Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
Heads. Up this one is
0:02
a bout a fictional computer
0:05
from a nine These Children
0:07
This Enemy Television series. For
0:10
like. Thirty minutes? Just about that.
0:13
So. That's
0:15
what this is. And.
0:19
Now I'm gonna play the Pokemon theme
0:21
music but all playing in a mall
0:23
at night I find. Any
0:28
pilot of the very popular
0:30
Nineteen Ninety Seven animated television
0:32
series, Pokemon we introduced to
0:34
a portable computer basically like
0:36
the smartphone of the pre
0:38
smartphone world of Pokemon called
0:40
a poker decks. Just
0:43
like it decks. For.
0:50
Some of you me even explaining what
0:52
a pokemon x is exceptionally weird because
0:54
it's very obvious. For others it would
0:56
be equally weird if I didn't explain
0:58
what a poker Texas. The
1:02
Balkan X has evolved endlessly over
1:04
the years, but the original is
1:06
a small red portable computer if
1:08
you can google image it safely
1:11
giver. Otherwise, it has a book
1:13
style form factor with you know,
1:15
a small screen, a couple buttons
1:17
almost looks like a half a
1:20
calculator on the right side and
1:22
what's basically a Gameboy lay out
1:24
on the left side. And then
1:26
I was like a camera sensor,
1:29
something which The Pocket X could
1:31
importantly. Used to identify pokemon. the the
1:33
general use shown in the show is
1:35
like you walk up to like a
1:37
squirrel. you hold it up and it's
1:39
like that that the squirrel and it
1:41
tells you just enough about the pokemon
1:43
that the person watching the show now
1:46
knows what. that pogo on his. Which.
1:48
Implies that it has computer vision. And
1:51
text to voice synthesis because
1:53
it could tell the protagonist and
1:55
by extension that young audience.
1:57
Important exposition about the Pokemon. For.
2:01
What now to me as an adult seems like
2:03
marketing purposes. The
2:16
screen. Very Ninety Ninety Seven,
2:18
but it was. Other was a
2:20
very futuristic portable computer. He
2:23
could do stuff the would have been pretty
2:25
challenging for a portable computer to do until
2:27
like the last few years. And
2:30
ape we've been hearing
2:32
recently hacked when Together.
2:37
And I think if you look at
2:39
the history and might evolution of hardware
2:42
and devices that we deal with, a
2:44
lot of them take inspiration from fiction,
2:46
their science fiction ideas or whatever that
2:49
eventually became feasible and I think a
2:51
lot of a time what what's really
2:53
happening in sites technologies progressing and is
2:56
enabling things that are non obvious and.
2:58
I. Need a lot of sense of people who come
3:00
up with these ideas in the real world. Get.
3:03
The inspiration for that aware set at that
3:05
want that adds to that make said. You
3:07
know that last ten percent that really makes
3:10
a notable? They get that from fiction. The.
3:13
Six your computer for my use is
3:15
now a homebrew d I y Project
3:17
Street. He printed the body use the
3:19
school blend of still relatively new A
3:22
P Eyes were going to talk about
3:24
adult A custom poker decks are west.
3:26
kind of like stitch it all together.
3:29
So. Can opener petted x.
3:31
Point. It at a pokemon or
3:34
rather sculpture toy. drawing. any rendering
3:36
of a pokemon he will identify
3:38
it, displayed stats and tell you
3:40
about it. in the original voice
3:42
of the token decks we're getting
3:44
to the point with really printing
3:46
technology and access to a programming
3:48
tools and like these cheap parts
3:50
and all stuff where. It. Would
3:52
be feasible. For. You to
3:54
really build something that you used
3:56
every day but like from scratch
3:58
and quotes. You could make a
4:01
laptop you could make up taste for you that you
4:03
could make these things I would love to see. More
4:05
people seeing a project like this and having that
4:08
reaction that is like oh that's really cool. I
4:10
to do something like that or I could take
4:12
this idea, do my own thing. Sometimes.
4:15
We've got to talk about heavy
4:17
cyber crime type stuff on the
4:19
shelf. You were even be getting a few
4:21
of those are the next help Lab results. And
4:23
sometimes I get sad sack about stuff like
4:26
this. About. A
4:28
six your computer. From. And ninety's
4:30
Enemy. That. A packed together.
4:34
This is a fun one. Is
4:36
my conversation with ape haskins? a kid
4:38
A at a today on you tube?
4:41
About the D I y poco next. Year.
5:00
Saves ages much for us and down with me. Faint.
5:03
Your thanks for having me. Because. Of the
5:05
broadly speaking, for anyone that doesn't know
5:07
what is a token decks in the
5:10
games on the animated show originally, what
5:12
is it and what is it do.
5:14
So a pocket X is a small
5:17
device given by a poker mon professor.
5:19
It's when you when you decide as
5:21
an eleven year old that you're going
5:24
to go capture. poke him on I
5:26
guess for for a living free job.
5:28
Yeah, yeah, for your job they're like
5:30
I'd your eleven years old You don't
5:33
know anything about how. come on here
5:35
is this magical ah device bad. Somehow.
5:38
Has all the information by Parker Man and
5:40
the the general use shown in the show
5:42
is like you walk up to like a
5:44
squirrel your hold it up and it's like
5:46
that's that's a squirrel and it tells you
5:49
just enough about the puck a that the
5:51
person watching the show now knows what that.
5:53
oh come on it's if is. But
5:56
it's basically just like a it's a
5:58
tool in the show too, My
6:00
pokemon are opaque technically a tool but
6:02
also Agatha for the audience understand and
6:04
learn about yeah the dozens and dozens
6:06
of pokemon that accessed via physically small
6:08
little red device hand can wix, I
6:11
got a palm pilot. Are you say
6:13
like something kind A great yeah ninety
6:15
this. Most of the designed to these
6:17
are from the late nineties early two
6:19
thousand so they're kind of. That
6:21
version of a toy. a computer
6:24
kind a login thing. Ah,
6:27
And legacy there is some other. Stuff.
6:29
In there in some episodes Dell Dell actually
6:31
like use the on screen thing delight. Call.
6:34
Someone or something bell rang as if that
6:36
really is not used much. it's it's like
6:39
one random time they needed a plot device
6:41
and he could become that. but that most
6:43
slates just identifying dopamine is it's main thing.
6:45
Yeah it's funny when I was thinking about
6:47
this before a conversation I was just remember
6:50
that others are basically a it's an encyclopedia
6:52
pokemon be you are right it it functions
6:54
as a little bit of an exposition or
6:56
device like it is a means by which
6:59
they can just transmit information the audience. Here's
7:01
what a right to his or whatever. Yep,
7:03
Yep, exactly yeah. and I think that's what
7:05
I think a big problem in the say
7:08
Originally designed like we have a hundred and
7:10
fifty pokemon reacted, explain and we have to
7:12
get kids to love and correlate with as
7:14
the the other interesting is already attained of
7:17
in the pocket acts by the other thing
7:19
They did help people understand other pokemon was.
7:22
On. Commercial breaks. they would do this for
7:24
them and will come on segment where they
7:26
show an outlier milana an end up and
7:28
then you come back. So that is also
7:31
helping kids to memorize all the pokemon and
7:33
they'll start a song at the end which
7:35
was us hilarious Ninety Ninety eight rap song
7:37
about a have a different pokemon to also
7:40
have you remember to. they were really clearly
7:42
trying to each other some to the i
7:44
don't know if they were just like. We.
7:47
don't know which pokemon are going to be important in
7:49
the show or if they were just like we're trying
7:51
to get people to remember every single one but there
7:53
were multiple things including the bug in acts which were
7:55
just designed to get people to the point where they
7:58
knew what elaborate was which and probably failed because most
8:00
people I don't think know what a Lapras is. But,
8:02
you know, they were clearly putting a lot
8:04
of work into it and this device was part
8:07
of that. Yeah, kind of the beating heart.
8:09
So, okay, I could get pulled off onto
8:11
tangents about the Poker App all
8:13
day. The Poker App is good, you should look it
8:15
up. It's really good, it is really. It's vitally funny.
8:17
You should have it on your iPod if
8:20
we're talking about nostalgic tech. But just
8:22
to zoom in on the tech itself, the
8:25
show premiered in, I checked this before we talked, April
8:27
1st, 1997 here in the West. At
8:31
that time, this technology, which was a Palm
8:34
Pilot type device with
8:36
like a camera that
8:38
could identify a thing in the real world and then
8:40
talk to you, 1997, this
8:42
technology, a portable computer, this was not
8:45
possible. Just not to put too
8:47
fine a point on it, but like what
8:49
was impossible about the Pokedex when this came
8:51
out? Given that you did make one, really.
8:53
What was impossible? So, we had small
8:56
form factor device, right? So,
8:59
we had small things that were battery powered,
9:01
that had some level of operating system
9:03
on them. Some of them had speakers
9:05
and screens.
9:07
So, we had like several of the
9:10
core chunks, right? We had the form
9:12
factor of a Palm Pilot, PDA kind
9:14
of device. Things
9:16
like cellular phones were starting to
9:19
show up. They still weren't commonplace
9:21
in 97, but like
9:23
they were getting there for sure. And
9:26
so, you have a couple
9:28
core components there, right? You have a screen, you
9:30
have the ability for it to make noise, you
9:32
have input of some form, you have the form
9:34
factor of the battery. The cellular network to assume
9:36
like somehow this is like fetching
9:38
information from that professor who gave it
9:41
to you or like it's accessing an
9:43
encyclopedia somewhere that isn't actually on the
9:45
device. All of that was
9:47
close to possible. At least possible if like
9:50
a company the size of, whatever
9:53
Pokemon company in the show makes these with like
9:55
doing the research, like we could get close.
9:57
The part that really was not. there
10:00
was the computer vision and
10:03
the ability to take a photo and have any
10:05
idea what was in it. That
10:09
was like the technology they had at
10:11
the time was like, maybe they
10:13
could take a photo and like
10:15
see the shape of a person in it.
10:18
And like maybe they could do some
10:21
very basic analysis, but like cameras
10:23
weren't really there, webcams, any sort
10:25
of digital photography just as a
10:28
whole was not really there.
10:30
The first digital cameras were coming
10:32
out in the 90s really, there might have been a
10:34
couple early ones, but they were around that time that
10:36
they were like becoming feasible and
10:39
like they weren't good. So like
10:41
these pictures not only are very bad pictures
10:43
compared to traditional photography, but we also didn't
10:45
have the computational knowledge or power
10:47
to like analyze them and be like, this is
10:49
the thing. The computer vision
10:52
aspect of it and the understanding of what
10:54
it is seeing, that wouldn't come around to
10:56
anything close to capable of doing that until
10:59
probably 2014 ish. So
11:02
it was quite early, but everything else
11:04
like the hardware, it was
11:06
not that far off. The hardware was close to
11:09
reasonable. Relevant
11:11
to your project in the last year and
11:13
change, boom, we all get pretty
11:16
easy API access to really,
11:18
really, really good computer vision.
11:22
And you were able to use that amongst
11:24
a couple other things to DIY hack together
11:27
this device, the Pokedex. Do you
11:29
want to take me through the
11:31
basic sequence of technologies that goes
11:33
into building this thing? Yeah.
11:37
So the problems you have to solve
11:39
if you're building any device, any handheld
11:42
device like this, whether it's a little
11:44
game console or something like this, there's
11:47
kind of those table stakes of like
11:50
you need battery for it. You need
11:52
some sort of processor microcontroller, probably
11:54
need some sort of display and need
11:57
some sort of like button input. All
11:59
of that. They're not
12:01
really like solved problems, but they're things that
12:04
people have done. There's nothing that crazy about
12:06
that. You can kind of look up, you
12:09
know, how can I add a rechargeable
12:11
battery to my microcontroller project and you
12:13
can figure it out. So the core
12:16
components there are, like I said, microcontroller,
12:18
that is actually the brain of the
12:20
little device, some sort of screen. I
12:22
used a little OLED screen in mine.
12:25
You know, there's a bajillion options that you could go
12:27
with. All of that will get you
12:29
to the point where you have a device with a screen and
12:31
a button. And after
12:34
that, that's when you get a little
12:37
bit more into uncharted territory. And
12:39
it comes to that like integration
12:42
with those APIs that are available,
12:44
chat GPTs, GPT-4 Vision API. There's
12:47
other models available that have, you
12:49
know, vision capabilities. And
12:52
that is stuff that although
12:54
technically reasonable to do on
12:56
these really low power,
12:58
like a microcontroller, you
13:00
know, a modern microcontroller, you can
13:02
compare the processing power it had
13:04
to probably like somewhere
13:07
between a Game Boy Advance and a PSP. So
13:10
they're pretty light
13:12
device, like they're pretty weak devices, but
13:15
they're not crazy weak. Like they're okay,
13:18
especially considering that they cost like $6. So
13:23
the part that gets tricky is
13:25
writing code that can deal
13:27
with these APIs and talk
13:30
to these models and deal
13:32
with the unexpected nature
13:34
of large language models that sometimes
13:36
just don't answer your question or
13:38
they do random stuff. And
13:41
writing that in and writing code that
13:43
is actually like reliable and consistent and can do that.
13:46
And again, like this is not uncharted
13:48
territory for one of
13:50
these little controllers. Most of the time, these
13:52
little microcontrollers are used to like read a sensor
13:54
and post that up to a server somewhere.
13:57
We're not doing something that different reading a camera
13:59
and post. but sending it up to chat
14:01
GPT. But like
14:03
the devil is really in the details
14:05
of the software here to make it
14:07
nice, to make it feel reliable, you
14:10
know, even dealing with stuff like wifi
14:12
connections or something we take for granted
14:14
on a mobile phone, it, you know,
14:16
really flawlessly switches between cellular and wifi
14:18
and stuff like that. These microcontrollers, they're
14:20
much lower level. There's times
14:22
when like wifi just won't connect, you have
14:24
to deal with that. All sorts of little
14:27
details. So yeah, but
14:29
that's kind of where you get to, you get like the hardware
14:31
that is not that, you know, new.
14:33
People have been building little portable devices
14:35
for a long time in the hobbyist world. And
14:38
then just kind of pushing into, you know,
14:40
what type of safety and what do we have
14:42
to design in those large, like when interacting with
14:44
those large language models to make it feel
14:48
complete and feel like a thing that actually works
14:50
versus like something that works like 10% of the
14:52
time and crashes and all that
14:54
stuff. But
14:56
yeah, that's kind of where the difficulty lies
14:59
and where the balance kind of
15:01
is. It's funny, when I first read coverage
15:03
of it, I was thinking this is so cool, but
15:05
I can imagine how it works. You know,
15:08
it's you using chat GPTs, computer
15:10
vision, you sort of imagine it. And then
15:12
you realize you actually watch the way it works. You're like, no,
15:15
yes, you can analyze the
15:17
image using that API, but
15:20
you have to get the answer back out and
15:22
then display it as this low resolution version of
15:25
the thing. Like the art of it really was
15:27
in recreating the Pokedex OS for lack of a
15:30
better term to make sure that that process of
15:32
using it felt seamless and analogous to
15:34
what was in the show. Yeah, and I
15:36
think that's the case a lot of time
15:39
where with technology especially, and
15:41
I think if you look at the history
15:43
and like evolution of hardware and devices that
15:45
we deal with, a lot of them take
15:48
inspiration from fiction, their science fiction
15:50
ideas or whatever that eventually became
15:52
feasible. And I think a
15:55
lot of the time, what's really happening is
15:57
like technology is progressing and
15:59
it's enabling things. that are non-obvious, that
16:01
are that extra percent you're talking about
16:03
where it's like, yes, we
16:05
technically could do this, but there's a
16:07
design element to it. There's a warmth
16:10
to it that makes it magical to
16:12
a consumer. And I think
16:15
a lot of times the people who come up with
16:17
these ideas in the real world get
16:19
the inspiration for that OS, that
16:21
warmth, that edge that makes it
16:23
that last 10% that really
16:25
makes it notable, they get that from
16:27
fiction. So, you know, the
16:29
Pokedex is not a really great example of
16:31
this, but it's kind of the same thing
16:34
that like what makes this project interesting is
16:36
its connection back to
16:38
that and the great design
16:40
of the original device. That it's so
16:42
simple, like it's a really cool little
16:44
thing that they designed, and
16:47
now I can kind of stand on those shoulders
16:49
and make a realer version of it, but it's,
16:52
you know, because of that original design work that was
16:54
so good that I can then be like, okay, now
16:56
when I have all these random components that aren't that
16:58
interesting on their own, if I make it into this
17:01
thing that was already cool, now
17:03
it, you know, feels like something grander
17:05
than the sum of its parts. It's
17:07
poking at some deep core memory. Warps
17:10
is such a good word for it. I'm, every
17:14
so often will be traveling. You know, the phone is
17:16
such a regular part of your life, it kind of
17:19
just becomes invisible. It's just wallpaper at a certain point.
17:21
But whenever I go traveling and I find myself
17:23
in an unfamiliar environment and I get the SIM
17:25
card plugged into it and I bring up Google
17:28
maps, I have this Hitchhiker's Guide to the Galaxy
17:30
moment where I'm suddenly like, I'm out there in
17:32
the world, but I have this thing and it
17:34
gives me access to all of the information I
17:36
could ever need. And the capability and the utility
17:38
that is inherent in this thing that is again,
17:40
invisible to me, you know,
17:42
99 days out of a hundred, suddenly it
17:46
becomes clear again. It
17:48
is only that connection to a fictional device,
17:51
rooted in sci-fi through
17:54
which I'm able to do that. It's an interesting,
17:56
is an interesting point. Yeah. And it's cool to
17:58
have those moments, right? Where you realize. you
18:00
know, the like really
18:03
sci-fi world we're living in. Because
18:07
yeah, things like just even having a map in
18:09
your pocket, I mean, that's a completely novel idea.
18:12
It's amazing that it works
18:14
as well as it does. You
18:17
know, I used to work at Google, I
18:19
worked near the Google Maps teams and I,
18:22
you know, always really respected
18:25
the design of Google Maps specifically in terms
18:27
of product design, visual design as well, keeping
18:29
it as simple as it is. But you
18:32
know, having this really, really
18:34
hard technical problem of digesting
18:37
a world of map
18:39
data and taking
18:42
it and making it so simple
18:44
and so approachable that everyone in
18:46
the world, one
18:48
of their top like three use cases of their
18:50
phone is using Google Maps.
18:52
Like that's crazy.
18:54
And it's a really,
18:58
Google Maps specifically is a really magical
19:00
thing that summarizes a world of information
19:02
into something that it's so approachable. It's
19:04
really a notable example of this type
19:06
of thing where it's more than the
19:09
sum of its parts, but it's also
19:11
hiding so much complexity under the surface.
19:13
Well put, you use the term warmth.
19:15
And I think that so much of the, when I
19:18
think of that device from the 90s show, I
19:20
picture the hardware and I think about the voice. Because
19:23
as you said, it was an expositional device from the TV
19:25
show. Use the service called I think
19:27
Play HT for the voice. Take me
19:30
through that cloning process. How did you get this thing
19:32
talking? Yeah, I mean the cloning was, I mentioned in
19:34
the video, but it was trivial. I
19:36
looked up a clip of the Pokedex
19:38
on YouTube, downloaded that clip and
19:41
like converted it to a WAV file. And it
19:43
was a really bad clip too. It was like
19:45
maybe 10, not even 10
19:48
seconds of voice and
19:50
handed that over to that language
19:52
model and it immediately got, what
19:54
I would consider somewhat close, depending
19:58
on who you ask, it's closer. farther
20:00
from the original voice. But
20:02
yeah, the actual process was
20:05
functionally nothing. It was minutes
20:08
of work to get it to be able to
20:10
generate something that sounded reminiscent of
20:12
the original voice. It was unreal.
20:14
We've every so often, every couple of
20:16
years we've done check-in episodes on the
20:18
state of voice synthesis on this show.
20:22
And from where we were even two years ago, where
20:24
I would have to record 15, 20, 30 minutes of
20:27
speaking a specific script into a system to get a
20:29
result that I would call like a C, C
20:32
plus maybe, to
20:34
where we are now where I can shout
20:36
at a phone for 15 seconds
20:39
and it will do a recent, like a
20:41
not bad version of my
20:43
voice. It's pretty wild. Yeah, it's incredible. I mean,
20:45
it's definitely a very recent improvement. I was amazed
20:47
when I did it, that it worked as well
20:49
as it did. That one's a little bit of
20:52
a magic trick. And then the other part of
20:54
the pipeline, and this was the, I
20:56
hadn't heard of Play HD for the voice cloning,
20:58
but it was similar to things I had bumped
21:00
into. I'm familiar with the
21:03
OpenAI API. And
21:05
then I came across the Pokemon API.
21:09
Hadn't heard of that one. Can you tell me a
21:11
little bit about it? Yeah, the Pokemon
21:13
API is, it's like
21:15
a fan project to
21:17
provide every bit of information
21:20
about Pokemon you could
21:22
possibly want in a
21:24
structured format. In other words, like
21:26
a machine readable API format. It's
21:30
really mostly used for like
21:32
educational stuff. It's really common
21:34
for university students who need a big data
21:36
set to just like poke around with
21:38
like something they're mildly familiar with. It
21:41
gets used a lot for that. It
21:46
has more detail
21:49
about Pokemon than you could ever possibly
21:51
imagine. Like down to the
21:54
sprites for every single game every Pokemon's
21:56
ever been in, including all the offshoots.
21:58
It has every bit of
22:01
flavor text that every Pokemon has
22:03
ever been described as in every
22:05
Pokemon game. So
22:07
the amount of that that I used
22:09
was really minimal. I basically pulled Pokemon
22:12
names and associated those with sprites that
22:14
I thought would look good on the
22:17
black and white screen, but it's
22:19
really a very impressive dataset. I
22:21
think it's like mildly out of date now,
22:23
like it hasn't been, you might be trailing
22:25
a little bit with the new games, but
22:27
I mean it's a really cool resource and
22:29
something great for like people who are beginning
22:31
the program because it's just so much data
22:34
to poke around with and you know build
22:36
an app on top of or whatever. Okay,
22:38
that makes a lot more
22:40
sense now, the educational use case,
22:42
because on the you go
22:44
to the website it says it's serving a billion
22:46
API calls each month, and I was like I
22:49
get that Pokemon is popular. Lord, what are people
22:52
doing with this thing? Yeah, there's a bit
22:54
of a catch-22 with things
22:56
that are examples for programmers.
22:59
The same issue happens with example.com.
23:02
example.com serves an
23:04
absurd amount of requests because it's
23:07
used in every textbook and every
23:09
example as like the default domain,
23:12
so it's kept up basically as a public
23:14
good for programmers
23:16
to practice against. Right,
23:19
sure. That's yeah, you don't want that one falling
23:21
into the wrong hands because a lot of textbooks
23:23
suddenly become kind of insidious. Yeah, exactly, yeah. You
23:25
got to keep it safe. I think Microsoft owns
23:28
it. I might be wrong though. microsoft.com is also
23:30
a very common one where they were
23:32
like please stop using us as an example because
23:35
it's just sending absurd amounts of traffic,
23:37
especially for young programmers who like write
23:39
an infinite loop that curls the website
23:41
a thousand times a second. Okay, they're
23:44
like please stop doing that.
23:47
Okay, so you got play HD doing
23:49
the voice, you got chat GPT doing
23:51
your computer vision, you
23:54
got poke API with the hard data.
23:57
Tell me a little bit about the hardware. How do you put this
23:59
thing all together physically? Yeah, so
24:01
like I said, the electronic
24:04
components, nothing too new
24:06
there. You
24:09
can get screen
24:11
modules that have good documentation that are
24:14
basically like wire it to a microcontroller
24:17
in this way. Buttons
24:19
are easy. Speakers
24:21
are a little trickier just because
24:23
any time a component is moving, which
24:26
a speaker does, it has the magnetic
24:28
element of it that's physically moving to
24:30
produce sound. You
24:32
get a little sketchy because any time things move,
24:34
it's a little harder. But all of
24:37
that, pretty reasonable.
24:40
You could Google any given component of that and
24:42
find a dozen tutorials explaining how to wire it
24:44
up. So that part was all pretty straightforward. And
24:46
then the big issue with these types of projects
24:49
is just if
24:51
you have an iPhone, that
24:54
internal circuitry is super
24:56
precisely engineered and very
24:58
carefully laid out and
25:00
designed to fit in that small package.
25:03
Even something the size of the Pokedex, which
25:05
is, I don't know, maybe three quarters of
25:08
an inch thick or something like that. That's
25:11
pretty tight when you're dealing with like
25:13
off the shelf screens or
25:15
you're dealing with off the shelf batteries
25:18
or whatever. So
25:20
the tricky part there is really just
25:22
designing a 3D case and printing it
25:24
in a way that once you have
25:26
that, all your components actually lay perfectly
25:28
and the device has a form factor
25:30
and the thickness that you want. It's
25:33
not that difficult, but it is the type of
25:35
thing that if you don't really think ahead, you'll
25:38
be like, all right, my project's almost done. And
25:40
then you go to close the cover and it
25:42
just doesn't actually close because you didn't think about
25:44
the wires or you didn't think about two things
25:46
sitting on top of each other. Like the solder
25:48
joints, which add like a millimeter of height or
25:50
whatever, you're like, dang it, now that doesn't quite
25:52
fit. But
25:55
yeah, the main process, as outlined in
25:57
the video, you figure out
25:59
the gist of the components you want to use, go
26:02
do a 3D model case, print out
26:04
the case, hope everything fits, iterate
26:06
on that when things inevitably don't fit,
26:09
and you can get there pretty quick. I
26:12
caught a short that you posted where you were
26:14
talking about how the sort of like primary use
26:17
case for this is pointed at a toy, 3D
26:19
object of the Pokémon, and then you did sort
26:21
of a quick pencil crayon sketch
26:23
of a Gengar, and you
26:25
took a snap of it. I guess this
26:28
is basically just a question about the reliability
26:30
of the chat GPT API, but to
26:33
get abstract here, if
26:35
you were to come across a Pokémon in the
26:37
real world, just like a Charmander
26:39
chilling, do you think it would
26:41
work? Do you think that it would correctly
26:43
identify a living animal? It's
26:45
a good question. I would
26:47
say probably. We'd
26:50
have to get a real Charmander to
26:52
test this, but expensive. Yeah,
26:55
expensive. We'd have to genetically engineer
26:58
a Charmander. Probably worth
27:00
it for the test. If you
27:02
got the money. Yeah. No,
27:05
I think it's feasible to think
27:07
it would. These
27:11
models are really good at understanding
27:13
things that are adjacent to things, right?
27:15
So if you're looking at a cartoon
27:17
drawing of a Pokémon, that is adjacent
27:19
to a 3D rendering of
27:21
a Pokémon, something like Detective Pikachu, where you're
27:24
like, these are not super
27:26
realistic models, but they're real. They're
27:30
close enough. And then you can imagine
27:32
that is adjacent to what an actual photo of
27:34
a real Charmander would look like. I
27:36
think it's reasonable to think it could make that
27:38
leap through those things.
27:41
I think if we
27:43
had less data about Pokémon out there,
27:45
if, for example, we only had the
27:47
sprites from the original games, which are
27:49
just black and white pixel art, that
27:52
jump it would never
27:54
make. But because we have 30 years
27:57
of fan art and fan renderings and
27:59
movies and all of these
28:01
things, like its understanding of the
28:03
geometry of a Charmander is probably
28:05
actually quite good, but it probably
28:07
also breaks down with newer
28:10
and newer Pokemon because it has less and
28:13
less resources to pull from
28:15
and understand. Right, and later
28:17
generations are quite as carefully
28:20
and thoroughly indexed across decades and
28:22
decades of media. Yeah,
28:24
exactly, and they're just newer, even if it was
28:26
the same rate, right? They haven't had
28:29
the same time to get that exposure. Eh,
28:32
but a Charmander specifically, I think you'd have a
28:34
very good chance. This
28:39
is Jess Betancourt, the host of DNAID, the
28:42
only true crime podcast that exclusively covers
28:44
cases solved using forensic genealogy. DNAID goes
28:46
behind the headlines to answer your questions
28:48
about this remarkable new crime solving tool,
28:50
how it works, how cases are selected,
28:53
why the cases were unsolved for so
28:55
long, and how the justice system is
28:57
addressing it. I include input from law
28:59
enforcement to give you the inside scoop
29:01
that we all crave with a straightforward,
29:03
no-nonsense delivery. You can find DNAID on
29:06
any podcast platform. Episodes come out weekly
29:08
on Mondays. You
29:13
obviously do a lot of DIY open source
29:15
projects like this. How
29:20
has the response from the community been to this
29:22
in particular? Have you thought about collaborating with others
29:24
on something like a Gen 2 of this? Do
29:27
you think about open sourcing it? How
29:29
have people received your Pokedex? No,
29:32
I mean, I think people really like it. I think there
29:34
is a really
29:37
positive reaction for most people. I
29:39
do think there's an interesting cultural
29:43
thing that I think is
29:45
interesting. I've
29:48
had with a few other projects, which is where I'll
29:51
create something like the Pokedex, which I just
29:53
do basically for fun. I mean, it's part
29:55
of my job as a YouTuber, but it
29:57
is mostly for my own entertainment. And
30:01
the reaction people have is like, can I
30:04
buy that? Like will you sell it to
30:06
me? Are you planning on turning it into
30:08
a product? You know, that type
30:10
of thing. And
30:13
to me, the answer is obviously no
30:15
to both of those things. One, because, you know,
30:17
it took me a long time to
30:19
make it. I'm not gonna say anything you would buy it
30:21
for would not be worth it to me to sell. But
30:23
also, like, that's not the
30:25
point. And I think that's
30:27
kind of the interesting reaction people have to
30:29
this type of project. And something that I
30:32
think would be really cool to work on
30:34
as a culture is to say
30:37
like, we're getting to the point with
30:39
3D printing technology and access to programming
30:41
tools and like these cheap parts and
30:43
all this stuff where it
30:45
would be feasible for you to
30:48
really build something that you use every
30:50
day, but like from scratch
30:52
in quotes, like you could make a laptop. You
30:54
could make a case for it. You could make
30:56
these things. And I would love to see more
30:59
people seeing a project like this and having that
31:01
reaction that is like, oh, that's really cool. I
31:03
could do something like that. Or I could, you
31:06
know, like take this idea and do
31:08
my own thing versus the reaction I get now
31:10
a lot, like I said, which is like, can
31:12
I buy this, which is also great. And I
31:15
appreciate that support and enthusiasm. But
31:17
my real goal is to have
31:19
people walk away inspired to like
31:21
learn about these technologies and to
31:24
do their own thing because we're at a point more
31:26
than ever where it is feasible to like make your
31:28
own stuff. And I think that's super cool. Yeah.
31:32
Can I buy that is so flattering for
31:34
a hardware project you built out of parts
31:36
like this. It's really,
31:38
really cool. And I would have, I'm reticent to
31:40
even bring this up because it feels like summoning
31:43
a very litigious Bloody Mary, but it is also
31:45
the point at which an open source DIY project
31:48
intersects with a large corporation that
31:50
does have lawyers and IP that
31:53
they like to protect. Like how do
31:55
you, how do you navigate creating an
31:58
open source project like this? knowing
32:00
that it is based on that pre-existing
32:03
intellectual property. Yeah, you
32:05
know, I think it's always a fine
32:07
line. Nintendo specifically, who's
32:10
obviously the big owner of the Pokémon
32:12
company, who owns the Pokedex IP, they
32:16
are known for being particularly litigious,
32:19
which I get. And I respect, as someone
32:21
who's worked for a lot of big companies,
32:23
done a lot of complicated projects, I get
32:25
why they need to do that. I
32:28
think with something like
32:30
this, you have a bit more leeway,
32:33
especially if you're not selling something, if
32:35
you're not, you
32:38
know, really monetizing it
32:40
in any way. And especially
32:43
something like this, where it's
32:45
clearly a prototype, clearly a demo,
32:49
you know, I'm not going out and saying that
32:52
this is, you know, anything.
32:57
Even something like fan games that people will
32:59
sometimes do, you know, they'll remake Mario or
33:01
whatever, and Nintendo gets mad about that. Even
33:03
that's quite a bit different, because it's closer
33:05
to a real product in a real space
33:08
that Nintendo would get into. But,
33:10
you know, I'm
33:12
not particularly concerned about that. You
33:15
know, I would never sell, like, I
33:17
would never productize this and sell it
33:19
to begin with, but I would also
33:21
not do it, because, as you point out, it's not my IP.
33:24
And I think there's just, like, a bit of an ethical quandary
33:26
there where I'm like, I'm not particularly
33:29
interested in doing that. But if someone wants to take
33:31
this, make their own version,
33:33
do whatever, make it something like a Pokedex,
33:35
but not, you know, a lot of
33:37
people suggested in the comments, like, oh, it'd be great
33:39
if it's worked with real animals, or if it worked
33:42
with plants, or whatever. Like, if you want to take
33:44
this and run with this idea and
33:46
turn it into something like that, that's an original idea
33:49
that is not, you know, it might now have a
33:51
reminiscent form factor, but it's, you know,
33:53
it's its own thing. So I think those types of things
33:57
are really cool, but, yeah, it's always
33:59
a little bit of a problem. The. Minefield.
34:02
If you're doing anything that isn't completely
34:04
original, I'd be very to me. this
34:06
is such a good. Even more so
34:08
then I think that those other games
34:10
built on. You know obviously they're
34:12
reworked versions of the pre existing games
34:14
but even that is so much more
34:16
based on had creative output from the
34:18
Nintendo corporation that this a sample Certain
34:20
this is you know the out is
34:23
somebody just love spoke I'm on him
34:25
wanted to make a thing from the
34:27
show and had access to the technology
34:29
to make it. Ah yes. Yeah.
34:31
I view it more akin to like a
34:33
prop making and cosplay as going I like
34:36
as you would never look at someone dressed
34:38
up as messenger chief and deal I
34:40
wait a minute be about to get suicide
34:42
yeah but not at the be I got
34:44
it's cool. It's cool that you made that
34:47
armor as a as his nails are getting
34:49
more functional but yeah same same kind of
34:51
space the system as in layers sprinting
34:53
around a clean with this account is tackling
34:56
totally one person after another. Yeah. Before
34:59
he figure out there's more of us that there are of
35:01
them. When. He bring
35:03
up an interesting point in oh I
35:05
think about you're talking at the very
35:07
start of a conversation about size I
35:09
technologies. it inspires and at some point
35:12
becomes just checked knowledge. He and Nine
35:14
and I think about this device which
35:16
is basically a multi modal ai with
35:18
computer vision that you carry around in
35:20
your pocket and then I think about
35:22
the Med or a dance. Which.
35:25
Are a pair of glasses with a camera
35:27
said does basically the same thing beacon and
35:29
it's take a photo and tell you what
35:31
you're looking at. I think about ah a
35:33
device like the rabbit ai which none of
35:36
us really notice any good yet but. Where
35:39
do you think devices that would evoke a
35:41
poker decks you know twenty years ago are
35:43
going next? What do you think we're at?
35:45
We might be carrying around in our pockets.
35:47
And ten years, that sort of of votes.
35:50
The. Classic Bucket x. Yeah. I
35:52
mean a smartphone is the obvious sure of is
35:54
parallel and I ask you to enter and I
35:56
don't think that's a form factor will be will
35:58
be leaving any time. When I'm.
36:02
An. Addict The Truth. As for
36:04
the vast majority of people, even if devices
36:06
like. Bad. A rabbit
36:08
mob. Do. A
36:10
hickey is really novel and really
36:13
sick assaults like a It's much
36:15
more likely that. The you
36:17
why are you acts Yes! Improvements that
36:20
that thing has are gonna have merged
36:22
back into the devices we already used.
36:24
Some I'm carrying around a rabbits ah.
36:28
But. I didn't that kind of fun of
36:30
center when I was saying earlier about this idea
36:32
of like home brewing devices and electronics becoming more
36:34
accessible because I i think. There. Is
36:37
a value proposition as we get
36:39
more connected to in a big
36:41
companies and clouds and all that
36:44
stuff they're they're becomes more value
36:46
to detach from. Matt. And.
36:48
Be like know my. You.
36:50
Know device I used to text my
36:53
friends is not something I bought to
36:55
up from Apple that's connected to Apple's
36:57
clouds. It's this you know, slightly esoteric
36:59
device but I know the only people
37:01
who can possibly read did this chat
37:03
log is me and my says who
37:06
also have this device or whatever and
37:08
I think we I would love to
37:10
see more stuff like that come up
37:12
more of that. Hobbyists,
37:14
Hacker culture. Because.
37:16
It is more accessible now and I think
37:19
we could get a lot of creativity and
37:21
and novel ideas. I think most people would
37:23
never get into that type of space, but
37:25
I do thank. You. Know it
37:27
would be. Feasible.
37:29
Turn to see people going down that route
37:32
more. we make it's medical innovation out of
37:34
at. but yeah, practically. I imagine we'll see
37:36
a lot more Smartphones and a few Vr
37:38
headsets are beyond that. The assisted by I
37:40
don't see any other huge wave coming down
37:42
the road, you know. I
37:44
think I think you put it well
37:46
that that hobbyist hacker side of it
37:49
is. You know I appreciate the Bleeding
37:51
Edge slick, polished thirty five hundred dollar
37:53
headset as much as the next guy,
37:55
but I'm definitely much more mostly in
37:57
destin what people are able to put
37:59
together. Yeah, and share it on a
38:01
youtube video. It's just it's I mean you
38:03
use the word warmth earlier. It's just there's
38:06
a warmth to it Yep. Yeah, it's the
38:08
electronics equivalent of like cottage core. There's a
38:10
reason people want to be
38:13
You know out in the woods on a fire and
38:15
it's because there is there's a warmth to it There's
38:17
a joy and I think All
38:19
these ai models and algorithms and stuff. They
38:21
they do Great in a lot of
38:24
ways you find a lot of stuff. That's really cool to them, but
38:27
They do lose some of that Organicness
38:30
and and I think it would be reasonable
38:33
to see a pushback and and see more
38:35
people hunting for that again Through
38:37
their electronics and the sites and the communities
38:40
and everything that they get involved with Yeah,
38:42
who is the musician the the sort of
38:44
like archetypal bonnie ver go into the woods
38:46
and record an album in a cabin type
38:48
thing That's the uh, it's the hacker equivalent
38:50
of that Yeah, definitely. It's
38:52
that same human thing of like
38:54
wanting small community and wanting small
38:57
You know connected feelings
38:59
versus being like part of Instagram
39:02
where you're like it's cool. I have an instagram account, but
39:05
It's this gigantic thing. I can't understand, you know
39:08
Well, maybe we wrap up there. So
39:11
in terms of hacking stuff together Computational
39:14
cottage core. Uh, what are you building
39:16
next? I mean, I
39:18
always got a hundred things going I
39:20
was just talking about the answer the other day and she
39:22
was saying like I don't know how you Move
39:25
all these projects forward and it's just You
39:28
know you get blocked on one and you jump to the
39:30
other you get upset about one and you do the other
39:32
one for a while So there's a
39:34
lot coming down the pipeline um You
39:38
know explorations of historical tech and
39:40
things that are now simple that
39:42
were very very difficult a few
39:44
decades ago, I think is uh
39:48
A trend I would like to keep working
39:50
on so similar to the pokedex how that
39:52
was not feasible But has become reasonable. I
39:54
think like that's cool. So i'm
39:57
doing explorations around 3d printing that
40:00
store data on them. So basically
40:03
creating my own version of CDs out
40:06
of 3D printed filament, or
40:09
doing similar
40:12
types of devices that are
40:14
cheap, low powered devices, but
40:16
customizable and have that warmth
40:18
to them, that nostalgia to
40:21
them. So I'm gonna be
40:23
doing a bunch more stuff with that
40:25
as well. But yeah, nothing specific to
40:28
announce, but there's always so many cool things
40:31
to build. The
40:33
list of ideas is pages and pages
40:35
long. So should be a lot of
40:37
fun stuff. It has 3D printed data
40:39
storage on it, that rules. Yeah,
40:42
if I get that to work, it's gonna be cool. Yeah,
40:44
that's really neat. Please send that to me if you do
40:46
it. I'm staring at
40:48
the 3D printed disc right now,
40:50
thinking about how it doesn't quite
40:52
work yet. But hopefully we'll
40:55
get that sorted. But yeah, that's that
40:57
same type of thing though, that physical
40:59
media, that feeling of, I
41:01
have my favorite movie in my hand
41:03
and all that stuff. It's that same warmth. And
41:05
I think that might be the overarching
41:08
thing of all my projects, is trying to
41:10
get technology back to being cozy
41:12
and fun again in that way. Abe,
41:15
thank you so much for taking the time to talk with me. I
41:17
love this project and the chat was a lot of
41:19
fun. Yeah, thank you. Thanks for having me. Thanks
41:22
for watching. I'll see you next time.
41:25
See ya. Bye. Bye.
41:27
Bye-bye. Bye. Bye.
41:30
Bye. Bye. Bye bye.
41:32
Bye bye. Bye bye. Bye
41:35
bye.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More