Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
So there's a handful of people that I'll schedule like
0:02
a month monthly facetime call with and in most of
0:04
them, you know, almost all of them are in fact
0:06
all of them are not local. And
0:08
then there's a handful of people that I try to do
0:10
lunch with like once a month and my
0:12
good friend Sam, he and I had our monthly
0:14
lunch today and we went to a place. But
0:18
I had a problem. During
0:21
lunch, there was music
0:23
outside, which was good. But
0:28
the Jack Brown burger joint trolled me
0:30
because they were playing a fish
0:33
album during the entire lunch.
0:37
And all I could do was think about how
0:39
happy you would be if you were there or
0:41
if you at least knew this was happening as
0:43
it was happening in retrospect. I should have like
0:45
FaceTime or something just to be like, listen to
0:47
this joke. And the
0:50
worst part of all, the worst part of
0:52
all, you could tell
0:54
me what songs I heard and I would probably
0:56
be like, sure. But
0:58
there were a couple of songs that even
1:00
I recognized as fish songs. Like, you know,
1:02
not only did it have the vibe of
1:05
fish, but I like had heard the
1:07
songs before and recognize them. And
1:09
I forget which ones they were. The
1:11
only one I know by name is bouncing around the room and that
1:13
was not it. But there was one
1:15
and I'm sure this is describing half a
1:17
fish catalog, but where it was repeating the
1:19
same phrase over and over again. And it
1:21
was very catchy. That's fairly common. Yeah, exactly.
1:23
But anyways, David Bowie, maybe I'm
1:26
kind of proud of you that you recognize that
1:28
it was fish. I'm not sure I could do that.
1:30
I don't know any of their songs. Maybe
1:33
I could pick it up based on vibe, but I don't think I've
1:35
even heard that much. So like you must. When
1:38
are you listening to fish so much that you recognize
1:40
songs? I can give you a good heuristic, John. If
1:43
you hear a song that you don't recognize, you don't
1:46
think you've ever heard it on the radio before. Look
1:49
around the room. And if the
1:51
widest guy in the room is slightly
1:54
bopping his head to it. That's me,
1:56
though. And zero other
1:58
people are. There's a
2:00
decent chance. It's fish It
2:03
could be anything like I don't know I
2:05
think my chances of record Spontaneously recognize like
2:07
you're at a restaurant these music playing the
2:09
background spontaneously recognizing fish I think my odds
2:11
are very low. I guess I'd have
2:13
to look for somebody with the little red
2:16
blood cell You know pattern
2:18
on there on their clothing and if they were bopping to
2:20
it or something then I could figure it out Right now.
2:22
I know what that is That's the one thing I can
2:24
recognize Marco taught me what that is and now I see
2:27
it on people's like license plate surrounds I'm like, oh one
2:29
of them Anyway,
2:32
the worst part Marco the worst part of
2:34
this entire lunch and about the only bad
2:37
part of this lunch because I really do
2:39
Enjoy Sam so very much and I guess
2:41
yes, did you like some of it? It
2:44
wasn't bad So
2:49
we have a new member special we have
2:51
gone back to the well and
2:54
we have done another ATP tier list
2:56
John. Can you remind us all what is
2:58
a tier list? I can't remind you all
3:00
because everybody knows what a tier list except
3:02
for all people who Listen
3:05
to this podcast, but then they've also heard the specials
3:07
before so it's a tier list You rank things you
3:09
put them in tiers multiple things can be in a
3:11
single tier the top tier is s why nobody knows?
3:13
Except somebody knows but we don't really care the point
3:16
is is better than a it's a tier list and
3:18
it's grading It's like a through f and then s
3:20
on top of a mm-hmm And
3:23
we graded all the iPods No,
3:25
at least most of them anyhow and so
3:27
I am pretty confident that we did a
3:29
pretty good job on this There was a little bit of horse
3:31
trading involved, but I'm pretty happy with where we ended up We
3:34
made a handful of people that we know
3:37
very upset and I'm sorry that you're upset,
3:39
but we're right So if you are
3:41
curious to hear this tier list or any
3:43
of the others You can go to ATP
3:46
dot FM slash join and if you join
3:48
even for about a month But you should do
3:50
more then you can get to all of the
3:53
members specials We've been trying to do one a month
3:55
for what like a year or two now I forget exactly
3:57
how long it's been but that we've racked up a fair
3:59
number over the over the course of the last several months.
4:02
There's a handful of tier lists. We
4:04
do ATP Eats among
4:06
other things. There's a lot of, there's a lot of
4:08
good stuff in there and some silly stuff. So ATP
4:10
tier lists. And if you are a member and you
4:12
would like to watch the tier list
4:14
happen, which is not required, but is occasionally
4:17
helpful. There is a super
4:19
secret YouTube link in the show notes for members
4:21
where you can go and watch it on YouTube
4:23
as well. Please do not share that. It's the
4:25
honor system, but you can check it out
4:28
there as well. It's in the show notes
4:30
for the member special. Sorry, yes, thank you. When you
4:32
go to the iPod tier list member special, look in
4:34
the show notes, the first link will be the YouTube
4:36
video. I like this tier list because they always, we
4:38
always seem to, I think they reveal something about the
4:40
things that we are ranking. Something that
4:42
we, at least I usually didn't know going in. You
4:44
think, oh, you're just gonna rank them and people are gonna,
4:47
you know, have controversies over which is good and which is bad.
4:49
But I think in the end, when you look at the whole
4:52
tier list and you kind of look at the shape of it
4:54
and how it's worked out and how contentious the choices would be,
4:57
you learn something about it. Like I think our connectors tier
4:59
list was like that. And I think the iPod one turned
5:01
out like that too. And the reason
5:03
we made some people angry is because we
5:05
know a lot of really weird tech people
5:07
with very specific and often very strange opinions,
5:10
specifically than iPods. I think you could also
5:12
say incorrect opinions. Like
5:14
they have their reasons. At least most
5:16
of them have reasons that make some sense. I think
5:18
one of the things we learned, not to spoil too
5:20
much, is that a lot of people
5:22
have, you know, all the things that
5:24
we put in tier lists, people can have personal
5:28
sentimental reasons for. We all certainly do.
5:31
And, you know, listeners do as well. And I
5:33
think iPods, more than anything we've done before, like
5:36
the people who had opinions, they swayed
5:38
heavily into the sentimental, right? It
5:41
was, you know, it was like, this was my first
5:43
iPod. I really love this thing, right? Much
5:46
more so than the past tier
5:48
list we've done. So I think, you know, maybe
5:50
the iPod at that point was the most personal
5:52
product Apple had ever made. Yeah,
5:55
I mean, honestly, like, I had a lot of
5:57
fun with this one because, like. Even
6:00
though I hardly ever really
6:02
used iPods because by the time I
6:04
could really afford decent iPods, it
6:07
was only very shortly before the iPhone really
6:09
took over. So I only really had a
6:11
couple of years with iPods, but those couple
6:14
of years, I really liked the iPods. And
6:16
this was actually fun. So, and just for
6:18
coincidence sake, I happen to have bought a
6:21
couple of iPod Nano's off of eBay a couple
6:23
of years back, just to kind of play around
6:25
with. And I took
6:28
them out the other night after we recorded
6:30
this episode and charged them up. And well, the ones that
6:33
we'll accept to charge at least, charged them
6:35
up and got to play around
6:37
with the old iPod Nano. And I will
6:39
just say, I stand
6:41
by everything I said on that episode, everything.
6:44
So feel free to listen and tell
6:46
us how wrong we are. And you too listener can
6:48
pay us $8 a month to
6:51
yell at your podcast player just a little bit more.
6:53
So we encourage you to do that. That's
6:56
absolutely great marketing. Thank you, Marco.
6:58
And by the way, our membership
7:00
episodes are DRM free. And
7:02
so if you happen to use an
7:04
iPod to listen to your podcasts, we
7:07
are fully compatible. So you can pay us
7:09
$8 a month to listen to our member
7:11
content on an iPod if you actually have
7:13
one. And you can honestly buy one on
7:15
eBay for only a few months worth of
7:17
membership fee because they're pretty cheap these days.
7:21
Indeed. And hey, what would you listen to
7:23
on an iPod if not a podcast? Well,
7:25
you could listen to music and you could
7:27
listen to music on a U2 iPod. And
7:29
so Brian Hamilton wrote in with regard to
7:31
the red and black colored U2 iPod. We
7:33
were wondering, I thought we were wondering on
7:35
the episode, certainly there was some mumblings about
7:37
it on Mastodon afterwards, how did they get
7:39
to red and black for the color scheme
7:41
of the U2 iPod? And Brian
7:43
wrote into her mind, John and us, about
7:46
how to dismantle an atomic bomb, which was
7:48
released November 22 of 20, or
7:51
excuse me, of 2004. And the
7:53
color scheme on the cover art for that album
7:56
is red and black. We're worried on that one, John, Mr. U2.
7:59
Yeah, I remember. I heard it once I was
8:01
reminded of it. I mean, here's the thing. Like I
8:03
said on the episode, it's not as if red and
8:05
black became the iconic colors of the band. This was
8:08
one album that was released, obviously, at
8:10
the same time as the iPod as part of a
8:12
promotional thing, like the iPod, the U2 iPod, the first
8:14
U2 iPod was released in October, and the album came
8:16
out in November. So it's a tie-in, right? And then
8:18
there were future U2 iPods, and they were also red
8:20
and black. But at that point, U2 hadn't released a
8:23
new album. So they're all just tied to this one
8:25
album. But they have released a lot of
8:27
albums, and they were future albums, and they were past albums,
8:30
I can tell you that this one and this color
8:32
scheme did not become heavily associated with the band. But
8:34
that's the reason. That's why they went with red and
8:36
black, because of the cover of the album. Are you
8:38
saying that as an assumption? I'm genuinely asking. Are you
8:40
saying that as an assumption? No, once
8:43
I was reminded of it, I'm like, oh, yeah, that's why they did
8:45
it. I mean, it's not a great reason, but I'm pretty sure it's
8:47
the reason. Fair enough. Max
8:50
Velasco Knott writes in
8:52
that there's also another feature, and I'm using
8:54
air quotes here, on the U2
8:57
iPod. Max writes, the U2 iPods featured
8:59
signatures of the band members on the backside. I was
9:01
fine with the black-red color scheme, but couldn't stand seeing
9:03
Bono and company on the back whenever I turned them
9:05
over. Yeah, I'd forgotten about that
9:07
as well. I mean, obviously, it's a shiny back end
9:09
that doesn't show up that much. But if you really
9:12
just wanted a red and black iPod and didn't care
9:14
about the band, the signatures on the back messed
9:16
it up a little. Indeed. Nikolai
9:19
Bronvol Ernst writes to us with
9:21
regard to the DMA and Apple's
9:23
Cut. Nikolai writes, I really enjoyed
9:25
your last show, 593, not
9:27
a European lawyer. I'm also not a European lawyer, but I
9:29
am a citizen in the EU and wanted to provide a
9:31
single European's point of view. The DMA
9:34
has nothing to do with Apple's Cut in
9:36
the App Store or how much money Apple
9:38
earns from selling their hardware. It only has
9:40
to do with ensuring fair competition, citizens' rights
9:42
to freely choose services they want to use without
9:44
vendor lock-ins on interoperability, portability, and your own
9:46
data, which we here in the EU believe
9:48
belongs to the user. That was a pretty
9:51
good summary. A lot of people have written in
9:53
to say this, but I think people will get hung up with the
9:55
idea when they're like,
9:57
Apple's Cut and how The
10:00
EU is trying to control that and they're like, the
10:02
EU is not trying to tell Apple how much money
10:04
it can make, it's just trying to do this other
10:06
thing. But the
10:09
reason it gets mixed up and the reason
10:11
people send us these emails is
10:13
because what Apple did to, you
10:17
know, supposedly comply with the DMA
10:19
while also trying to prevent competition
10:22
is an application of fees. So it's that
10:24
we, you know, OK, well, the EU says
10:27
you have to allow for competition. Apple says, OK,
10:30
sure, we'll allow competition. But all of our
10:32
competitors have to pay us an amount that
10:34
makes it so they can't compete
10:36
with us, right? And the cut we're
10:38
talking about is not Apple's cut from its
10:40
own App Store. Like when you
10:42
sell through the App Store, you pay Apple
10:44
some cut. It's the cut Apple demands from
10:46
the App Stores and the people selling
10:48
through App Stores that are not Apples on App Store,
10:51
that are selling through third party App Stores. Apple is
10:53
using money, using fees to
10:56
make the competition less competitive.
10:58
And that's what we're talking about. I know it's
11:01
even confusing when we're talking about Apple collecting its
11:03
money or Apple having its fees and stuff like
11:05
that. So I think maybe that's the source of
11:07
the confusion. And the other thing is, by the way, that
11:10
plenty of countries, including the EU, do
11:12
actually tell companies that they can't
11:15
make a certain amount of money on a certain
11:17
thing that they do. Someone wrote in to give
11:19
us the example of like credit cards, like a
11:21
MasterCard and Visa, the two big credit card networks.
11:24
I think in the EU, the fees
11:26
they charge stores to
11:28
process the credit cards are essentially
11:30
capped. And the EU
11:32
has basically said, Visa and MasterCard
11:34
own the market. You can continue to do
11:36
that, but you can't charge merchants
11:39
any more than 0.1%. The
11:43
EU has not done that to Apple. They
11:45
haven't said to Apple, hey, Apple, you can't
11:47
charge more than 10% in your own App Store.
11:50
They haven't said that at all. They haven't said anything about what
11:52
Apple can charge in their App Store. What they just want is
11:54
more competition, and Apple is saying, OK, there
11:56
can be other App Stores, but they all have to give us an amount
11:59
of money that makes it up. unattractive. And
12:01
yeah, we'll see how that flies. Again, the
12:03
EU has not yet ruled on the
12:06
core technology fee and all the other things that
12:08
they're investigating. So far, they've only ruled on the
12:11
steering provisions about how Apple
12:13
restricts the way apps in
12:16
its own app store can link out to third party
12:18
payment methods. But we'll see how
12:20
those other decisions come out in the coming months and
12:22
years. I don't know how long this is going to
12:25
take. But right now it's not
12:27
looking good for the core technology fee. Let's say that. We
12:30
asked for mostly tongue in cheek,
12:32
but we asked for Brexit style
12:34
names for Apple leaving the EU.
12:38
Jared counts was the first
12:40
we saw to suggest I
12:42
leave Frederick B Jormann suggested
12:44
accent and provided a truly
12:46
heinous but hilarious. I
12:49
presume AI generated image for this. My
12:52
personal favorite though, was suggested several times.
12:54
First we saw was from Oliver Thomas.
12:56
I quit. Yeah, that's pretty good. I
12:58
leave and I quit. We had many more suggestions. These are
13:00
thought these were the top three. The I leave and I quit are
13:03
cute, but I kind of like accent
13:05
because it's as close to Brexit and
13:07
the axe thing like the picture has
13:09
like a EU themed
13:12
Superman holding an axe and an apple. And
13:14
yes, it does look AI generated. It's interesting
13:16
how due to the way the various models
13:19
that we're familiar with have been trained, most
13:21
people can now look at an image and
13:23
identify it immediately as AI generated based
13:26
on like the shading and the
13:28
weirdness of hands and all sorts of other stuff.
13:30
It is kind of strange how quickly that happened.
13:32
But anyway, I kind of like
13:34
accent, but I don't
13:36
think we get to pick this name. So I mean,
13:38
I don't, MacBook one didn't really catch on. Neither did
13:41
a MacBook. Oh please, it sure did. Well, within our
13:43
little circle of podcasts. Yes. But I don't see the,
13:45
uh, the New York
13:47
Times running with accent or I quit. Yeah,
13:49
we don't really seem to have naming power
13:51
in the, in the greater ecosystem. If
13:55
we try hard enough, we can make fetch happen. All
13:57
right. Uh, someone anonymously wrote in
13:59
with. So
18:00
maybe there's some other factors there. How does it, how
18:02
would it possibly be measuring the sound on the inside
18:04
of your ear? Is there a microphone that's facing the
18:06
inside of your ear? I think there might be. Isn't
18:09
that how they do some of the calibration stuff? So
18:11
anyway, the point is my experience actually
18:14
using them, it
18:16
really does not feel like I'm
18:19
hearing a 95 decibel concert for three hours.
18:22
It feels like what it says of 85. Well,
18:25
how loud was the concert outside of the
18:27
year? From your seat, did you
18:29
look at the decibel meter? If I had nothing
18:31
on, what would the level be? Yes. So I did
18:33
a couple times where I would take the AirPods
18:35
out and put them
18:38
away so they turn off and just listen and watch
18:40
and see how the watch measures the concert fully. And
18:44
it was, I don't remember exactly, but
18:46
I remember it was somewhere in the high 90s, I think.
18:48
So not quite as loud as there. So maybe
18:51
the difference is that they were
18:53
coming from 105 decibels. And
18:55
they came out to 95. And I was coming, I think, from somewhere in
18:57
the 90s down to 85. So maybe that's
18:59
the cause. Or it could just be differences
19:01
in fit. I don't know exactly
19:04
how good is the seal with their artificial ear
19:06
setup compared to my actual ear. I don't know.
19:08
There's no good way to know that. So
19:11
I think the conclusion to draw here is, first
19:13
of all, what we kind of already knew,
19:15
which is they provide
19:17
some protection, suitable
19:20
for occasional concert goers, not suitable if you're
19:22
going to be working in a factory every
19:24
single day. There's different degrees of
19:26
protection that you might need. This
19:28
is not everyday protection. But also, it probably
19:31
varies a little bit between both
19:33
fit and between what exactly you're
19:36
actually listening to, like how loud
19:38
is your environment. Maybe
19:41
it can't bring down 105 decibels, but
19:44
maybe it can bring down 95 decibels. So
19:47
obviously, there are other variables here. So
19:49
I think the advice that
19:51
I would give remains the same, which is,
19:54
if you have really serious hearing protection needs
19:56
or very frequent hearing production needs, get real
19:58
hearing protection. an
20:00
occasional concert goer like me and you want
20:02
basic hearing protection for occasional concerts, this
20:05
is probably fine unless you are standing
20:07
like directly next to the giant PA
20:09
speaker. Maybe you might need a little
20:11
bit more protection. But this
20:13
seems fine to me. And every time I've
20:15
used them, I feel great afterwards and my ears
20:18
don't ring at all. And it doesn't, there's no
20:20
fatigue. So like, it seems to be
20:22
working. So maybe it just has a limit to
20:24
how much it can work. Apple
20:26
is apparently using Google Cloud infrastructure
20:29
to train and serve AI. This
20:31
is from HPC wire. Apple is
20:33
two new homegrown AI models, including
20:35
a 3 billion parameter model for
20:37
on device AI in a larger
20:39
LLM for servers with resources to
20:41
allow, to answer more queries.
20:44
The ML models developed with TensorFlow were
20:46
trained on Google's TPU. John, remind me
20:48
what TPU stands for. Tensor processing unit,
20:50
some of that we talked about, the
20:53
actual hardware on a past show and how many
20:56
billions of computations or whatever they do
20:58
and how many different operands are in each
21:00
operation. But I think it's like a Tensor
21:02
processing using unit or something. It's basically, so
21:05
Google doesn't buy its GPUs from Nvidia and
21:07
put them in, it makes its own silicon
21:09
to do machine learning. It has for many,
21:11
many years, it's not a new thing. They're
21:13
called TPUs. And that's what they're currently using
21:16
to train Gemini and stuff. And if
21:18
you pay them, just like you pay AWS or whatever,
21:20
you pay Google Cloud, I believe
21:22
they will rent you their TPUs and you can train your models
21:24
on it. And that's what Apple did. Indeed.
21:28
Apple's AXLearn AI framework used to
21:30
train the homegrown LLMs creates Docker
21:32
containers that are authenticated to run
21:34
on the GCP or Google Cloud
21:36
something. What is that? Google Cloud
21:38
computing? I don't know. Computers. GCP
21:42
is like AWS. It's Amazon Web
21:44
Services, but Google. Anyway,
21:46
to run on the GCP infrastructure,
21:48
AXLearn supports the Bastion orchestrator, which
21:50
is supported only by Google Cloud.
21:53
This is a quote from their
21:55
GitHub documentation. While
21:58
the Bastion currently only supports Google. Cloud Platform.
22:00
There you go. I should have kept reading. My bad. Google
22:02
Cloud Platform Jobs, its design is cloud-agnostic. And
22:05
in theory, it can be extended to run
22:07
on other cloud providers, Apple stated on its
22:09
AX Learn Infrastructure page on GitHub. Yeah,
22:12
so this is, I mean, we didn't
22:14
put this in the notes, but the rumors
22:16
are that the deal between Apple and Google
22:18
to use Gemini as part of iOS 18
22:20
as an option alongside a chat GPT, that
22:22
deal is reportedly getting closer. But this is
22:24
from the past of like, hey, Apple's got
22:26
these models, the one that's going to be
22:28
running on people's phones or the various ones
22:30
that are running on their phones, which are
22:32
smaller. And the big ones, they're going to
22:34
be running on their private cloud compute. And
22:36
these are Apple's own models, and they train
22:38
them themselves. And how do they train them?
22:40
They paid Google to use TPUs to train
22:42
their models. And so I feel
22:44
like this is interesting in that Google's
22:50
unfriendly relationship, let's say, with Nvidia
22:52
continues. And
22:54
their friendly relationship with Google continues. It's kind of a
22:56
surprise that Google didn't do the deal. Maybe the rumors
22:59
are, I think we talked about this on a past
23:01
show, that nobody's paying anybody for
23:03
the open AI thing, whereas maybe Google wanted
23:05
to be paid. So we'll see how this
23:07
works out. But yeah, there seems to be
23:10
a cozy relationship between Apple and Google, because
23:12
apparently Apple either doesn't have yet or doesn't
23:14
plan to have fleets of
23:16
massively parallel machine learning silicon that
23:18
they can train their models on.
23:20
But Google does. We
23:24
are brought to you this episode by
23:26
Photon Camera, the ultimate manual shooting app
23:28
for iPhone photography enthusiasts. Whether you're a
23:31
seasoned pro or just getting started with
23:33
photography, Photon Camera is designed to give
23:35
you all the tools you need to
23:38
elevate your iPhone camera experience to new
23:40
heights. Photon Camera is a beautifully designed,
23:42
easy to use manual camera app for
23:45
iPhone, perfect for both beginners and professionals.
23:47
You can say goodbye to confusing buttons
23:49
and hidden gestures. Photon Camera is very
23:52
intuitive and comes with a comprehensive manual
23:54
to help you learn the basics of
23:56
photography. They've also just launched Photon Studio,
24:00
your iPad or a monitor connected to
24:02
a spare iPhone for a big screen
24:04
preview while you shoot. It also allows
24:06
you to favorite or delete images in
24:08
real time, view metadata, and even zoom
24:10
in to expect details closely. And Photon
24:13
Enhance is the new powerful photo editor
24:15
for iPad and Mac that's also now
24:17
available. Both Photon Studio and
24:19
Photon Enhance are included free with
24:21
your Photon Camera subscription. And
24:24
here, of course, is the best part. For
24:26
our listeners, Photon Camera is offering an exclusive
24:28
deal. You can get up to 50% off
24:31
your first year by visiting
24:33
Photon.cam slash ATP. That's Photon,
24:35
PHOTON.cam, C-A-M slash ATP. Go
24:37
there, Photon.cam slash ATP, to
24:40
claim your discount and start
24:42
exploring the power of manual
24:44
photography on your iPhone today.
24:47
Thank you so much to
24:49
Photon Camera for sponsoring our
24:51
show. John,
24:56
I hear that you have
24:58
asked Apple for help and they have said,
25:00
you know what you need? You need a
25:03
Mac Studio. Because why would anyone need a
25:05
Mac Pro? This went around,
25:07
I think, a week or two ago. Apple's got
25:09
a page, apple.com/Mac slash best hyphen
25:11
Mac. And the title of the page is
25:13
Help Me Choose. Answer a few questions to
25:16
find the best Mac for you. And
25:19
when this was going around, the first thing
25:21
I did was launch this page and
25:23
I wanted to go through the little wizard and answer
25:25
a bunch of questions to see if
25:27
I could reach the win condition, which is
25:30
having this tool recommend the Mac Pro. Is
25:32
that the win condition? It is the win condition. Are
25:35
you sure? And the answer
25:37
was very clear. And I was mostly telling the truth, but
25:39
occasionally I would exaggerate to
25:41
make sure I go on the Mac Pro path. And
25:44
I did not end up at a Mac Pro. It
25:46
recommended Mac Studio to me and that a bunch of
25:48
other people pride. So a bunch of people tried to
25:50
use this tool to get Mac Pro. Nobody could do
25:52
it. And Julia Montier tried it and found out how
25:54
to cheat to win the game. If
25:57
you look at the source code, you can
25:59
see that a JSON file that
26:01
defines the options for
26:03
the endpoints. And that JSON, it's
26:05
not a JSON file, but it's not a JSON. That JSON does
26:08
not contain the Mac Pro. It
26:11
contains pretty much every other Mac that
26:13
Apple sells, but there
26:15
is no way to get to the Mac Pro because
26:17
the Mac Pro is not one of the options. That's
26:20
weird. Is it? No, this
26:22
is Apple telling you that literally nobody wants this
26:24
computer and nobody should have it. We all agree
26:26
on this show that the current Mac Pro is
26:28
not a great computer. But it is
26:31
a computer that exists. And on
26:33
top of that, there is at
26:35
least one very specific
26:37
reason why someone might want to use it. If
26:40
one of the questions had asked, hey, do
26:42
you have a bunch of PCI Express cards that you
26:44
need to use? If
26:46
the answer to that is yes, it's literally the only
26:48
computer Apple sells that you can do that on. And
26:51
that is really the only thing to recommend. Do
26:53
you think the people who made this quiz know
26:55
what a PCI Express card is? I
26:58
mean, it's Apple. They
27:01
have questions and answers for every other computer. It
27:03
just seems weird to me. Now, again,
27:05
I can understand saying, well, this is not a
27:08
great computer. And really, honestly, no one should really
27:10
buy it. I agree with all of that. But
27:12
when you make a help me choose tool on
27:14
your website, you should have
27:16
all of the things as endpoints. And yeah,
27:18
make the Mac Pro pretty much impossible to
27:20
get to unless you need it. But there
27:23
is a reason someone might need it. If
27:26
someone is going through this tool and saying, I
27:28
don't know what I'm going to do. I've got all these audio cards
27:30
that I need to use for, you
27:32
know, my old Mac is dying. Is there some
27:35
other computer that I can use? How would you
27:37
determine that Apple still
27:39
sells computers with card slots
27:41
in them? Everyone on Mass.
27:43
I'm saying, OK, well, the people who need the Mac Pro
27:45
know it, and so they don't need to use this tool.
27:48
That's not how these tools work. You could say the
27:50
same thing about, well, the people who need an iMac know they
27:52
want an all-in-one things that they don't need to use this tool.
27:56
And if you know which computer you need, yes, you don't need
27:58
this tool. But the tool exists to lead you. to
28:00
whichever product that Apple sells is best suited
28:02
for you. And it's weird to
28:04
leave just one out. And
28:08
I would just love to know if the thinking behind that process is
28:10
like... Look, if Apple doesn't want to sell them,
28:13
don't sell them, right? But they're selling them. You
28:15
can buy them for a huge amount of money.
28:18
And the tool can make
28:20
it difficult or almost impossible to get there because when
28:22
it says, how many PCI Express cards do you need
28:24
to use? The default choice should
28:26
be zero or I don't know what a PCI Express
28:29
card is. Like have a million options that regular people
28:31
will click and they will lead them off that path
28:33
and say you shouldn't buy this. But if the person
28:35
says three or any
28:37
number other than zero, you have to leave them to
28:39
the Mac Pro because it's literally the only computer they
28:41
sell with card slots. I mean, you're
28:43
gonna hate this, but so I did the whole quiz
28:45
trying to get to the Mac Pro before
28:48
you said it wasn't an option and just putting
28:50
in all like the highest requirements. Like, you know,
28:52
I need all that. I do 3D editing and
28:54
content creation and video editing and audio editing. I
28:56
need all these tools. I need to connect a
28:58
bunch of stuff to my Mac and
29:01
it recommended exactly what I'm
29:03
using right now. The Macbook Pro
29:05
16H. I
29:08
thought for sure I'd at least get a Mac Studio, but
29:10
nope. Well, no, because the question it asks is do you
29:13
do all your work in a single location or do you
29:15
need to be portable? Did you say all I do all
29:17
my work in a single location? I said like the on
29:20
the one desk option, the very top option where it's
29:22
like I do everything at the same place on a
29:24
desk. Like I even I thought for sure I'd at
29:26
least get a Mac Studio. I think a lot of
29:28
the endpoints recommend two computers. Like I didn't
29:30
just get the Mac Studio. I got recommended the Mac Studio and
29:32
the Macbook Pro. Oh, I also got two computers,
29:35
the Macbook Pro $4,000 configuration and
29:38
the Macbook Pro $3,500 configuration. I
29:42
don't know how you didn't end up with desktop because
29:44
there must have been some question that's differentiating portability.
29:46
Obviously, if you mention you ever need to take it
29:49
somewhere, they're not going to recommend it at all. Yeah,
29:51
I don't know. I don't know how great this tool
29:53
is. Wizards in general are not great.
29:55
I like their comparison ones like for the phones where it does
29:57
like columns and you can list all the features and scroll and
29:59
see they are different from each other. This
30:02
doesn't do that. But I do
30:04
think it's very strange to not
30:06
have a single one of your
30:08
computers. Remember when they were selling the trash can for
30:10
years and years, and really nobody should be buying that,
30:12
right? But if you needed
30:15
whatever GPUs it came with, for a while
30:17
it still did have the most powerful GPUs
30:19
you could buy in an Apple computer. And
30:22
if you needed those GPUs and they had a tool that was
30:24
asking you a bunch of questions, they should have had a question
30:26
that said, do you use Maya at
30:28
Pixar and need this much GPU power and then it
30:30
will lead you to the trash can? But I
30:33
don't know. It's weird. Anyway, if someone at Apple knows why
30:35
the Mac Pro is emitted from this tool, please
30:38
tell us. I'm sure it's the obvious reason, which is like,
30:40
no one should buy that. And we kind of agree, but you're selling
30:42
it, so put it in the tool. Pretty
30:44
sure it's very clear why it's emitted.
30:48
Even the very first day this Mac Pro came out, nobody
30:50
should be buying it, let alone now. It's
30:54
not nobody. It is the only computer with slots.
30:57
That's not a great reason for it to exist, and it's not
30:59
a reason for you to pay twice as much as Mac Studio,
31:02
but especially since they don't support, I believe
31:04
they don't support at all anymore, the PCI
31:07
Express breakout boxes like they used to on
31:09
the Intel things, it's literally
31:11
your only choice if you have cards. And
31:14
that's one of the reasons they should continue to make it. And do continue
31:16
to make it, and they just never ask about that. Oh,
31:19
yeah. It made me laugh quite
31:21
a bit that nobody was coming up with the Mac
31:23
Pro. I don't know. Maybe that's
31:25
a feature, not a bug. All
31:28
right. For the main, main
31:30
topic this week for your main course, we
31:33
have a plethora of different AI
31:35
related topics. And I'm going
31:38
to try to take us on a journey.
31:40
We'll probably fail, and that's OK. But basically,
31:42
this next section is AI. Huh.
31:45
That's a thing, isn't it? And
31:48
so we start on the 17th
31:50
of June, for what it's worth, with our
31:52
friend John Voorhees at Mac Stories, which
31:56
is them saying, hey, the article is entitled
31:59
How We're trying to protect Mac stories from AI
32:01
bots and web crawlers and how you can too. And
32:04
it seems like both John and
32:06
Federico are getting very wrapped around
32:08
the axle with regard
32:10
to AI stuff. And I'm not saying I don't
32:12
mean to imply that they're wrong or that's bad,
32:14
but they are getting ever
32:16
more perturbed on what's
32:19
going on with AI crawlers. And
32:21
I mean, to a degree, I get it. So,
32:23
uh, that was on the 17th of June. John
32:25
says, here's how you can protect yourself from crawling.
32:27
And then on the 21st of June, business insider
32:29
writes, it says, Oh, open
32:32
AI and anthropic seem to be ignoring
32:35
robots.txt. And if you're not familiar, if
32:38
you have a webpage or website, I
32:40
guess I should say, where, um, where
32:42
you control the entire domain, you can
32:44
put a file called robots.txt at the
32:46
root of the domain. So,
32:48
you know, it would be
32:50
marco.org/robots.txt and any self-respecting
32:52
and ethically clear crawler will start
32:55
crawling marco.org or whatever the case
32:57
may be by attempting
33:01
to load robots.txt and seeing if
33:03
there's anything there. And if so, it,
33:05
there's a, a mechanism, a schema, if you will,
33:07
by which, um, the
33:10
robots.txt will dictate who or really what
33:12
crawlers should or should not be allowed
33:15
to crawl that site. And it's by, it's by path.
33:17
They can say everything in this directory, you shouldn't crawl
33:19
everything here. You can crawl it. So you can sort
33:21
of subdivide your site to say which parts are accessible.
33:24
Yeah. And I have thoughts on that, but we'll come back to that. Yeah.
33:27
I mean, whenever you're ready to interrupt, to be honest,
33:30
feel free. Okay. Let's talk about robots.txt. Uh,
33:32
so, so, well, let me just actually very quickly, I apologize.
33:34
I gave you the, we gave you the green light and
33:36
I'm giving you the yellow light. Just very quickly, it's already
33:38
in the intersection. It's
33:40
important to note that robots.txt has never been
33:42
enforced in any meaningful
33:45
way. It's been kind of a, a, a, a, a, a, a, a, a, a, a, a, a, a, a, a, kind
33:51
of a friendly agreement amongst pretty
33:53
much the entire world wide web. But there's
33:56
never been any real, um, wood behind the
33:58
arrow or, or, or, or, or, whatever the
34:00
turn of phrase we call advisory. Yeah.
34:02
Advisory locking. It is a
34:04
scheme that people who agreed to
34:06
that scheme can use that scheme to collaborate
34:08
and work together, but there is no actual
34:10
mechanism stopping anyone from doing anything. It is
34:12
literally just a text file that you can
34:15
choose to read or not. Right.
34:17
So with that said, Marco carry on. Yeah. And,
34:19
and so robots.txt is
34:22
basically a courtesy. It is
34:24
a website saying, please
34:26
maybe follow these rules if you, if
34:29
you would, you know, but it
34:32
is not a legal contract. It is not a
34:34
legal restriction. Um, it
34:36
is not technically enforced or
34:39
enforceable really. It is
34:41
also not universally used and respected.
34:43
And so, and, and I can
34:45
tell you, I operate crawlers of
34:47
a sort and I don't use robots.txt. So
34:50
when, when overcast crawls podcast feeds, I
34:53
don't even check for robots.txt. I just
34:55
crawl the URL as the users have entered them
34:57
or as they have submitted them to iTunes slash
35:00
Apple podcasts. What robots.txt
35:02
advisories are were originally
35:04
for was not
35:06
like, Hey, search engines, don't
35:09
crawl my entire site. That's not what they were for.
35:12
What they were for was mostly
35:14
to prevent like runaway crawls on
35:17
parts of a site that were potentially infinitely
35:19
generatable. So things like if you had like
35:21
a web calendar and you can just click
35:24
that next month, next month, next month button
35:26
forever if you want to. And so a
35:28
web crawler that like, you know, indexes a
35:30
page and then follows every link on that
35:33
page. If it's hitting like a web calendar,
35:35
it can generate basically infinite links, uh,
35:37
as it goes forward or backwards in time.
35:40
So the main purpose
35:42
of robots.txt was to kind
35:44
of advise search engines. And
35:47
it was specifically for search
35:49
engines. It was to advise
35:51
them areas of the site that
35:53
crawlers should not crawl mostly for
35:55
technical reasons, occasionally for some kind of privacy
35:57
or restriction reasons, but usually it was just.
38:00
Was some kind of like legal contract that
38:02
said you must obey my rules That really
38:04
has never been tested until fairly recently like
38:06
that that was never really something that really
38:08
ever came up I mean there were been
38:10
a couple of things here and there was
38:12
like Google News and news publishers in certain
38:14
countries and stuff But like for the most
38:16
part that the basic idea
38:18
of robots.txt was really just
38:21
please Like that's it. It was like, please
38:24
do this or don't do this And
38:26
even then like it was often
38:28
used in ways that harmed The
38:32
actual customers using using things or or
38:34
did things that were unexpected this is
38:36
why I don't use it for overcast
38:38
feed crawlers because if you
38:40
publish an RSS feed and Submit
38:42
it to Apple podcasts. I'm
38:44
pretty sure you intend for that to be a public
38:46
feed And so I
38:48
feel like it is not really my place
38:51
to then you know Put up an alert
38:53
to my user to say hey this person's
38:55
robots.txt file actually says you know It's disallow
38:57
star on this one path that
38:59
this feed is in and so I actually
39:01
can't do this for you Like that that
39:03
would feel like I would have
39:05
I would have first of all no incentive to
39:07
do that and second of all because of It's it
39:11
because of its intention and context as a standard
39:13
for search engines, which I'm not This
39:15
doesn't really apply to me and my use and
39:18
and there were all sorts of things over the
39:20
years too Like you know that you could specify
39:22
certain user agents like alright Google bot do this
39:24
Yahoo bot Do this like and even and that
39:26
was also problematic over the years too because it
39:29
Disadvantaged certain companies if you just had
39:31
like bad behavior once or if a
39:33
site owner Just had like
39:36
one bad thought about one of these companies
39:38
once and then like never revisited it or
39:40
whatever Like then that company was allegedly like
39:43
disallowed from crawling this site. Why well, I
39:45
mean, it's not even that It's like, you
39:47
know for people in all of the technology
39:49
behind it You know
39:51
don't allow Google bot the way you identify Google
39:54
bot is by the user agent string Which is
39:56
part of the HTTP request and anybody can write
39:58
anything there and more
52:00
and more applications over time of
52:03
technologies like AI summarization and
52:05
action models and things
52:07
like that, where some
52:09
fancy bot basically is going to
52:12
be browsing and operating
52:14
a web page on behalf
52:16
of a user. That is kind
52:18
of like a browser, but it's
52:20
a very different form that I
52:22
think breaks all those assumptions with publishers.
52:25
This is one thing that I faced when I was
52:27
making Instapaper a thousand years ago. Instapaper would save the
52:29
text of a web page to read later and only
52:31
the text, not like all the ads and the images
52:33
and everything like that. I was very careful,
52:35
though, to not make features
52:38
that would enable somebody to
52:40
get the text of a page without having
52:42
first viewed the page in
52:45
a browser or a browser-like context. It would
52:47
load the whole page. They would see the
52:49
page. If there were ads, those
52:51
ads would load on the page. They would see
52:53
those ads. Then they could save what they were
52:55
seeing. Then part of that would be saved
52:57
in Instapaper and shown to them later. That
52:59
was always a very tense
53:03
balance to try to maintain
53:06
because what I didn't
53:08
want was widespread scraping of people's text
53:10
without loading their ads, but
53:13
I figured that seemed like an okay trade-off because that
53:15
was literally just saving what was already
53:17
sent to the browser and what the user was already
53:19
looking at. But a lot
53:21
of these new technologies, first
53:23
of all, I probably wouldn't attempt that today, but a lot
53:25
of these new technologies
53:29
break a lot of those little details.
53:31
If you have some kind of bot
53:33
that's doing something on
53:36
a website, suppose
53:38
it's one of these action models where you're saying, all right, book
53:41
me a flight. This stupid
53:43
book me a trip thing that all of
53:45
these AI demos from these big companies keep
53:47
trying to do even though nobody ever wants
53:49
that. Suppose you have a
53:51
book me a trip kind of thing with an AI model and
53:53
the idea is that model will go behind the scenes
53:56
and will go
53:59
operate experience. media or orbits behind the
54:01
scenes for you and manipulate things back
54:03
there to find the best flights and
54:05
hotels and whatever else. Well,
54:07
those sites make some of their
54:09
money via ads and affiliate things
54:11
and sponsor placements on those pages.
54:14
If you have some bot operating the site for
54:16
you, kind of clicking links for you behind the
54:18
scenes in some kind of AI context, that
54:21
bot is not going to see those ads. It's not
54:23
going to click those affiliate links. It's not going to
54:25
pick the sponsored listing. It's going to just
54:27
kind of get the raw data and that's it. And
54:29
that will be violating those sites business models that that
54:31
happened. That really has
54:33
not happened at massive scale until fairly
54:35
recently. So this really has not been
54:37
challenged. This really has not been legally
54:40
tested that much. This really has not
54:42
been worked out. Like what are the
54:44
standards? What are the laws? What are
54:46
the legal precedents? How much of this
54:48
is fair use versus not? You
54:50
know, for the most part until very recently,
54:52
we could pretty much just say, all
54:54
right, if you serve
54:56
something publicly via
54:58
public URLs and anybody can
55:00
just download it, then nothing
55:03
bad would really happen to you and your business
55:05
model for the most part. If
55:07
some bot came by sometimes and parse that page for
55:10
some other purpose, it wasn't a big deal. But
55:13
now there's a pretty
55:15
significant difference in scale
55:17
and type of replacement. Now
55:20
with a lot of these AI products
55:22
and with Google search itself, you know,
55:24
increasing over time and then more recently,
55:26
rapidly increasing, what we're seeing now is
55:28
full out replacement of the need for
55:30
the user to ever look at that
55:32
page. That's a
55:34
pretty big difference. And it's really
55:36
bad for web publishers and kind of,
55:38
you know, then consequently really bad for
55:41
the web in general. We
55:43
have a pretty serious set of challenges on
55:45
the web already, even before this new wave
55:47
of LLMs came by to
55:50
further destroy the web. We
55:53
already had a pretty bad situation for
55:55
web publishers for lots of other reasons
55:57
over the years to have
55:59
something that reminds us of the web.
56:01
moves the need for many people to
56:03
visit a page at all, that is
56:05
going to crush publishers. And so
56:07
it does make sense why everyone's freaking out about this.
56:09
It makes a lot of sense. I
56:12
do caution people though,
56:14
I don't think it's a
56:16
very good business move or a very good technology
56:18
move to say, I'm going to just block AI
56:21
from being able to do any, to see any
56:23
of my stuff. Because
56:25
that's a pretty big hammer and that's
56:27
a pretty big blanket statement. And
56:30
you can't actually block them anyway. Like that's when
56:32
it comes down to technically speaking, you can't, you
56:34
literally can't stop them, right? Unless
56:36
you stop everyone from viewing your website, in which case you
56:38
don't have a website. Right.
56:40
So I think it is wise to
56:43
focus on trying
56:45
to prevent uses of your content that
56:47
remove the need to visit your page.
56:50
Because that is a direct attack on your business model. That
56:52
makes a lot of sense. I don't
56:54
think it's wise to say, I don't
56:56
want any AI training or any AI
56:59
visibility of my page. That
57:01
I think is probably
57:03
short sighted and probably a bit too
57:06
much of a blanket statement. And that I
57:09
don't think it's good for any
57:11
party involved to
57:13
have that kind of blanket ban on it. I know a
57:15
lot of people want though that what people, the publishers
57:18
in particular want is
57:21
they want an ecosystem
57:24
of members who do agree to
57:26
some rules of politeness and
57:28
say, look, we should agree on a system that
57:30
lets me tell you that you shouldn't do
57:33
X, Y and Z on my site and you should agree to it
57:35
and we'll feel better about you if you do that. And part
57:38
of the reason I think Instapaper, your
57:40
example was not a particularly big
57:42
problem is like you said, scale. And anything
57:44
with AI in the name these days, people
57:46
flip out about it and think this is
57:48
going to be as big as Google. Instapaper was
57:50
not as big as Google. No,
57:53
it did not have billions and billions and billions
57:55
of users. If it did, if Instapaper
57:57
had Google scale, I bet there would have been a.
57:59
hell of a lot more scrutiny on even the very
58:01
conservative things that you did. But because it was small,
58:03
it's not a big deal. Like that's, that's part of
58:05
the sort of the ecosystem of the web is there's
58:08
all sorts of small things that don't have
58:10
particular big scale to do and all sorts
58:12
of weird stuff. Nobody cares about them. We
58:14
allow them to exist. It's fine. But now
58:16
these big names and AI, AI is the
58:18
next big thing. You're an AI company, you
58:20
have a lot of funding. Everyone looks at
58:22
them and think that could be the next
58:24
Google. That could be the next thing with
58:26
billions and billions of users. So we better
58:28
take whatever weird stuff they're doing way more
58:30
seriously than we would take overcast. And with
58:32
even with Google, the, you know, the current
58:35
giant in the world of search and they're, you
58:37
know, trying to replace sites and giving answers on
58:39
the side or whatever. Neil I Patel coined a term,
58:41
I think it was says about this called Google zero,
58:43
which is the point at which publisher
58:46
websites get zero traffic from Google search,
58:49
right? Because it's been going down and down over the
58:51
years because, hey, you'd write type Google search and look,
58:53
the answer to my question that I typed it's right
58:55
on the Google results page. It's unattributed. And I don't
58:57
have to, if it was attributed, I don't have to
58:59
click on any link to get to it because the
59:01
answer is right there. And so Google has been sending
59:03
less and less traffic to websites and Google zero is
59:05
when you notice, hey, you know what, you know how
59:07
much traffic we're getting from Google searches? Zero.
59:09
I don't know if it's absolutely zero for everybody, but
59:11
it's sure going down. And it's a scary
59:14
world to have
59:16
what was once the massively largest
59:18
source of your traffic to your website disappear.
59:23
But yeah, like whether or not it is wise
59:25
to exclude, to try to, to ask
59:28
to be excluded from pick, you
59:30
know, whatever, whatever AI crawler thing
59:32
from whatever open AI perplexity or whatever.
59:35
I think most publishers just simply want that
59:38
choice. And to have that choice, the,
59:40
the crawlers need to agree. Because again, there is no
59:42
technical way to stop this short of doing like putting
59:44
your entire site behind a paywall. And even that's not
59:46
going to stop them. Cause they'll just pay another crawler,
59:48
go through it. Like it's, that's, that's the thing about
59:50
publishing on the web. You do, it's
59:52
like DRM. You want people to
59:54
see your movie. You can't make it impossible to
59:56
see your movie. You have to give the viewer
59:58
an ability to see. your movie. But once you
1:00:00
give the viewer the ability to see your movie,
1:00:02
they can see your movie. Like, what
1:00:05
if they see it, but also record
1:00:07
it? I want them to see it,
1:00:09
but not be able to see it. Can I do that? And
1:00:11
the answer is no. Right? So if you're
1:00:13
publishing on the web, you have
1:00:15
like, it's like anything else. That's why Marco was right to
1:00:17
call this illegal thing. Like things are published
1:00:19
all the time. They were published in paper, you know, like
1:00:21
the books or whatever. It's like, but I can take the
1:00:24
book and look at it. I can see all the letters
1:00:26
in it. Haha. The book is mine. Well, no, actually we
1:00:28
have laws about the stuff
1:00:30
that's in that law, that book. We
1:00:32
have this thing called copyright. And even
1:00:34
though you can technically read it and
1:00:36
you can technically copy it increasingly more
1:00:39
easily over time with technology, we have laws surrounding it
1:00:41
to control what you can do it. And
1:00:43
ROASD text, people who think of ROASD text as some
1:00:45
kind of like technological bank vault. It's no more of
1:00:47
a bank vault than you could put on a book.
1:00:49
Like you do want people to read it and you
1:00:51
can't stop them from being able to copy it. And
1:00:54
these dates is really, really easy to copy a book, especially
1:00:56
if it's an ebook, right? Setting aside
1:00:58
the whole DRM thing. What
1:01:00
you want is some either in
1:01:03
a sort of polite society, an agreement
1:01:05
among the large parties that actually are
1:01:07
significant to get along and then
1:01:10
failing that you want laws to provide whatever protections
1:01:12
you think are due to you. And yeah,
1:01:14
the Google search stuff has, I feel like been
1:01:16
hashed out probably in the Altavius today's, but who
1:01:19
knows? And the AI stuff has not
1:01:21
yet been hashed out. And so to move on to this next
1:01:23
one, because we have a lot of these items, Microsoft,
1:01:25
at least someone in Microsoft has
1:01:27
a very interesting notion of
1:01:30
what the deal is on the
1:01:32
web and potentially what the law should be surrounding
1:01:34
it. So this is
1:01:36
a post on the verge by
1:01:38
Sean Hollister, who writes, Microsoft AI
1:01:41
boss Mustafa Suleiman incorrectly
1:01:43
believes that the moment you publish anything on
1:01:45
the open web, it becomes quote unquote freeware
1:01:47
that anyone can freely copy and use. When
1:01:49
CNBC's Andrew Ross Sorkin asked him whether AI
1:01:52
companies have effectively stolen the world's
1:01:54
IP, Mustafa said, I
1:01:57
think that with respect to content that's already
1:01:59
on the open web, the social contract of
1:02:01
that content since the nineties has been that
1:02:03
it is fair use. Anyone can copy it,
1:02:05
recreate it with recreate it, reproduce with, sorry,
1:02:07
recreate with it, reproduce with it. That
1:02:10
has been freeware if you like, and
1:02:12
that's been the understanding Microsoft
1:02:14
is currently the target of multiple lawsuits alleging
1:02:16
that it and open AI are stealing copyrighted
1:02:18
online stories to train generative AI models. So
1:02:20
it may not surprise you to hear Microsoft
1:02:23
exec defend it as perfectly legal. I
1:02:25
just didn't expect them to be so very publicly and
1:02:27
obviously wrong. And I'm not a lawyer writes Sean, and
1:02:29
that's also true for me, but I
1:02:31
can tell you that the moment you create a work,
1:02:34
it is automatically protected by copyright in the U S
1:02:36
you don't even need to apply for it. And you
1:02:38
certainly don't void your rights just by publishing it on
1:02:40
the web. In fact, it's so difficult to rate to
1:02:42
waive your rights that lawyers had to come up with
1:02:44
special web licenses to help. This
1:02:46
is so gross. Like I'm
1:02:49
not as riled up as a lot of
1:02:51
people about people about, you know, these AI
1:02:54
bots crawling my website, like sitting
1:02:56
here now, I don't find it that off
1:02:58
putting. I don't love it, but whatever this
1:03:01
though, this is disgusting. So this is such
1:03:03
a weird statement because everybody knows how copyright
1:03:05
works. I'm sure this person knows as well.
1:03:07
But to say that like, Oh, it's once
1:03:09
you put it on the web, it's freeware,
1:03:11
which is a term that mostly applies to
1:03:13
software. But like, the idea is you can
1:03:16
recreate it, reproduce it, uh, you know, copy
1:03:18
it. Like, no, no, no.
1:03:20
Like there, those are specifically
1:03:22
the things we actually do have laws around. Well, we
1:03:24
don't have laws around or the more complicated things like,
1:03:26
well, can I train AI on it or whatever? And
1:03:28
we'll get to that in a little bit. But like,
1:03:30
it's such a weird thing to say that like, Oh,
1:03:32
as everyone knows since the nineties, once you put it
1:03:35
on the web, you forfeit all ownership. That's
1:03:37
not true at all. And I think that's, that's, like,
1:03:39
it's one of the things that's great about the web
1:03:41
is, Oh, it's just like books. It's printed word, right?
1:03:43
And especially in the beginning, it was just a bunch
1:03:45
of words and we already have laws surrounding that, right?
1:03:47
And that's why there were cases about search engines. Like
1:03:50
our search engines, copying it because, you know,
1:03:52
we got this whole, you know, giant library
1:03:54
of laws about copying texts. My website has
1:03:56
text on it and Google's copying it and
1:03:59
they've had to do good. out and say,
1:04:01
actually, what Google's doing is fine within these
1:04:03
parameters, blah, blah, blah. But
1:04:06
that fight was fought because it was an
1:04:08
example of copying. But yeah, this... I
1:04:12
mean, obviously, the Microsoft AI
1:04:14
leadership, this guy is not a
1:04:16
lawyer either. But
1:04:19
that's not how you should defend this. You shouldn't
1:04:21
defend it by saying, you know, every single web
1:04:23
is a free for all. Because that's never
1:04:25
the way it's been and it's not the way it is now. There's
1:04:29
another foot in the mouth problem from Microsoft. I'm not
1:04:31
sure what's going on over there, but they really need
1:04:33
to take a lesson from
1:04:35
Apple and maybe try to speak with one
1:04:37
voice instead of having individual lieutenants make really
1:04:39
terrible statements of the press. Yeah.
1:04:42
So, Louis Mantia writes with regard to permissions on
1:04:44
AI training data from the 22nd of June. Louis
1:04:48
writes from John Gruber today on the 22nd of June.
1:04:52
It's fair for public data to be excluded
1:04:54
on an opt-out basis rather than included on
1:04:56
an opt-in one. And then Louis
1:04:58
continues, no, no, it's not. This
1:05:00
is a critical thing about ownership and copyright in
1:05:03
the world. We own what we
1:05:05
make the moment we make it. Publishing text or
1:05:07
images on the web does not make it fair
1:05:09
game to train AI on. The public in public
1:05:11
web means free to access. It does not mean
1:05:13
free to use. Also, whether reposting
1:05:16
my content elsewhere is in good faith or not,
1:05:18
it is now up to someone other than me
1:05:20
to declare whether or not to disallow AI training
1:05:22
web crawlers in the robots.txt file
1:05:25
to allow, excuse me, to add insult to injury,
1:05:27
that person may not have the knowledge or even
1:05:29
the power to do so if they're posting content
1:05:32
they don't own on a site that they also
1:05:34
don't own like social media. So this
1:05:36
is, he's so close to getting to the crux
1:05:38
of this. In the first
1:05:40
little paragraph here, he's basically declaring
1:05:43
that training AI on your data
1:05:45
is exactly the same
1:05:47
as copying and reproducing it. And that is
1:05:49
not something that the world agrees on. His
1:05:52
opinion is that it is. The courts have
1:05:55
not yet weighed in. I think
1:05:57
to the average person they would say, are those
1:05:59
the same things? seem like they might be a
1:06:01
little bit different. Kind of in the same way
1:06:03
that indexing your content in Google is a little
1:06:06
bit different than just literally copying it and reposting
1:06:08
it on the website, right? But
1:06:10
anyway, if you agree that it's the same as copying,
1:06:12
then yeah, sure. But then the second bit is getting
1:06:14
to even more of the heart of it here, which
1:06:16
is like, okay, so let's say we do agree that
1:06:19
it's the same much, you know, not
1:06:21
proven yet, but anyway. What
1:06:24
about when somebody like posts a link to
1:06:26
your site on the social media network, and
1:06:28
on that website, they do a little embedding,
1:06:31
inlining of like the first paragraph or whatever,
1:06:33
like what if someone copies and pastes a
1:06:35
paragraph of your thing on another
1:06:37
website, right? Even if you had absolute,
1:06:39
somehow magical technical control to stop
1:06:42
AI crawlers crawling your website, if
1:06:45
people can read your website and quote
1:06:47
from it or embed little portions of it or
1:06:50
screenshot or do whatever on other websites, of course,
1:06:52
you don't control those other websites. And so if
1:06:54
they allow crawling, your stuff's going to end up
1:06:57
in the Google search index in the
1:06:59
AI training model or whatever, even
1:07:01
though you disallowed it from your
1:07:03
website. And I would say
1:07:05
that for the most part that we also
1:07:08
have laws covering can someone take a portion
1:07:10
of the thing that you made and
1:07:12
quote it elsewhere, there's all legal framework
1:07:14
deciding whether that is fair use or
1:07:16
not, and it's complicated. And the
1:07:19
law is not a deterministic machine, as the other
1:07:21
Patel, who I mentioned before, is always fond of
1:07:23
saying, but we do have a legal framework to
1:07:25
determine, can I copy and paste this
1:07:27
paragraph from this thing on this person's site and quote it
1:07:29
on my site so I can comment on it? Yeah,
1:07:32
in general, you can. Can I make a
1:07:34
parody of this article on my
1:07:36
website? Yeah, you can. There's a whole bunch
1:07:38
of things around that that have been fought
1:07:40
out in court that we have a system
1:07:42
for dealing with. But all of
1:07:45
those things, say the court determines, you sue them
1:07:47
and they say, actually, this person was allowed to
1:07:49
quote that snippet, right? You lost your fair use
1:07:51
case because it's pretty open and shut. That's fine.
1:07:54
That just got indexed by an AI
1:07:56
training pot because that person's website allows
1:07:58
them, you know, polite AI bots or
1:08:01
whatever, nevermind again, nevermind that you can't stop them.
1:08:03
Anyway, right? That's
1:08:05
just the nature of publishing.
1:08:08
No matter what, you do not
1:08:11
have absolute control over every single character
1:08:13
that you made. You
1:08:15
do have control over the entire work and the
1:08:17
reproduction of the entire work, but you don't have
1:08:19
control over other examples of fair use. And
1:08:22
Louie's saying, oh, it shouldn't be like, I shouldn't have
1:08:24
to opt out. The default should be that nobody can
1:08:26
crawl me. I mean, that's just like,
1:08:30
not only is it technically impossible, but
1:08:32
like, that's not the
1:08:34
way the web has ever worked. It has
1:08:36
always been, we're going to crawl you unless
1:08:38
you tell us don't. And
1:08:41
even the polite ones, you know, they'll
1:08:43
read the thing that you said not to do it, but
1:08:45
by default, they're going to crawl you. And I think asking
1:08:47
for a world where everything you
1:08:49
publish on your website is not only
1:08:51
not crawlable by the things
1:08:53
you don't want it to crawl up, but also not able
1:08:55
to be quoted by other people is clawing
1:08:59
back rights that we've
1:09:01
already decided belong to other people through fair
1:09:03
use. So then the music
1:09:05
industry decided to get involved. Yeah. Multibillion
1:09:08
dollar companies are entered the chat, as they would say.
1:09:11
We talked about this before of like, Hey, Louie Mantian doesn't
1:09:13
want people to crawl on his website. What
1:09:15
can he do about it? He's just one person. Uh, the
1:09:18
music industry, they have a lot of money. They
1:09:20
have a lot of IP. This is
1:09:22
where the stuff really starts
1:09:25
going down. Yeah. So reading
1:09:27
from our Stendica on the 24th of June, universal
1:09:29
music groups, sunny music and Warner records
1:09:31
have sued AI music synthesis companies, Oudio
1:09:34
and Suno for allegedly committing mass copyright
1:09:36
infringement by using recordings owned by the
1:09:38
labels to train music generating AI models.
1:09:40
The lawsuits filed in federal courts in New
1:09:43
York and Massachusetts claim that the AI companies
1:09:45
use of copyrighted material to train their systems
1:09:47
could lead to AI generated music that directly
1:09:49
competes with and potentially devalues the work of
1:09:51
human artists. So from
1:09:54
the verge article, there's a quote from
1:09:56
RA a chief legal officer, Ken Doroshow.
1:09:58
And that quote is these. are straightforward cases
1:10:01
of copyright infringement involving unlicensed copying of
1:10:03
sound recordings on a massive scale. Suno
1:10:05
and Udeo are attempting to hide the
1:10:08
full scope of their infringement rather than
1:10:10
putting their services on a sound and
1:10:12
awful, lawful, excuse me, footing. And again,
1:10:15
that was the RIAA chief legal officer.
1:10:18
Mikey Shulman, the CEO of Suno, says
1:10:21
the company's technology is transformative
1:10:23
and designed to generate completely
1:10:25
new outputs, not to memorize
1:10:27
and regurgitate preexisting content. Shulman
1:10:30
says Suno doesn't allow user prompts
1:10:32
based on specific artists. Reading
1:10:35
from the lawsuit, the use here is far
1:10:37
from transformative as there is no functional purpose
1:10:39
for Suno's AI model to ingest copyrighted recordings
1:10:42
other than to spit out new, competing music
1:10:44
files. That Suno is copying
1:10:46
the copyrighted recordings for commercial purpose
1:10:48
and is deriving revenue directly proportional
1:10:50
to the number of music files
1:10:52
it generates further tilts the fair
1:10:55
use factor against it. Andy
1:10:58
Baio writes, 404 Media pulled together a video
1:11:00
montage of some of the AI-generated examples provided
1:11:02
in the two lawsuits that sound similar to
1:11:04
famous songs and their recording artists. Then
1:11:08
finally, we'll put a link in the show
1:11:10
notes to a Verge article that discusses what
1:11:12
the RIAA lawsuits mean for AI and copyright.
1:11:14
You know, I saw somebody say this a
1:11:16
few days ago, I don't remember who exactly
1:11:18
it was, but what's
1:11:20
going on if the RIAA are suddenly the good guys?
1:11:22
This is a weird place to be. Well, are they
1:11:25
though though? So here's the thing. Like this is the
1:11:27
tricky bit with this and we talked about this with
1:11:29
the image generators or whatever. So this is significant because
1:11:31
they're big rich companies and you have to take them
1:11:33
seriously when they bring a lawsuit because this is the
1:11:35
kind of like who can stop open AI and Google
1:11:38
and whatever. Well, you know, it's clash
1:11:40
of titans. You need other titans in here to be
1:11:42
duking it out, right? I
1:11:47
think this needs to
1:11:49
be fought out in a court in some way. I
1:11:51
say that before we see what the result will be
1:11:53
because maybe the result does not probably want
1:11:55
to happen. But like as
1:11:58
with the image things, These companies that you
1:12:00
type in a string and they produce a song for
1:12:03
you, right? These models
1:12:05
are trained on stuff. And
1:12:07
these record labels say, yeah, you trained them on
1:12:09
all our music, right? Gets back to
1:12:12
the question, is training something? Is
1:12:14
AI training? How does that relate to copying? Is
1:12:16
it just like copying? Is it not like copying
1:12:18
at all? Is it somewhere in the middle? Do
1:12:20
any of our existing laws apply to it? And
1:12:22
we've discussed this on past episodes as
1:12:24
well, especially when
1:12:27
the company doing the training, then
1:12:29
has a product that they make money on.
1:12:32
And as I said with the image training, these
1:12:35
models that make songs are worthless without
1:12:37
data to train them on. The
1:12:39
model is nothing without the training
1:12:41
data. This company that wants to make money,
1:12:43
you pay us x dollars, you can make
1:12:45
y songs, right? That's their business model. They
1:12:47
can make zero songs if they have not
1:12:49
trained their model on songs. So
1:12:51
the question is, where do those songs
1:12:54
come from? If they've licensed them
1:12:56
from somebody, if they made the songs themselves,
1:12:59
no problem, right? Again, Adobe training
1:13:01
their image generation models entirely on
1:13:03
content they either own or licensed.
1:13:06
Nobody's angry about that. That's the thing
1:13:09
you're doing. You own a bunch of images, you license
1:13:11
them from a stock photo company or whatever, you
1:13:14
train your models on them, you put the feature
1:13:16
into Photoshop, you charge people money for Photoshop, they
1:13:18
click a button, it generates an image. Whether people
1:13:20
like that feature or whatever, legality seems
1:13:22
fine. These other situations
1:13:25
where it's like, hey, we crawled your site because
1:13:27
we don't care about your robust ed text. We
1:13:29
trained our models on your data, on
1:13:31
your songs, on your whatever, right? And by the way,
1:13:33
we have no idea if these companies
1:13:35
actually paid for all the songs. Let's just
1:13:37
assume they did. They bottle the songs from,
1:13:39
you know, Sony Music
1:13:41
Warner Records or whatever, or they paid for a training
1:13:43
service. They got all the songs, they trained their model
1:13:45
on them, they're charging people to use their model, right?
1:13:49
Just like the image processing, I've always thought that if
1:13:53
you have a business that
1:13:55
would not be able to exist without content
1:13:57
from somebody that you did not. pay
1:14:00
anything for, that is very
1:14:02
different than are we trained in
1:14:05
AI model for research purposes, are we trained
1:14:07
it for, you know, for some purchase that
1:14:09
is not like literally making money off of
1:14:11
you. And this particular case is like, okay,
1:14:13
not just that they're making money, but the
1:14:15
thing they're providing is, quote, not transformative. They
1:14:17
keep using that word because that's one of
1:14:19
the tests for like fair use. Is the
1:14:21
work transformative? Have they taken the
1:14:23
thing that existed but made something new out
1:14:25
of it? And that's the argument in court,
1:14:27
whether it is not transformative.
1:14:29
And also, is it a substitute? There's
1:14:32
another one of the fair use tests. Is it
1:14:34
a substitute for the product? Is someone not going
1:14:36
to buy a Drake
1:14:38
album because fake Drake sounds
1:14:40
just as good and they just listened
1:14:42
to fake Drake, right? Is it a substitute
1:14:44
for it? It doesn't mean does it sound exactly like it? That's
1:14:46
a whole other sad area of
1:14:49
law of like, does song A sound too much
1:14:51
like song B and they have to pay them
1:14:53
whatever when they're all made by humans, right? This
1:14:55
is like, would someone pay for
1:14:57
this instead of paying for this?
1:14:59
Is one a substitute for the
1:15:02
other? And that's what they'll be
1:15:04
duking it out about. But I think
1:15:06
at its root, it is sort of like,
1:15:08
where does the value of this company come from?
1:15:12
Every company has to take input
1:15:14
from somewhere. They manufacture something and they sell it to
1:15:17
you or they have a service, they wrote the software
1:15:19
for it, they pay someone to run the servers and
1:15:21
they sell it. There's a there's sort of a value
1:15:23
chain there. And a lot of these companies are like,
1:15:27
we would make more money if we
1:15:29
don't have to pay for the things that make
1:15:31
our product valuable. So we don't want
1:15:33
to have to license all the music in the world, but
1:15:35
we do want to train an AI model on all the
1:15:37
music on the world so that we can make songs that
1:15:39
sound as good as all the music in the world. But
1:15:42
we don't want to have to pay for any of that. And
1:15:44
that seems to be not
1:15:47
a good idea for from my perspective. And this is one
1:15:49
of the different ways you can look at this. All
1:15:52
ethical, legal, I think one
1:15:54
of the frameworks that I've
1:15:56
fallen back on a lot is practical.
1:16:00
If, you know, for any given thing, say, if
1:16:02
we allowed this to happen, would
1:16:04
it produce a viable,
1:16:07
sustainable ecosystem? Like,
1:16:11
would it produce a market for
1:16:13
products and services? Would it
1:16:15
be a rising tide that lifts all boats? Or
1:16:17
would it like, burn the forest to the
1:16:19
ground and leave one tree left in the middle? Right? You
1:16:21
know what I mean? Like, that
1:16:23
practical approach, people like to jump on, like we talked
1:16:25
about before with the Vittici and Max stories and everything,
1:16:27
like they want to go to the moral and ethical
1:16:29
thing. They're stealing from us. It's our stuff. They have
1:16:31
no right. And even when I was saying before, like,
1:16:33
oh, they're, they don't want to pay for this stuff,
1:16:35
but they want to make money off of it or
1:16:37
whatever. But practically, and this is not the way the
1:16:39
law works, but this is the way I think about
1:16:41
it. Practically speaking, I'm always asking myself, if
1:16:44
this is allowed to fly, what
1:16:46
does this look like? Fast forward this. What,
1:16:48
you know, is this viable? Right? What
1:16:51
if, if everyone's listening to fake Drake,
1:16:54
does Drake, the next Drake, are not able to
1:16:56
make any money? Does human
1:16:59
beings making music become an unviable business? And
1:17:01
all this is just an increasingly gray soup
1:17:03
of AI generated stuff that loops in on
1:17:05
itself over and over again. Right? Like
1:17:07
where are the, you know, and we have the same
1:17:10
thing with publishing on the web. Like, does Google destroy
1:17:12
the entire web because no one needs to go to
1:17:14
websites anymore? They just go to Google. Right. Unfortunately,
1:17:18
when these cases go, you know,
1:17:20
to court, no one is thinking that that's not how
1:17:22
the law works. The law is going to be, is
1:17:24
this fair use, whatever? Just does Congress pass new laws
1:17:26
related to this or whatever. But what I really hope
1:17:28
is that the outcome of all these things and the
1:17:30
thing I'm always rooting for is can we get to
1:17:33
a point where we have an
1:17:35
ecosystem that is sustainable, which
1:17:37
means it's probably, you know, whatever they're suing for is like, I think they
1:17:40
want like $150,000 for every song or something. That is not a sustainable solution.
1:17:45
You can't train an AI model when you pay $150,000 for each song that you
1:17:47
trained it on because you need basically
1:17:49
all the songs in the world. That's a
1:17:51
big number. That's stupid. We do
1:17:53
want AIs that can make little songs, right?
1:17:55
I think that is a useful thing to have, right?
1:17:58
So we need to find a way where we can... have that, but
1:18:01
also still have music
1:18:04
artists who can make money making actual music, setting aside
1:18:06
the fact that the labels take all the money and
1:18:08
the artists get rarely anything anyway, which is separate
1:18:10
issue. Right. And there was a good article
1:18:12
about that recently about how the labels, the
1:18:15
label Spotify and the artists and the terrible relationship
1:18:17
there that screws over artists. Anyway,
1:18:22
I think I really hope that the outcome
1:18:24
of this is some kind of situation where,
1:18:26
where there's
1:18:30
something sustainable. There's like, I keep using ecosystem, but
1:18:32
it's like, you know, you have to have enough
1:18:34
water, the whole water cycle. This animal
1:18:36
eats that animal. It dies. It
1:18:39
fertilizes the plant. Like the whole, you know, a sustainable
1:18:41
ecosystem where everything works and it goes all around in
1:18:43
a circle and everything is healthy. And there's growth,
1:18:45
but not too much and not too cancerous. And it's
1:18:48
not like everything is replaced by a model culture and
1:18:50
only one company is left standing and all that good
1:18:52
stuff. Right. But right now
1:18:54
the technology is advancing in a way that if
1:18:57
it's not, if we don't do
1:18:59
something about it, uh, the,
1:19:01
the individual parties involved are not
1:19:04
motivated to make a sustainable ecosystem.
1:19:06
Let's say, I mean, that's kind of what the DMA is about
1:19:09
in the EU and these AI companies definitely
1:19:11
are not motivated to try to make sure they have
1:19:13
a sustainable ecosystem. They just want to make money. And
1:19:16
if they can do it by taking the world's music and
1:19:18
selling the ability for you to make songs that sound like it
1:19:20
without paying anything to the music that they ingested, they're going to
1:19:22
try to do that. Yeah.
1:19:27
It's, I don't know. It's all just so weird
1:19:29
and gross. And it's, it's hard because I don't
1:19:32
want to be old man who shakes fist at clouds.
1:19:35
Right. And it
1:19:37
seems like AI for
1:19:39
all the good and bad associated with it
1:19:41
is a thing. It's certainly a flash in
1:19:43
the pan for right now, but I
1:19:46
get the feeling that where blockchain
1:19:50
and Bitcoin and all that sort of
1:19:52
stuff was very
1:19:55
trendy, but anyone with a couple of brain
1:19:57
cells to rub together would say, eh, that's
1:19:59
all good. a fade or it's certainly not going to
1:20:01
work the way it is today. I
1:20:03
think there's a little of that here, but I
1:20:05
get the feeling that this is going to stick
1:20:08
around for a lot longer. And I think
1:20:11
that there needs to be some wrangling
1:20:13
done, some legal wrangling. And I
1:20:16
get the move fast and break things mentality of
1:20:18
these startups that are doing all this, but I
1:20:21
don't know. It just feels kind
1:20:23
of wrong. Like again, I'm not nearly
1:20:26
as bothered by it as some of our peers
1:20:28
are, but it just doesn't feel right. And
1:20:31
it definitely doesn't feel sustainable. Like practically, like if,
1:20:33
if, like regardless of how you feel about right
1:20:35
or wrong, if you say, if we just let
1:20:38
them do this, like, and
1:20:41
these, you know, these models get better and better and
1:20:43
produce more and more acceptable content, you can
1:20:45
see that it's taking, again, regardless of
1:20:47
how this lawsuit ends up with the whole
1:20:49
record labels, you can see that it is
1:20:51
taking value away from human beings, making
1:20:54
music and pushing that value to models,
1:20:56
making music. But those models are absolutely worthless
1:20:58
without that human generated music, at least initially,
1:21:00
right? Again, maybe in the future, there will
1:21:02
be models trained entirely on model generated music,
1:21:04
but then you have to trace it back
1:21:06
to where that model get trained. Like in
1:21:08
the end, these models are trained
1:21:11
on human created stuff and
1:21:13
there's, there may not be
1:21:15
enough officially licensed human created stuff to train them
1:21:17
on at this point. I
1:21:19
think that's, you know, I think we, we
1:21:22
want these tools. They are useful for
1:21:24
doing things. Even if you think, oh, they make
1:21:26
terrible music. Sometimes people need terrible
1:21:28
music, right? Sometimes people just need a little jingle. They
1:21:30
can describe it. They want it to be spit out.
1:21:32
Right. By most people's definitions, all of my music is
1:21:34
terrible music. They
1:21:37
do useful things like, unlike, you know, cryptocurrency,
1:21:39
which does a very, very small number of
1:21:42
useful things that is not in general purpose.
1:21:44
The AI models do tons of useful things.
1:21:46
Apple's building a bunch into their operating systems.
1:21:48
You know, people use them all the time.
1:21:50
They do tons of useful things, right? We
1:21:53
should find a way for them to
1:21:55
do those things without
1:21:58
destroying the ecosystem. I think
1:22:00
we can find a way for that
1:22:02
to happen. If you look at the
1:22:04
awful situation with Spotify and record labels
1:22:06
and music artists, that's a pretty bad
1:22:08
version of this. And yet still,
1:22:10
it is better than Spotify saying, we're going to stream
1:22:13
all these songs for free and not
1:22:15
pay anybody. I wish I could find that article for
1:22:17
the notes. I'll try to look it up. But
1:22:19
even that is better than the current situation
1:22:22
with AI, which is like, we're
1:22:24
just going to take it all for free. Come sue us. And
1:22:27
they say, okay, we are suing you. And they'll
1:22:29
battle it out in court. And like, either way
1:22:31
this decision goes with the music, they could go
1:22:33
bad in both directions. Because if they say, oh,
1:22:35
you're totally copying this music, all AI training is
1:22:38
illegal. That's terrible. That's
1:22:40
bad. We don't want that, right? And
1:22:42
if they say, no, it's fine. It's transformative. You can
1:22:44
take anything you want for free. That's also bad. So
1:22:47
both extremes of the potential decision that a
1:22:49
court could make based on this lawsuit are
1:22:51
really bad for all of us for the
1:22:53
future. So that's why I hope we find
1:22:55
some kind of middle ground. Like again, with
1:22:57
Spotify, they came up with a
1:22:59
licensing scheme where they can say, we want to
1:23:01
stream your entire catalog of music. Can
1:23:04
we figure out a way to exchange money
1:23:06
where you will allow that to happen legally?
1:23:09
And they came up with something. It's not a great
1:23:11
system. They came up with, again, if I can
1:23:14
find that article, you can read and see how bad it is. But they
1:23:16
didn't just take it all for free. Right. And
1:23:18
they also didn't, the music labels didn't say, okay, but every time
1:23:20
someone streams one of these songs, it's 150 grand. Right.
1:23:23
That's also not sustainable. So obviously they're
1:23:26
staking out positions in these lawsuits and they're trying to
1:23:28
put these companies out of business with big fees or
1:23:30
whatever. But yeah, it's
1:23:32
like, it's scary. It's scary when Titans clash.
1:23:35
And I do worry about how the results of these cases
1:23:37
are going to be. But I think, I think
1:23:40
we have to have these cases or, and I
1:23:42
know this is ridiculous in our country, or we
1:23:44
have to make new laws to address this specific
1:23:47
case, which is different enough from
1:23:49
all the things that have come before it, that we
1:23:51
should have new laws to address it. It
1:23:53
would be better if those laws weren't created by
1:23:55
court decisions, but our ability
1:23:57
and track record for creating technology.
1:24:00
technology related laws for new technology
1:24:02
is not great in this country.
1:24:04
So there's that. Yeah.
1:24:06
And then it continues because
1:24:08
Figma, a popular, um, I
1:24:11
don't know how to describe this, like a user interface,
1:24:14
uh, generation tool, um, yeah,
1:24:17
design tool. Thank you. Uh, they pulled
1:24:19
their AI tool after criticism that it blatantly
1:24:21
ripped off Apple, uh, Apple's weather app. So
1:24:23
this is the verge by Jay Peters. Figma's
1:24:26
new tool make designs. Let's use our users
1:24:28
quickly mock up apps using generative AI. Now
1:24:30
it's been pulled after the tool drafted designs
1:24:32
that looked strikingly similar to Apple's iOS weather
1:24:35
app. In a Tuesday interview with
1:24:37
Figma, CTO, Chris Rasmussen, I asked him point
1:24:39
blank if make designs was trained on Apple's
1:24:41
app designs. His response, he couldn't say for
1:24:43
sure. It
1:24:46
was not responsible for the training AI models that
1:24:48
used it all. Who knows who trained it? It's
1:24:50
just our model. We don't know. Do you know
1:24:52
who trained? I don't know. Does anyone know who
1:24:54
trained it? It's just, we just found it on
1:24:56
our doorstep and this is a model with the
1:24:58
real trainer. Please stand up. Uh,
1:25:01
quote, we did no training as part of the
1:25:03
generative AI features. Rasmussen said the features are quotes
1:25:05
powered by off the shelf models and a bespoke
1:25:07
design system that we commissioned, which appears to be
1:25:09
the underlying issue. So if you commissioned it, then
1:25:11
you should know we had someone else do it
1:25:13
and they gave it to us and we just
1:25:15
took it and we're like, we didn't ask him
1:25:17
any questions. It's fine. Whatever, whatever
1:25:19
you got, just give it as it's probably
1:25:22
fine. Uh, the key AI models that power
1:25:24
make designs are open AI's GPT four O
1:25:26
and Amazon's Titan image image generator G1. According
1:25:28
to Rasmussen, if it's true that Figma didn't
1:25:30
train its AI tools, but they're spitting out
1:25:32
Apple app lookalikes anyway, that could suggest that
1:25:34
open AI or Amazon's models were trained on
1:25:37
Apple's designs. Open AI and Amazon
1:25:39
didn't immediately reply to a request request for
1:25:41
comment. This is seriously the like the
1:25:44
Spiderman pointing at other Spiderman's
1:25:47
image. It's just, it's not my fault. It's their fault.
1:25:49
Well, it's not my fault. It's their fault. Oh, no,
1:25:52
no, no, no, no. It's not my fault. It's their
1:25:54
fault. It was that company. I think it was open
1:25:56
AI or whatever the Sora model that does it makes,
1:25:58
makes movies essentially. Someone who was
1:26:00
responsible for that, was asked in an interview,
1:26:02
was your model trained on YouTube? They
1:26:05
didn't give an answer. Like, maybe, I don't
1:26:08
know. Listen, if you run an AI company,
1:26:11
figure out how and where your
1:26:13
models were trained. I don't know, maybe
1:26:15
you trained them on good things, bad things, whatever,
1:26:18
but have an answer. Don't say, we don't know.
1:26:21
Someone else did it. We like, this
1:26:23
seems like table stakes. You
1:26:25
should know where and on what your
1:26:27
model was trained on. Not like granular, like every single individual
1:26:30
thing, although ideally that would be great, but there's too much,
1:26:32
I get it, right? But when someone says, hey, did you
1:26:34
train on YouTube? You should be able to answer that with
1:26:36
a yes or no. Right, not
1:26:38
weasel about it. And this one, was this trained on Apple's
1:26:40
apps? I mean, anyone looking at it
1:26:43
is gonna be like, well, if it wasn't, this is the
1:26:45
world's biggest coincidence because it looks just like Apple's app. As
1:26:48
Gruber pointed out, right down to the really weird
1:26:50
like line chart that I never really understood until
1:26:52
I saw it explained in Apple's weather app, right?
1:26:54
Was obviously trained on Apple stuff, but
1:26:57
you have to have an answer, right? If you don't
1:26:59
have an answer, say, I don't know, but I'll find
1:27:01
out for you and then come back. But
1:27:03
like the bar is real
1:27:05
low here. Anyway, same
1:27:08
situation, different thing. Images, songs,
1:27:10
text, UIs, a
1:27:15
mock-up tool that makes UIs, it's
1:27:17
based on a model. That model
1:27:19
is worthless without being trained on
1:27:21
a bunch of UIs. Were you gonna get enough UIs to
1:27:23
train it from the world of
1:27:26
UIs that we take essentially without permission? Is
1:27:28
that okay? If we sell that as part
1:27:30
of our application, is that okay? I
1:27:33
mean, I wrote a
1:27:35
big post about this, what, in January? Excuse
1:27:37
me, I made this. Yeah, and
1:27:39
we talked about it on the podcast before
1:27:41
and I took a while to write
1:27:43
this because actually speaking of Neil Patel, I was listening
1:27:46
to the Decoder podcast and there was an episode where
1:27:48
I was debating with somebody about the New York Times
1:27:50
lawsuit at the time, like New York Times was suing
1:27:52
some company that had trained its AI on the New
1:27:54
York Times and they said, you
1:27:56
can't do that. And going back and forth about like, well, the
1:27:58
model is just doing what a person would do. it's learning and
1:28:00
blah, blah, blah. Is the person the same
1:28:03
as the model? Does the model have the same
1:28:05
rights as a person? And I was trying to
1:28:07
write up something related to that. And as usual,
1:28:09
writing helps me clarify my thinking, but it is
1:28:11
a fairly complicated circuitous route to sort of really
1:28:13
dig down into that thought to
1:28:16
get to what's at the heart of it. And
1:28:19
I wrote this thing and I think I did get to the heart
1:28:21
of it as far as I was concerned, but it's
1:28:24
complicated. So every time I try to like summarize it
1:28:26
on the podcast, I find myself like tongue tied and
1:28:28
rather, you know, you just quote from the paragraphs. Like
1:28:30
I think if you read the post, my thoughts are
1:28:32
in there, but a lot of people have read it
1:28:34
and like no one has commented on it. So maybe
1:28:37
I'm doing a poor job communicating it, but I
1:28:40
was coming at it from the other angle. We talked all about
1:28:42
training data in this section of the show here. I was coming
1:28:44
at it from the angle of like what
1:28:47
was then one of the hot
1:28:49
topics, which is say I use one of these
1:28:51
tools. Say I use the Figma tool to generate
1:28:53
a UI. I use the song tool to generate
1:28:55
a song or whatever. That
1:29:00
thing that I made, what
1:29:02
is the legal, ethical,
1:29:04
moral, practical ownership
1:29:07
deal with that? If
1:29:09
I use Figma to make that
1:29:11
auto create UI thing and it makes me
1:29:14
a UI and I put
1:29:16
that in my app, do I
1:29:18
own that UI? If I make a song with
1:29:20
the song making tools, do I have the copyright
1:29:23
on that song? There's been legal cases about this. And I think
1:29:25
the only ruling we have now is something like if you make
1:29:27
it with any AI generator tool, you don't have the copyright on
1:29:29
it or whatever. But
1:29:31
the reason I got to that, because I was getting with the whole
1:29:33
like, oh, training is just like what
1:29:35
a human would do. They read all these articles in
1:29:38
the New York Times, then you ask the human the
1:29:40
answer, and they read all those articles, and they have
1:29:42
the knowledge from reading those articles, and they give you
1:29:44
an answer. Well, that's just what our AIs are doing.
1:29:46
I'm like, yeah, but a human is a human, and
1:29:48
AI is an AI, and is that really what the
1:29:50
root of the thing it is? And I kept chasing
1:29:52
that down, chasing that thought down, and got to sort
1:29:54
of the thing that confers ownership. When
1:29:58
you make something, it's yours. You write something. something on
1:30:00
your blog, you have the copyright to it because you
1:30:02
created it. It's so clear, right? What
1:30:05
if you draw a picture on a piece of paper? Okay, you
1:30:07
got the copyright on the picture, right? What
1:30:09
if you use Photoshop to make a
1:30:11
picture? Well, now you use this software tool written
1:30:13
by a bunch of other people, just plain old
1:30:15
Photoshop, not like AI generally, like Photoshop 3.0, right?
1:30:18
With layers now. You
1:30:20
use Photoshop, but you didn't write Photoshop. A bunch
1:30:22
of people wrote software to make Photoshop, and you
1:30:24
then paid Adobe for, then they gave you that
1:30:26
software product. You use Photoshop to make a picture,
1:30:29
but still we say, well, you made that picture.
1:30:31
You have the copyright on it. You are the
1:30:33
creator. You own it, right? Then
1:30:36
we say, all right, but what if you can't draw?
1:30:40
What if you tell somebody, like, I can't draw. Here's
1:30:42
what I want. I want this picture, or whatever example I gave a
1:30:45
thing, a polar bear riding a
1:30:47
skateboard, but I can't draw. So I
1:30:49
asked somebody else, say, can you draw me a picture of
1:30:51
a polar bear riding a skateboard? So someone goes and they
1:30:53
draw a picture of a polar bear riding a skateboard. At
1:30:55
that point, the person who drew it owns it. Maybe they
1:30:57
use Photoshop, maybe they don't. They own it because they created
1:30:59
it. They drew it, right? But then you say, okay, this
1:31:01
was a work for hire. I'll give you 10 bucks. And
1:31:03
our contract says, I give you 10 bucks. You give me
1:31:06
the polar bear drawing. Now I own the polar bear drawing
1:31:08
because I paid you for it. That is
1:31:10
a market for creative works. Someone
1:31:12
was an artist. I can't draw. They
1:31:15
could. They drew it. They asked
1:31:17
for money. I gave them money. The
1:31:20
copyright is now mine, right? And
1:31:22
the act of creation is clear. The person who drew
1:31:24
it, they created it. I paid money for it. They
1:31:27
sold me their creation. Now I own it. All, you
1:31:29
know, normal, right? Now
1:31:31
I say, make me a picture of a polar bear
1:31:33
on a skateboard. But I don't say it to
1:31:35
an artist. I say it to an
1:31:37
image generator. It's the exact
1:31:39
same thing as I did before. Before
1:31:42
when I did it, it was clear that I don't own
1:31:44
anything until I pay for that, right? Now
1:31:46
when I do that exact thing, but instead of typing
1:31:48
it into an email to an artist, I type it
1:31:50
into an image generator and
1:31:52
I get an image back. Who
1:31:55
created that image? I didn't
1:31:57
create it. But if you're going to say I didn't create the
1:31:59
one that the artist drew for me. Because you just told the
1:32:01
artist what to draw, but you didn't create it. Well, if I
1:32:03
didn't create that one, I certainly didn't create this one because I
1:32:05
literally did the same thing. I just typed the text in a
1:32:07
different text field. It could literally be the same text. It could
1:32:09
be an AI prompt, emailed to an
1:32:12
artist or sent to an AI. So I'm not
1:32:14
gonna say that I am the creator of that.
1:32:17
The AI model can't be
1:32:19
the creator because computer
1:32:22
programs can't own things. They
1:32:24
don't have rights. Computer
1:32:28
programs are made by people who have rights, just like
1:32:30
people who wrote Photoshop. They have the rights to Photoshop
1:32:32
and so on and so forth. But the people who
1:32:34
wrote Photoshop have no rights to the things that people
1:32:36
made with Photoshop, despite Adobe's little snafu
1:32:38
with their license agreements recently, which they
1:32:40
clarified. But anyway. So
1:32:43
I didn't make that picture of the polar bear.
1:32:45
The large language model didn't make it. Who
1:32:50
owns that picture of the polar bear based on
1:32:52
the act of creation? Where is the act of
1:32:54
creation there? How did that model
1:32:56
create the polar bear? Well, it created the polar
1:32:59
bear picture because it had been trained on tons
1:33:01
of other images that maybe were
1:33:03
or weren't licensed. But still I'm looking around of
1:33:06
like if ownership is conferred by the act of
1:33:08
creation and there's no act of creation here, what
1:33:10
the hell are we, what's going on here? Who
1:33:12
owns the picture of the polar bear? Every
1:33:16
time I dig down into some kind of like,
1:33:18
oh AI's allowed you to do this and you're
1:33:20
allowed to train, it's just what people do or
1:33:22
whatever and computers aren't people. I always go through
1:33:24
to looking for how we confer ownership of stuff
1:33:26
like this, how we confer ownership of intellectual
1:33:29
property, how we exchange money for intellectual
1:33:31
property, how the market for intellectual property
1:33:33
works. And none of the
1:33:35
existence systems make any sense in a world where
1:33:38
I can say the same thing to a human and
1:33:41
a generator that is
1:33:43
clearly not me creating anything and yet I do
1:33:45
get a picture out of it that came from
1:33:47
somewhere and there's no like, there's no human act
1:33:49
of creation. It's an indirection, right? And
1:33:51
so I think we need new ways to think
1:33:53
about a new laws for that type of indirection
1:33:56
to say, what is the chain
1:33:58
of ownership here? It's kind of like, not
1:34:00
quite the same thing, but remember the whole
1:34:02
thing where the monkey took
1:34:04
a picture of itself with the camera? Do you remember? Oh, yeah.
1:34:06
It was like a camera set up in the jungle or whatever
1:34:08
and a monkey comes up to it and snaps a picture of
1:34:10
himself and the photographer is like, well, it's
1:34:12
my camera, so I own the copyright to the picture.
1:34:14
I'm like, well, doesn't the monkey own the copyright? Because
1:34:17
it took the picture, right? And it's like, but the
1:34:19
monkey can't own the copyright. It's not a person, right?
1:34:21
And believe me, a monkey is way closer to a
1:34:23
sentient being than an LLM, right? It's
1:34:26
a real living thing. No one's going to argue
1:34:28
in court that a monkey is not alive. And
1:34:30
they're going to say, well, does it have legal
1:34:32
rights? Well, I would say a monkey has more
1:34:34
legal rights than a large language model, which is
1:34:36
just a bunch of numbers in memory. And
1:34:39
so this is the kind of conversation we're having. And
1:34:42
honestly, this would be so much easier to
1:34:44
have if we had actual artificial intelligence, as
1:34:46
in sentient artificial beings. But we don't. That's
1:34:49
just science fiction. Large language models are not anywhere close
1:34:51
to that. That would be so much easier because you'd
1:34:53
be like, well, conscious beings have rights and we need
1:34:55
the, you know, whatever they always have names of this
1:34:57
in sci-fi movies, the AI consciousness act
1:35:00
of 2732 that gives rights to the AIs to avert
1:35:02
a global war and plunge
1:35:07
us into the matrix apocalypse. You know what I mean? Like,
1:35:10
it's so much easier when you say, well, people
1:35:12
have rights and computer programs that are basically people
1:35:14
have rights and it's straightforward, but
1:35:16
we're nowhere near there. So now we're arguing about monkeys,
1:35:19
if they have the copyright pictures,
1:35:21
and we're arguing about huge matrices
1:35:23
of numbers, whether they can
1:35:25
create anything or you're saying like, oh, you're saying
1:35:28
basically like the people who wrote Photoshop own every
1:35:30
picture that's made from because like, well, no, the
1:35:32
LM doesn't own it. And the person who wrote
1:35:34
the prompt doesn't own it. But you know who
1:35:36
does own it? Open AI because they wrote the
1:35:38
program that crawled all the pictures in the world
1:35:40
that train the model that you paid to use.
1:35:44
None of those answers are satisfactory
1:35:46
anyway. Like it doesn't feel right.
1:35:49
It doesn't seem right. It doesn't seem sustainable. And
1:35:51
yet we do need some kind of answer here,
1:35:54
even if the answer here is that anything again,
1:35:56
like that one wall precedent we had is like,
1:35:58
if you make something out of AI, you don't
1:36:00
own the copyright. It is not copyrightable. Nobody owns
1:36:03
it. It's garbage. It's slop. It's a thing that
1:36:05
exists, but nobody can claim that they own it
1:36:07
So it is free for anybody to take and
1:36:09
do whatever they want with but you
1:36:11
certainly can't like sell it to someone because you didn't
1:36:13
Own it. It's very confusing.
1:36:16
I know that I haven't made this any clearer You can try
1:36:18
reading my post to see if it becomes any more clear but
1:36:20
really this is this is a dizzying
1:36:23
topic if you think about it for any amount of
1:36:25
time and I think a lot of
1:36:27
people are doing a
1:36:29
lot of feeling about it, which makes perfect sense
1:36:31
and Honestly, it is more
1:36:33
straightforward to feel things about it than it is to
1:36:35
think about it because thinking about it gets you Into
1:36:37
some weird corners real fast It's
1:36:40
just it's a mess in It's
1:36:44
a mess and I don't know what the right
1:36:46
answer is right like it's it's so gray from
1:36:48
top to bottom And I just I don't know
1:36:51
I just don't know well And I think we're
1:36:53
gonna have to be fighting this and working this
1:36:55
out for a while I mean
1:36:58
look at how much disruption
1:37:00
to existing businesses existing
1:37:02
copyright law and existing norms was
1:37:04
caused by the web and then the
1:37:06
rise of other things on the
1:37:08
internet like this is Just this is
1:37:10
how technology Goes
1:37:13
there are massive disruptions to
1:37:16
what has been established what we what many
1:37:18
people have held dearly there
1:37:20
there's massive disruptions to that when new tech comes
1:37:22
around sometimes and Sometimes
1:37:24
it takes a decade or two to really
1:37:27
settle out and work out. What are the
1:37:29
norms? What should the laws be what is
1:37:31
copyright mean in this new world things like
1:37:33
that? like that takes a long time to
1:37:35
work out sometimes the rise
1:37:38
of these AI techniques and models is
1:37:41
Potentially as disruptive to
1:37:44
existing business models
1:37:46
and norms and perspectives as
1:37:48
the web was when it first came out a
1:37:50
thousand years ago, so I really think we're in
1:37:52
for a while of just
1:37:55
not knowing there's gonna be a
1:37:57
lot of damage and destruction
1:37:59
along the path to get from
1:38:02
where we are now to where kind of where
1:38:04
things settle out. It will destroy
1:38:06
a lot of businesses and it
1:38:08
will, you know, make it hard for a lot
1:38:11
of people to do what they've been doing. It
1:38:13
will also create a bunch of new businesses and
1:38:15
create a bunch of new value and new opportunities,
1:38:17
just like any other massive disruption. I think this
1:38:19
is, this is a very large disruption and it
1:38:22
is, it's mostly only going
1:38:24
to start to
1:38:26
become visible of like, you know, what the other side
1:38:28
looks like just after a bunch of time has passed
1:38:30
and we've gone through a bunch of messiness and we're
1:38:32
just, we're in such early days, it's really hard to
1:38:34
know where we're going to end up right now. I
1:38:37
feel like this is going to be in some
1:38:39
respects, not all, but in some respects, even more
1:38:41
disruptive than the initial web, because the initial web
1:38:43
was kind of like text.
1:38:45
We have laws governing that. It was
1:38:48
a massive shift of wealth. Obviously newspapers
1:38:50
go out of business. Craigslist gets
1:38:52
rich. You know what I mean? Like, like, but
1:38:54
we saw that giant shift in paper magazines, like
1:38:56
the shift of publishing, right, and web search and
1:38:58
doing all that or whatever. But during that entire
1:39:00
thing, people were upset and it was
1:39:02
a big turmoil because it was like, these things used
1:39:04
to be huge. Every city had 25 newspapers. A newspaper
1:39:07
reporter was a big job. And, you know, it was
1:39:09
like, and all of a sudden all that money's going
1:39:11
elsewhere to these.com things or whatever. But during that whole
1:39:13
process, there was mostly agreement
1:39:15
that like, newspapers own what
1:39:17
they publish, websites own what they publish. Like
1:39:19
we have existing copyright laws for this. There's the
1:39:21
whole Google search index thing that we can figure
1:39:24
out and, you know, fair use on the
1:39:26
internet and stuff. But in general, it was just
1:39:28
a massive shift of power and money from
1:39:30
older industries to newer ones, mostly
1:39:32
following along the shape of laws
1:39:35
and ideas and morals and
1:39:37
ethics and societal understanding about
1:39:40
the written word, mostly in the early days of the web,
1:39:43
especially before social media really came and mixed that up
1:39:45
a little bit, right. With the whole aggregation of humans
1:39:47
all talking to each other and quoting things and linking
1:39:49
out or whatever. Right. That in
1:39:52
hindsight, that seems much less disruptive, disruptive than
1:39:54
AI stuff, which is like, it's
1:39:56
a free for all. No one knows anything. No one knows
1:39:58
what's legal. What's not. What's sustainable? What's
1:40:00
not? What should we do? What can we
1:40:03
do? What are people doing? How valuable is
1:40:05
this? How useful is it? Like just
1:40:07
so many questions and we like all the laws that
1:40:09
we have that seems like they could apply this and
1:40:12
some of them do apply. It's like, yeah, but there's
1:40:14
these huge areas where it's like here be dragons on
1:40:16
the map and they draw the big dragon and the
1:40:18
thing is like, nobody knows what's there. And
1:40:20
there's a lot of money behind it. And a lot of people
1:40:22
running in that direction. And it's not even clear where
1:40:24
or how this will shift the power. Like
1:40:26
on the internet in the early days, it
1:40:29
was pretty clear paper newspapers, powers going away
1:40:31
from them and towards websites. Like that trend
1:40:33
was visible to anybody with a clue and
1:40:35
it was just a question of how fast,
1:40:37
how hard, you know, whatever. Here, is this
1:40:40
going to shift power massively to the record labels because
1:40:42
they own all the music, for example, or is it
1:40:44
going to destroy them because everything they have is now
1:40:47
worthless because AM models can be trained on it and
1:40:49
it's a perfect substitute for what they previously made and
1:40:51
no one wants anything like you can't even tell which
1:40:53
direction it's going to go at this point. It's so
1:40:55
early and I just don't think that was true of
1:40:57
the web. So this is an exciting
1:41:00
time to be alive in many
1:41:02
ways, especially if you're in any
1:41:04
industry, any creative industry that involves
1:41:06
intellectual property that AI touches at
1:41:08
all. And at this point, that's
1:41:11
nearly all of them. Right. And
1:41:14
right now, what it does
1:41:16
is not, you know, not
1:41:18
particularly amazing, but it is good enough
1:41:20
for so many use cases. And this
1:41:22
stuff generally doesn't get worse over time.
1:41:25
Thank you to our sponsors this
1:41:27
week, One Password and
1:41:30
Photon Camera. And thank you to our
1:41:32
members who support us directly. You can
1:41:34
join us at at.fm slash join members,
1:41:37
get a bunch of perks, including ATP
1:41:39
overtime. This is our weekly
1:41:41
bonus topic. That's an extra segment
1:41:44
that only members get to hear
1:41:46
ATP over time. This week is
1:41:48
going to be about a rumor
1:41:50
reported by the information and Mark
1:41:52
Gurman about some
1:41:54
changes and plans to what Apple is going
1:41:56
to be working on for the next Vision
1:41:58
Pro and kind of what they
1:42:00
can maybe do to make the next vision
1:42:03
pro cheaper and how they're going to possibly
1:42:05
do this and everything. That's what we're talking
1:42:07
about in ATP over time this week. Join
1:42:09
out a listen ATP.FM slash join. Thanks everybody.
1:42:11
And we'll talk to you next week. Now
1:42:17
the show is over. They
1:42:19
didn't even mean to begin
1:42:21
because it was accidental. Oh,
1:42:25
it was accidental. John
1:42:28
didn't do any research. Marco
1:42:30
and Casey wouldn't let him
1:42:32
because it was accidental. It
1:42:35
was accidental. And
1:42:38
you can find the show notes at
1:42:40
ATP.FM. And
1:42:43
if you're into mastodon, you
1:42:46
can follow them at C A
1:42:49
S E Y L I
1:42:51
S S. So that's Casey Liss M A
1:42:54
R C O A R M.
1:42:57
And T Marco R men. S
1:42:59
I R A C U S A
1:43:02
C. Recuse it's accidental.
1:43:07
They didn't mean to. Tech.
1:43:18
Not so real time follow up on my
1:43:20
earlier statement about Apple
1:43:22
Silicon Max not being able to use PCI
1:43:24
breakout boxes. That is not true. You can
1:43:26
use Thunderbolt PCI breakout boxes. Obviously you can't
1:43:28
use it. It's just not GPUs. Yeah. But
1:43:30
you can't use GPUs internally either. That's the
1:43:33
thing. Yeah. So still Apple should
1:43:35
have put the Mac Pro in the configurator. Or
1:43:37
I suppose they could have said, Hey, use PCI
1:43:39
cards buying it except you. You can use PCR cards.
1:43:41
Well, you can buy a Mac studio and also this
1:43:44
third party product that we don't even sell or
1:43:46
you could buy a Mac Pro, which is a product
1:43:48
in their lineup. I think two
1:43:51
things are simultaneously true. Number one,
1:43:53
they should keep making the Mac Pro because
1:43:56
it does have uses. And number two, absolutely.
1:43:59
Nobody should. by the Mac Pro effectively.
1:44:01
Like, anybody who's going to a
1:44:03
page on apple.com saying, what Mac should I buy?
1:44:06
None of those people should buy it. No,
1:44:08
the app, they should. That should be, look, the
1:44:10
whole point of this is it's a path that
1:44:12
leads to all of our products. And maybe there's
1:44:14
only one very lonely overgrown path that leads to
1:44:16
the Mac Pro, but it's got to be there.
1:44:19
I would say number three, your product chooser should let
1:44:21
you choose from any of the products, depending on which
1:44:23
things you answer. Put as many scary questions in there
1:44:25
as you want. There's just got to be a path
1:44:28
that lands in the Mac Pro. Because otherwise, what they're
1:44:30
saying with this is, no
1:44:32
one should buy this product. And I don't think Apple believes that.
1:44:34
If you ask them, they said, well, some people should. Like, OK,
1:44:36
great, but you have a tool that lets people choose, and it
1:44:38
has every single Mac you sell except for that one. That just
1:44:40
seems like a bug to me. Someone should report it, and they
1:44:42
should fix it. You should report it. I
1:44:45
just wanted the Mac Pro that's worth buying. I
1:44:48
mean, that's a bug. Maybe
1:44:50
they're working on that. We'll see. So
1:44:52
I feel like we covered this in
1:44:54
the past, but what are you waiting for? What
1:44:57
would make it worth buying? Am I waiting
1:44:59
for anything in particular? I don't know. Because
1:45:01
again, with the gaming situation on Apple Silicon
1:45:03
Macs is entirely unclear. If I did buy
1:45:06
a Mac with a big beefy
1:45:08
GPU, bigger than a Mac Studio GPU, that
1:45:10
would be a speculative purchase. It would not
1:45:12
be like my current Mac Pro, which I
1:45:14
literally knew I could run Windows games on
1:45:16
and do, and they work fine. And I
1:45:18
run literally put into Windows. That's not speculative.
1:45:20
That's a thing, right? If
1:45:22
I decide, hey, I want a bigger than
1:45:25
Mac Studio GPU in an ARM Mac, I
1:45:27
am crossing my fingers that some magical point
1:45:29
in the future, I will be able to
1:45:31
do interesting gaming things on it. I
1:45:34
don't know if I'm going to make that speculative purchase. I don't
1:45:36
know if Apple's going to make a Mac with a better than
1:45:38
Mac Studio GPU in it. And
1:45:40
maybe they make it, and it's just too
1:45:42
rich for my blood, and I can't spend
1:45:44
that much money on something speculative. Like I
1:45:46
said, my default is an M4 Mumble Mac
1:45:49
Studio is potentially the computer
1:45:51
I will replace this with whenever they release that like next
1:45:53
year or towards the end of this year or whatever. But
1:45:57
I would like to see, show me something.
1:45:59
Show me the Mac Pro. Pro show me something that's not
1:46:01
a Mac studio in giant cavernous case, right?
1:46:03
That's what I would like to see from them and then I can
1:46:05
decide is it worth it for me to get
1:46:07
that? Because it's not a slam dunk like the toy well,
1:46:09
it's not as big a slam dunk as 2019 was because
1:46:11
again, that's just not speculative but It's
1:46:14
just kind of wishful thinking at this point to think
1:46:16
you're gonna be running Windows arm games natively on you
1:46:18
know You're gonna be booting Windows for arm on your
1:46:20
you know Apple silicon Mac
1:46:23
Pro or you're gonna be running Windows
1:46:25
caliber games in Mac OS
1:46:28
Because Apple will have gotten all the triple-a game developers
1:46:30
on board. That is all just Twinkle
1:46:33
in someone's eye right now. It is not a real thing.
1:46:35
I just I feel like
1:46:37
and I'm
1:46:39
gonna say this and I know and I
1:46:41
understand why it's not appealing to you But
1:46:43
I feel like so many of your
1:46:45
problems which well, maybe not even not even problems But so
1:46:47
much of your life would be so much better if
1:46:50
you would just get a damn
1:46:52
Mac studio and a damn Windows PC
1:46:54
And I get it. I don't want to
1:46:56
run Windows anything I don't and I know
1:46:58
you are even worse than me in this
1:47:00
capacity But like that would make so many
1:47:03
things so much better in your life I would probably have a
1:47:05
gaming PC if I had a place in the house for it,
1:47:07
but I don't so I mean I
1:47:09
hate to break it to you But I really don't think that
1:47:11
there is ever going to be a Mac Pro that does the
1:47:13
things that your current Mac Pro does And I mean that may
1:47:15
be true like I'm rooting for it. But like right now the
1:47:18
outlook doesn't look so great Yeah, I
1:47:20
would definitely not hold your breath on that I
1:47:23
mean like the thing is it's actually kind of if
1:47:25
I thought like it two years ago like Predicted
1:47:28
how this would go actually I'm kind of
1:47:30
surprised at how much motion there is here
1:47:33
the copout plus PC how hard? Microsoft is
1:47:35
pushing into arm PCs after doing such a
1:47:37
bad job with Windows RT, right? Apple
1:47:40
with its whole game porting toolkit like
1:47:42
both those parties both Microsoft and Apple
1:47:45
are actually surprising me with how
1:47:47
Hard they're trying to make my
1:47:49
dream happen. They're just not
1:47:51
succeeding right, but they They're
1:47:54
trying more than I thought they would right? I
1:47:56
did not I didn't think they'd be like both
1:47:58
on both sides I have been pleasantly
1:48:00
surprised by the additional effort that they
1:48:02
are putting in. I think everyone kind
1:48:04
of is. It's just like I just
1:48:06
they're just not really doing
1:48:09
it. Alright, but I give
1:48:11
them kudos for the effort. If
1:48:13
I had to pick one thing like I
1:48:15
would I would wish that Microsoft would commit
1:48:18
to a transition to ARM but that's not what
1:48:20
they want to do. They seem to
1:48:23
think that they're going to have a they're
1:48:25
going to support x86 and
1:48:27
ARM forever off into the future
1:48:29
which I think is a dumb strategy but that
1:48:31
seems to be what they're doing and that doesn't
1:48:33
help me and that doesn't help Windows games get
1:48:35
ported to ARM. All that does is bifurcate their
1:48:37
market and say well all the all the AAA
1:48:39
games will still be on x86 with Nvidia cards
1:48:42
and ARM will just free for people's laptops and Microsoft may
1:48:44
be perfectly happy with that but it doesn't help me over
1:48:46
here with Apple Silicon. I mean
1:48:48
what what PC games are you playing with regularity
1:48:50
right now? Destiny. I don't know if you know
1:48:52
this but Destiny runs on PC. But
1:48:55
that's the thing like is there no
1:48:57
other appliance that you can buy to run Destiny
1:48:59
can't you do it on PlayStation? Destiny runs at
1:49:01
higher resolution and higher frame rates on gaming PCs.
1:49:03
I don't really play it on my Mac Pro
1:49:05
I play it on my PlayStation 5 for a
1:49:07
variety of reasons but it does run
1:49:09
better as defined
1:49:11
by resolution and frame rate even on my
1:49:13
Mac Pro. The PlayStation maxes out of 60
1:49:15
frames per second right and I can
1:49:17
get higher than that depending
1:49:20
on settings and you can go way higher you can
1:49:22
go I've played it I actually have played Destiny on
1:49:24
my PS5 at 120 frames per second on my TV
1:49:26
but it has to lower the quality substantially and I
1:49:28
generally don't play Destiny on my TV because it'll burn
1:49:30
it in right but I did try that just to
1:49:32
see what it was like. 120
1:49:35
frames per second is good all the like the
1:49:37
Destiny streamers who are out there playing Destiny they
1:49:40
occasionally have their frame rate displayed in the corner they're
1:49:43
triple digits always hundreds of frames per
1:49:45
second sometimes pushing them in 200 it
1:49:47
makes a difference it it looks and feels
1:49:50
smoother especially in PvP they're like that's you
1:49:52
know and even if I'm playing
1:49:54
on a controller because at this point sadly I'm better with
1:49:56
a controller in Destiny than I am with mouse and keyboard
1:49:59
and also controllers. way better for my RSI so
1:50:01
I'd be doing it anyway. But yeah, Destiny
1:50:03
is one choice. And games come out all the
1:50:05
time and they come out for PC. They
1:50:08
don't come out for the Mac until three years later when Apple
1:50:10
puts in the key note, right? So there's
1:50:13
past games, there's future games, there's my gigantic
1:50:15
Steam library that I still haven't played through.
1:50:19
I mean, I'll get a PlayStation 5 Pro, I'll
1:50:22
get a PlayStation 6, I do like
1:50:24
consoles, they're great. Maybe someday the gap
1:50:26
between PC and console will be
1:50:28
diminished. But even now I would say it's more diminished because
1:50:30
60 frames per second on PS5 is such a change from
1:50:32
30 on the PS4 that I
1:50:34
feel like the gap has narrowed. Because Destiny players
1:50:36
were playing at 100, 200
1:50:39
frames per second back when I was playing 30. Now
1:50:42
they're playing at 100, 200 frames per second and I'm playing at 60, right?
1:50:45
I'm gaining on them. So maybe at some point I'll be like,
1:50:47
you know what, I don't need a big GPU and I'll just
1:50:49
get a Mac Studio and be happy with it. And
1:50:52
that's looking like the most likely situation
1:50:55
right now. But we'll see. I
1:50:57
mean, to be clear, as much as I'm giving you a hard time,
1:51:00
I want you to have what you want.
1:51:02
I can make an argument, even I can
1:51:05
make an argument for the Mac Pro, for
1:51:07
a really beefy Mac Pro that's useful for
1:51:09
people that work outside of a music studio.
1:51:12
I'm not saying that your desires are wants as much
1:51:14
as I'm giving you grief about it. I'm not saying
1:51:16
your desires are wants or unreasonable. I
1:51:18
don't think Apple will be
1:51:21
achieving them, but I don't think they're
1:51:23
unreasonable. It's exciting that they did with 2019, because
1:51:26
again, I've said before, I don't, despite
1:51:28
my gaming things, this is not a purchase. This is not
1:51:30
a rational purchase. It's the same way that you don't need
1:51:32
a Ferrari to get to work faster. People just like fast
1:51:34
cars, they like fast cars. I just like
1:51:36
powerful computers because I like powerful computers. It's exactly
1:51:38
the same thing. Trying to justify a Mac,
1:51:41
me trying to justify a Mac Pro is like someone trying to
1:51:43
justify a Ferrari. It's like, well, I need a car this fast
1:51:46
to get to my work. No, you don't. Nobody
1:51:48
does. But people want them because they're cool.
1:51:51
Right. And that's fair and
1:51:54
that's totally fair. But I feel
1:51:56
like from to my eyes, we're
1:51:58
starting to cross. from, oh, it's
1:52:00
kind of adorable that John is still rocking
1:52:02
his Mac pro to like, man,
1:52:05
I kind of want you to move on to a Mac
1:52:07
studio. Cause I think you might enjoy it a lot more,
1:52:09
you know? Well, I mean, I'm not buying an M two
1:52:11
one at this point. Well, that's fair. No, that's fair. This
1:52:13
is not the time to buy a Mac studio. So it
1:52:15
isn't, it isn't hanging in there for the M four one.
1:52:17
Yeah. I think when the next one comes out, I think
1:52:19
that's your move. I can't, I just cannot
1:52:21
see a future in which they
1:52:23
make the Mac pro that you want. And
1:52:26
so you might as well get the Mac
1:52:28
studio, which is the Mac pro without slots.
1:52:30
Like that is the new Mac pro. I
1:52:32
can't say it enough. And with a wimpy
1:52:34
or GPU, but they, they just like the
1:52:37
Mac studio is the Mac pro. They should
1:52:39
have called it Mac pro. That is the
1:52:41
Apple Silicon Mac pro. They should not have.
1:52:44
No. Can you imagine the aneurysm he would
1:52:46
have had? It doesn't make sense. They
1:52:48
sell, I think all the Mac pro it's way bigger, but
1:52:51
it's the same computer. It's just a built-in
1:52:53
PCI breakout box. I know. It's still got
1:52:55
the slots. It's still anyway, but we'll see
1:52:57
how good, and by the way, by the
1:52:59
time I do get this, like my computer
1:53:01
is essentially five years old now already.
1:53:04
This is a pretty good run for a computer that
1:53:06
I bought at the, you know,
1:53:08
just before the, the, you know, processor
1:53:11
transition, right? Which is, you know, unfortunate for the, we
1:53:13
already said when it happened, like, Oh, my poor Mac
1:53:15
pro or whatever, but I love this machine and
1:53:17
I've already gotten five years out of it, which granted is half of
1:53:19
what I got out of my last Mac pro, but you know, processor
1:53:22
transition, right? So if
1:53:24
I ditch this machine at six years old, that's
1:53:26
longer than any of your laptops lasted, right? It's
1:53:29
a pretty good run. Hey, we're just excited
1:53:31
if Marco makes it six months, much less six years.
1:53:33
He's been pretty good with 16 inch. I think it's
1:53:36
almost two years old now, right? No, it's the M3
1:53:38
max. It's a, it's the black one. Sorry.
1:53:41
I mean, to be honest, lately I haven't
1:53:43
been much better, so I shouldn't be casting stones
1:53:45
in this glass house, but generally
1:53:48
speaking, Marco is much more frequent on, on
1:53:50
his purchases. So I mean, like no matter
1:53:52
what, like. I feel like I've gotten
1:53:55
a good run out of this Mac pro and I'm enjoying
1:53:57
it for, you know, as long as like I'm excited that
1:53:59
it's a. runs on it. That's
1:54:01
cool. Next year, probably not, right? So it's really
1:54:03
putting a deadline on this. Like I said, I'm
1:54:05
willing to run this with last year's version of
1:54:07
the operating system for some period of time if
1:54:10
I have to wait, right? But, you
1:54:12
know, well, we'll see what happens. Like I'm, you know,
1:54:14
I keep my cars for a long time and keep my Macs for a
1:54:16
long time.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More