exists this year, but
9:18
they have ditched those tiny cores entirely.
9:20
They could not keep the tasks on
9:23
that low-power island the way they want.
9:25
They were eating up too much battery.
9:28
They said, we're going to make some better medium
9:30
cores. They called
9:32
them the efficiency cores, the E cores. They've been
9:34
around for a while, but the
9:36
ones this year are so much
9:39
more powerful and potentially so much
9:41
more efficient. We would know for
9:43
sure if Intel ever labeled the axes on its
9:45
charts, but it doesn't. They
9:47
are so much more powerful and so much more efficient that
9:50
not only do they deliver I
9:52
think it's two times more
9:55
performance, up to
9:57
four times more performance than the tiny
9:59
cores do. last year, but
10:01
they also can deliver more
10:03
performance at the same power
10:06
as the P-cores, the performance
10:08
cores in last year's
10:11
chips. So today's E-cores more
10:13
powerful than last year's P-cores
10:16
at the kinds of lower clock speeds you typically
10:18
run a laptop at. Yeah, and so
10:20
that containment that you're talking about with Windows still
10:22
exists, but it just keeps them on these efficiency
10:24
cores that are at par with last year's power
10:26
cores. So most of the time, you'll
10:29
get that great battery life, it sounds like. Yeah,
10:31
we're not clear about the technical details,
10:33
the difference between last year,
10:36
the scheduler saying, oh, we think these apps
10:38
should be here in the low power cores.
10:41
And now we should move to the high
10:43
cores. And what they're now calling containment zones,
10:45
which is like we are keeping them on
10:47
the E-core. We are keeping Microsoft Teams will
10:50
run almost entirely in your E-cores
10:52
now. And so when you're doing
10:55
your video call, at least with
10:57
Microsoft's video calling solution, they
10:59
say you'll get 35% battery life out of that, which,
11:01
you know, before they were heating
11:04
up the entire chip, it was using up a
11:07
certain amount of power. Integrated
11:09
RAM is going to be new for Intel
11:11
users. People with Macs and
11:13
ARM processors in general have
11:15
gotten used to that. But this is only going to have
11:17
16 and 32 gigs of RAM. And that's all
11:20
you get, right? Yeah, this might be
11:22
the biggest shift for anybody who's who likes
11:24
to tinker with PCs, or build your own
11:26
PCs. If you expect to do that kind
11:28
of thing with your laptop, if you expect
11:30
to open up a hatch on the back
11:33
of your laptop and add additional memory sticks,
11:35
that is not happening with lunar like at
11:37
all. Now, again, this is not unusual for
11:39
smartphones, smartphones have had memory
11:41
on sitting on top of the
11:43
chip for many, many years now.
11:45
But now Intel is doing that
11:47
too. There's two chunks of memory
11:49
that'll physically sit below the CPU
11:51
that has all the
11:54
compute tiles and different tiles in it. The
11:56
memory will now be part of the same package that
11:58
the year heatsink goes. on top of. And
12:01
so if you want more than 32 gigabytes
12:04
of memory, you're out of luck. Thankfully, you get
12:06
a minimum of 16, whereas some
12:08
of the competition still ships laptops with 8
12:10
gigabytes of memory. Get 16 at a minimum.
12:12
Yeah. And Arrowhead Lake is coming. That's more
12:15
of a desktop PC. But they
12:17
say if you want to do upgradeable RAM, look for
12:19
that. I thought it was
12:21
just for desktop, but Arrow Lake
12:23
will be laptop and
12:25
desktop. They planned to
12:28
do that with Meteor Lake
12:30
too. There were Meteor Lake
12:32
desktop parts. That is Meteor Lake being last
12:34
year ships. They got canceled. They did not
12:37
come out. So folks on
12:39
desktop had to move directly to
12:41
a different system. But now Lunar
12:44
Lake may be exclusively for these thin laptops,
12:47
Arrow Lake, laptop and desktop. Yeah. And thank
12:49
you. It's Arrow Lake, not Arrow. Arrowhead is
12:51
water. Arrow Lake is the chip. There
12:55
is an Arrowhead Lake. Yes, exactly. So
12:57
yeah. Before we let you go, how
13:00
do you think this stacks up with
13:03
AMD's announcement with the Ryzen AI 300
13:05
announcement from yesterday? I
13:09
believe AMD is going to be
13:11
in a similar place as Intel
13:13
when it comes to that AI
13:16
co-processing. I
13:19
imagine that
13:22
the Intel chips could be more powerful, but
13:24
I feel like there's something that Intel isn't
13:26
telling us. I don't know what it is.
13:28
When I spoke to their marketing
13:31
guru, Rob Hallock, Intel said it, their
13:33
marketing guru, Rob Hallock, he told me,
13:35
I think his quote was that he
13:38
expects Lunar Lake laptops to be the
13:40
tippy top of the mountain. That's
13:42
what he said, tippy top of the mountain
13:44
when they ship later this year. And so
13:46
he's, when he says that, I verified with
13:48
him that he meant that he believes these
13:51
will be more performant, have better battery life
13:53
than AMD's Strix Point, which is those AI
13:55
300 chips, and Qualcomm. be
14:00
better than Qualcomm. But Intel
14:02
is not letting people touch lunar lake
14:05
yet. It has not shared
14:08
crucial things like how fast
14:10
will these chips be in terms of
14:13
gigahertz? Because if,
14:15
for instance, if you take for granted
14:17
that they're telling the truth that it will be
14:21
16% faster CPU speed, clock
14:23
for clock, compared
14:26
to mid-ear lake, it is
14:28
16% faster clock for clock, meaning if they
14:30
are at the same clock speed. But if
14:32
lunar lake runs at a slower clock speed
14:35
than mid-ear lake, you may not see
14:38
a 16% improvement in CPU performance.
14:40
It's possible that these get
14:43
better battery life, better AI, but maybe
14:45
don't have faster performance than last year's
14:47
chips. And I would say that for
14:49
many, many years now, I have
14:51
not felt like buying a new processor
14:54
for an x86 computer has made
14:56
my computer faster. It
14:59
feels like it's only keeping up with whatever the
15:02
increases in software. Yeah, yeah. Well,
15:05
you know, the lunar lake chips
15:07
are coming in Q3. Arrow
15:10
Lake's supposed to be in Q4. Lunar
15:13
Lake has an MPU with 48 tops. If
15:15
you're keeping score, that makes it co-pilot plus.
15:18
Wi-Fi 7, Bluetooth 5.4, two Thunderbolt 4 ports, just
15:21
as part of the chip set. Everything
15:24
else, every laptop will have 200-volt ports
15:26
now. The theoretically, left side, right side,
15:28
you plug in your cable, you'll get fast charging end
15:30
data, no matter which lunar lake laptop you buy.
15:32
That right there, something to celebrate, I
15:34
think. Yeah, yeah, indeed. We'll definitely go
15:37
read Sean's full article. He's got all
15:39
the details at The Verge. And
15:41
before we go, anything else you wanna tell folks about, Sean?
15:45
Yeah, I'm good. Thanks for having me. Go
15:48
check it out, theverge.com. John Hollister, thank you
15:50
so much. Have a great one.
15:53
Thanks, John. Well, Instagram
15:55
confirmed it is testing unskippable
15:57
ads in its feed, various
15:59
use. have come across the test. It
16:02
has a countdown timer letting you know how
16:04
long before you can resume scrolling. Instagram,
16:07
while confirming that that test was
16:09
actually real, didn't give any more
16:11
details on the test but said, hey, you know, if
16:13
we end up making this permanent we'll tell you all
16:15
about it someday. Anyone
16:18
besides an advertiser who would like
16:20
this to become permanent do you
16:22
think, Sarah? No, no. I don't
16:24
think anybody wants this. I also
16:26
think people are used to this.
16:29
Anybody who spends time on YouTube and doesn't
16:31
pay for YouTube Premium is
16:33
going to have at least a few seconds
16:36
of ads, even skippable ads. You still have
16:38
to sit through a few seconds of them
16:40
and in some cases, you know, you
16:42
don't even have the options to skip.
16:44
So this is not abnormal
16:47
and I really don't... I
16:49
mean, I was
16:51
about to say I don't fault Instagram for
16:53
doing this. I mean, it's as
16:55
a user, nobody wants this.
16:57
But it's sort of like, okay, Instagram
17:00
has a few options. You either make
17:03
people pay for an ad-free version of the
17:05
service, which could
17:07
possibly be a thing. You do
17:10
this sort of like, hey, you can't get
17:12
away from the ads and sometimes you can't
17:14
scroll past them and that's just part of
17:16
the deal. Or what's
17:19
the alternative that... because
17:21
the company wants to make money. That's
17:23
the bottom line here. So how do
17:25
you annoy
17:27
the users but not enough for the user to just
17:30
say like, okay, I don't want to hang out here
17:32
anymore. Would you partner
17:34
with popular creators? Take a
17:36
cut of the creators cut.
17:39
I mean, that's already happening. But you
17:42
know, you get to the point
17:44
with a platform
17:46
like Instagram and what else
17:49
do we expect? Yeah, I
17:52
didn't expect unskippable ads in Instagram because
17:54
it is contrary to the user experience.
17:56
It's one thing in a video where
17:59
we're all... used to pre-roll ads for
18:01
YouTube to say like guess what, you can't
18:03
skip the pre-roll ads anymore. It
18:06
is another thing in a system that's been
18:08
around for 10 plus years to be like,
18:11
we're built on keeping you scrolling, we
18:13
are now going to stop you scrolling. I
18:16
think that's risky. Like I totally get what
18:18
you're saying, you're absolutely right, they wanna monetize
18:20
and maybe the solution is, they'll be like,
18:23
hey, for $5 a year you can have
18:25
no ads in Instagram. Although Instagram
18:27
is one of those weird places where people seem
18:29
to kinda like the ads sometimes. Like I actually
18:31
saw people saying, you know, I use Instagram ads
18:33
and I've bought lots of things from them, but
18:35
I will stop paying attention to them if they
18:37
do this. It's almost like
18:40
people don't wanna get rid of the ads,
18:42
they like them the way they are in some
18:44
cases. Some of you are probably reviled by that
18:46
notion, but I personally,
18:48
I'm not emotional about it, but I will
18:50
say I go to Instagram as
18:52
sort of like a side diversion. And
18:55
if I'm scrolling through and suddenly it's like, oh,
18:57
you have to wait 20 seconds for this ad,
19:00
I'll just go somewhere else. I
19:03
won't go away mad, I'll just go away.
19:05
You know what I'm saying? Right,
19:07
yeah. Yeah, no, I'm with
19:09
you on that. I do think
19:12
Instagram is sort of
19:14
unique in the sense that, yes,
19:17
so many things that are Instagram ads
19:19
are things for sale, easy to buy,
19:21
look at the photo sets or a
19:23
video of the cute boots that you
19:26
were looking at a week ago because
19:28
we know where you've been, that kind
19:30
of stuff. It is,
19:32
it works on Instagram. I
19:34
have thought, I mean, not that many
19:36
things based
19:38
on Instagram ads, but it's definitely,
19:41
it captures your attention. If
19:44
nothing else, especially when Otis the
19:46
dog is on his Instagram, obviously
19:48
not me because it's his Instagram
19:50
account, but he gets served up
19:52
lots of, you know, like dog
19:54
bed and dog toy and cute
19:57
dog hoodie ads. And
19:59
I mean, It's all I can do not
20:01
just be like click so easy. Yeah, right
20:04
just one click But
20:06
but at the same time when
20:08
Instagram ads first rolled out and I think that
20:10
was what are we talking like? 2016
20:13
yes, I'm around there. It was it was right
20:15
around the time that Instagram introduced
20:17
the algorithmic feed which at the time those of
20:20
us who had been sort of right or die
20:22
for the platform were just Like this is the
20:24
worst and now we all just got used to
20:26
it. You know, it's not going anywhere. It's everywhere
20:29
But at the time I you know, I'd say it
20:31
hide ad Every ad
20:33
I said hide I'd be like,
20:35
yeah, we'll just run out eventually. No, that's
20:37
not what happened They just came back
20:39
in force and now I just I just deal with
20:41
them. Yeah. Yeah, we're all used to it now Yeah,
20:44
but yeah that whole sort of you
20:46
can't scroll without watching the ad That
20:50
tends to anytime that happens
20:53
to me in any similar form where
20:55
I'm sort of forced to deal with
20:57
an ad I'm just like la la la
20:59
la do not care Yeah, I
21:01
am fighting you now and you
21:03
know, it's just not a great experience.
21:05
I suspect That there's
21:07
a principle of the internet that more people complain
21:09
about a thing then will actually act on it
21:12
So I think that is worth taking
21:14
into account here But there
21:16
is a principle of it. I've seen many
21:19
people I think storied squirrel just said it I've seen
21:21
lots of other people say well if the ads are
21:23
good, you won't skip them anyway, so why force people
21:25
to watch them? And that
21:27
is a principle that if it were a hundred
21:30
percent true, there would never be unskippable
21:32
ads, right? Apparently unskippable
21:34
ads are effective because the advertising
21:36
world wants them and that's
21:39
why Instagram is testing it To
21:41
be fair to Instagram. They're testing
21:43
it. They're seeing does this cause
21:45
people to stop scrolling? And
21:48
if it does they won't continue with it and
21:50
they'll be able to tell their advertisers. Yeah, here's
21:52
the numbers We did a we did a search
21:54
and it doesn't work on our platform. That's kind
21:56
of what I'm hoping selfishly happens But
21:58
you know what if it doesn't doesn't stop people
22:00
from skipping away. If I'm wrong and
22:02
I'm like, you know what, even though
22:05
it's unskippable, I don't skip away. If
22:07
they're fine-tuned the exact right amount
22:09
of time for an ad on
22:11
Instagram to be unskippable before people do
22:14
close, we're probably gonna
22:16
be stuck with it. I
22:18
also, you mentioned something like $5 a year to
22:21
not get ads. I mean, if it was really that low,
22:23
I'd be like, yeah, of course, it's like $5 a month.
22:27
In this world where people are paying $20 a month for
22:29
Netflix, Instagram's gonna have to be really cheap
22:32
if all you're doing is taking away the
22:34
ads. Exactly, but it also, and
22:36
again, this is
22:38
completely different from the hoaxes that
22:41
go around every so often where they're like,
22:43
all right, Facebook is gonna make you pay.
22:46
Yeah, yeah. He's like,
22:48
no, meta, Facebook, WhatsApp,
22:50
anything in the meta
22:52
universe, they
22:55
are not going to force you to pay
22:57
for their product. They're going to force
22:59
you to pay for the version of
23:01
the product that you want, possibly,
23:04
but yeah, we're definitely in the
23:06
testing phase right now. And Larry in
23:09
Atlanta asked, will Instagram charge advertisers to
23:11
make their ads unskippable? Yes, yeah, they'll
23:13
be able to charge more. They'll be
23:15
able to charge a higher CPM,
23:17
as they call it, if an ad is unskippable,
23:20
right? That's the idea. I mean, imagine being in
23:22
the marketing room where you figured out
23:24
the perfect slogan, which is at the end of the
23:26
ad? Like, no, people need
23:28
to see the whole ad. We all worked
23:30
very hard on that. The advertiser believes that
23:33
the ad will be more effective if it's
23:35
unskippable. That's why they want it. And if
23:37
it turns out that that's true on this
23:39
test, then Instagram can charge more for it.
23:41
Yeah, exactly. Exactly.
23:44
Well, it's been an exciting week
23:47
at ComputeX already. And next week
23:50
is no different. Apple's Worldwide
23:52
Developers Conference, WWDC, is next
23:54
week. And Apple Vision Show
23:56
is on it. Each
23:58
week, Eileen Rivera and talk all
24:00
things Apple while also keeping an
24:02
eye on the competition as well.
24:05
We want to know what everybody's
24:07
doing in the space. Do join
24:09
us at applevisionshow.com. Hi,
24:16
this is Janice Torres from Yoquiero Dinero.
24:18
From a local business to
24:20
a global corporation, partnering
24:23
with Bank of America gives your operation
24:25
access to exclusive digital tools, award-winning
24:28
insights, and business solutions
24:30
so powerful you'll make
24:33
every move matter. Visit
24:35
bankofamerica.com/bankingforbusiness to learn more. What would you like
24:37
the power to do? Bank
24:39
of America NA. Copyright 2024. Hey
24:42
Dave. Yeah
24:44
Randy. Since we founded
24:46
Bombas, we've always said our socks, underwear,
24:48
and t-shirts are super soft. Any
24:51
new ideas? Maybe sublimely soft.
24:53
More disgustingly cozy. Wait, what?
24:55
I got it. Bombas.
24:58
Of certainly comfortable essentials for yourself and
25:00
for those facing homelessness. Because
25:02
one purchase equals one donated. Wow, did we
25:04
just write an ad? Yes.
25:07
Bombas. Big comfort for everyone. Go
25:09
to bombas.com/acast and use code ACAST
25:11
for 20% off your first purchase.
25:16
The Claude 3 model family from Anthropic
25:18
is your one-stop shop for enterprise AI.
25:21
With models at every point on the
25:23
price performance curve, you no longer have
25:25
to make tradeoffs between intelligence, speed, and
25:27
cost. Claude 3 Opus
25:29
sets new industry benchmarks for intelligence. Sonnet
25:32
strikes the perfect balance between skills and
25:34
speed and Haiku is the fastest and
25:36
lowest cost model in the market. Perfectly
25:39
designed for high volume, high speed use
25:41
cases. Join the thousands of enterprises who
25:43
trust Anthropic to keep them at the
25:46
frontier. Visit anthropic.com/Claude
25:48
today. More
25:52
than two dozen current and former
25:54
employees of OpenAI, Google's DeepMind, and
25:57
Anthropic signed a letter posted at
26:00
righttowarn.ai called a right
26:02
to warn about advanced
26:05
artificial intelligence. So the
26:07
URL is very apropos. Seven
26:09
of the 26 signees used their name. The
26:11
rest were anonymous. The letter was
26:13
also endorsed by A.I. pioneers Yoshua
26:16
Benjio, Jeffrey Hinton and Stuart
26:18
Russell. The letter called
26:20
on advanced A.I. companies to
26:22
commit to four principles, allowing
26:24
employees to anonymously raise concerns
26:26
about A.I. risks without fearing
26:28
retaliation of being sued for
26:30
violation of confidentiality agreements. The
26:32
letter specifically called for the
26:35
protection of trade secrets and
26:37
intellectual property when raising concerns
26:39
in four specific categories. So
26:41
specifically, no prohibitions on disparagement
26:43
of a company over risk related concerns.
26:45
So if I work
26:48
for a company or otherwise know
26:50
about something that the company is
26:52
doing, I don't have to
26:54
worry about legality. A process for
26:57
anonymously raising concerns to company boards,
26:59
regulators or appropriate independent
27:01
organizations. Kind of piggybacking
27:03
off of step one, but same
27:05
idea. Number three, support for a
27:07
culture of open criticism. We
27:10
should all say how we feel. Number
27:12
four, promise to not retaliate for raising
27:15
risk related concerns in public if all
27:17
other processes have failed. Yeah,
27:19
so this is this is this is tricky because
27:24
if you are a lawyer
27:26
and I've worked
27:28
with many great lawyers who've expressed that
27:30
this is often the way lawyers work,
27:33
you tend to say, let us
27:35
protect against all possible eventualities. Let's
27:37
set the line as far away
27:40
as possible from from us getting
27:42
sued. And so that
27:44
disparagement thing that you were talking about
27:46
in that first point is
27:48
one that they will do when they are giving
27:51
someone severance pay, for example, or
27:54
giving someone stock options. They'll say you you
27:56
will get these options. Granted, you will get
27:58
the severance pay if you agree. that you
28:00
will not disparage the company in any way
28:02
for the next whatever five to ten years.
28:05
That's boilerplate. I've
28:07
signed one, Sarah signed one, Roger signed
28:09
one, you know, we've all gone through
28:11
that. I've signed more than one. Yeah,
28:14
exactly, right? And so what
28:16
the lawyer is doing there is saying we don't
28:18
care what the disparagement is, we're just going to
28:20
set the line out there as far as possible
28:22
to avoid that damage, right? What
28:24
these folks are saying is we get that. And
28:26
we're, and they say in the letter, we
28:29
don't want people to reveal intellectual property, we
28:31
don't want people to reveal trade secrets, but
28:34
if they're concerned, legitimately
28:37
concerned about any kind
28:39
of harm, and they're very broad
28:41
here from imminent things like bias
28:44
or prejudice to existential
28:46
threats to humanity, right? They're like,
28:49
whatever the legitimate harm they're concerned
28:51
about, they should have a pathway
28:53
to do it. And an NDA,
28:55
a confidentiality agreement, a non-disparagement
28:57
agreement, shouldn't prevent them from
29:00
warning people about serious
29:02
risks. It's all going to
29:04
come down to like, okay, but who gets to define what
29:06
a serious risk is? Right. In
29:09
the letter specifically, speaking
29:11
of serious risks, the
29:13
letter, Sainese said, we
29:16
also understand the serious
29:18
risks posed by these technologies.
29:20
These risks range from the
29:22
further entrenchment of existing inequalities
29:24
to manipulation, misinformation, the
29:26
loss of control of
29:28
autonomous AI systems, potentially
29:30
resulting in human extinction.
29:35
So yeah, not mincing words here. And it's
29:37
not as if this letter was
29:39
written saying humans are about to
29:41
become extinct, and these companies need to shut
29:45
down immediately. Not at all. There are a lot
29:47
of people working in this field who
29:50
care very
29:53
much and who know what they're talking about
29:55
and have worked in various systems that
29:57
are creating AI systems that they're not. the
30:00
rest of the world will then be using. And
30:03
if there is an issue, we
30:06
should be able to voice that issue for
30:08
the good of humankind. Yeah. There's
30:10
some pretty respectable pioneers in AI. Jeffrey Hinton
30:12
is the one who actually stepped down from
30:15
Google because he wanted to be independent to
30:17
shine light on these kinds of risks. And
30:20
they're saying not there are
30:22
problems. I think a lot of people are going to
30:24
look at this that way. They're saying if
30:27
there are problems and there will
30:29
be some problems of all manners
30:32
of levels, right
30:34
now there's no way for people to
30:36
alert folks outside of their
30:38
immediate manager. And essentially what they're saying
30:40
in this letter is we don't think that's good enough because
30:43
there is a financial interest and a
30:45
stock investor pressure to err
30:48
on the side of not revealing
30:50
a risk that could cause a
30:53
damaging event to
30:55
happen. And they're saying we
30:57
need to work on a process that
30:59
allows people to do whistle blowing if
31:02
things are risky. That doesn't mean they're
31:05
saying there are risks right now, but
31:07
they're saying the possibility is there and
31:09
there's no path for someone to warn
31:11
everyone about it that doesn't go through
31:13
people who have a vested interest in
31:15
not warning people about it. Well,
31:18
Tom, knowing, you know, having
31:21
read the letter and I think
31:23
some very valid points were brought
31:26
up and I understand what
31:28
the goal is here. How
31:32
much do you think this is going to
31:34
change what OpenAI is doing at its next
31:36
board meeting, for example? OpenAI
31:38
was the only one that responded to this.
31:41
And they said we have a hotline that
31:43
people can call anonymously if they are concerned
31:45
about something. We're very
31:48
responsible. Trust us. A
31:50
hotline? Yeah, for an internal employee. They do. They
31:53
have a hotline that you can call. Oh, for
31:55
an internal employee. Yeah, for an anonymously. But you're
31:57
warning someone within OpenAI and that's kind of beside
31:59
the point of this. This is saying if they
32:02
go to a regulator, if they go to a
32:04
journal, a responsible journalist, they
32:06
shouldn't be retaliated against. It's the
32:08
start of a conversation. It's not
32:11
necessarily a conversation that's going to go anywhere.
32:13
Anthropic and Google have not responded, at least
32:15
that I have seen as of this recording.
32:19
But it is definitely trying to put some
32:21
pressure on this and it's going
32:23
to need someone else to pick up the
32:25
baton, at least on the regulatory side, if
32:27
not on the independent agency side, to put
32:30
pressure on the companies to say you need
32:32
to create something like this, something akin to
32:34
what Meta does with its oversight board for
32:36
something much less potentially
32:38
risky, social networks in
32:40
my opinion. AI I don't think
32:42
is dangerous right now but certainly
32:45
has the potential to be used
32:47
in dangerous fashions. I support
32:49
everything in this letter the way it
32:52
was written for sure. Yep, yep, yep.
32:54
For a site rather than hindsight. Indeed.
32:56
Well patrons, stick around for the
32:58
extended show Good Day Internet. We're
33:01
going to talk about the fact that scientists
33:03
at Cambridge have developed an extra thumb for
33:06
humans to use. That's right. You want three
33:08
thumbs? Stick around. We'll tell you how. No,
33:11
wouldn't you have four thumbs? Oh, just
33:14
the one on each hand and then yeah.
33:16
Can't wait to hear more, Tom. I love
33:18
thumbs. Why not? Why not? Who doesn't? Yeah,
33:21
just a reminder you can catch
33:23
the show live Monday through Friday
33:25
at 4 p.m. Eastern, 2,200 UTC.
33:27
You can find out more at
33:29
dailytechnewsshow.com/live and we're back to it
33:31
all again tomorrow. Scott Johnson joining
33:33
us. See you then. The
33:41
DTNS family of podcasts, helping
33:44
each other understand. Simon Club hopes
33:46
you have enjoyed this program. What
33:55
does it take to drive innovation
33:57
in today's AI-driven world? Join exactly
34:00
executives from across industries as
34:02
they dive deep into how
34:04
technological advancements are redefining leadership.
34:06
From generative AI to cybersecurity,
34:08
cloud computing to sustainability, the
34:10
Conversations with Leaders podcast from
34:12
Amazon Web Services covers it
34:14
all. Join the conversation and
34:16
subscribe today, available on all
34:18
major podcast platforms. Hi,
34:23
this is Janice Torres from Yoquiero Dinero.
34:25
If you own or operate a business,
34:27
whether it's a local operation or
34:29
a global corporation, partnering
34:33
with Bank of America could be your smart move. By
34:36
teaming with Bank of America, you'll enjoy
34:38
exclusive digital tools, award-winning
34:40
insights, and business solutions
34:43
so powerful you'll make every
34:45
move matter. Position your
34:47
business to capitalize on opportunity in
34:49
a moment's notice. Visit
34:52
bankofamerica.com-bankingforbusiness to learn
34:54
more. What would
34:56
you like the power to do? Bank
34:58
of America, N.A. Copyright 2024.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More