System76 Transmission Log

Join Alex and Emma as they discuss the latest System76 community, hardware and software news. Also, listen to Shantanu, machine learning specialist at Trossen robotics, talk about the company's innovative machine learning robot kits and how they use System76.


00:36 Alex at All Things Open conference
02:13 Factory Tours and meetups
03:32 Thelio Astra at ATO
05:04 What's new in COSMIC
07:06 Begin interview with Shantanu from Trossen Robotics
07:30 Learning about Shantanu's job and main projects
10:49 What excites Shantanu about robotics industry today
11:27 Aloha Unleashed
15:02 Standout robotics projects
18:18 Using the Serval WS with Aloha Stationary V2
20:04 Meerkat and Trossen join forces
21:34 Experience with System76 products and support
22:39 Emma and Alex play a game

What is System76 Transmission Log?

Welcome aboard the System76 Transmission Log. The space station orbiting the Linux computer manufacturer, System76: home to handcrafted Open Source hardware and the makers of Pop!_OS.

3, 2, 1, 0.

Welcome aboard the System

76 transmission Log.

Our broadcast is about to begin.

This is the latest on System 76
computers, manufacturing, and pop OS.

Now for your in-orbit crew.

Hey, Alex.

Tell me about your experience
at All Things Open.

It was my first time.

I was helping work the booth with Stetson
and Aaron, who are from the sales team

and support team, respectively.

It was pretty fun.

We had a lot of attention at the booth.

We were the only consumer hardware company
that was there that I'm aware of.

We had a lot of excitement.

Ato was raffling off
three of our in laptops.

We got a lot of questions about that.

We were able to give the spiel
about System 76 and our open-source

mission, our Denver manufacturing.

It was great to educate
the enterprise masses about us.

Cool.

I feel like we were here
recording just yesterday.

The month flew by pretty fast.

Yeah, it's almost Thanksgiving.
It's crazy.

Yeah, and when it's close to Thanksgiving,
we celebrate our birthday

because System 76 was born
or founded on November seventh.

We celebrated our 19th birthday
this month, and we had a feast with cake

and presented Carl with a nice canvas
that had a company photo on it.

Now he can see our smiling faces
on the wall every single day.

Yeah, walking into the office
is like walking into grandma's house.

You see your photo from
fifth grade up on the wall,

and you're like, Oh, good times.

Thanks, grandma.

Grandma Carl.

I like when we go on tours and
we stop by the photo wall on the way out,

and I can tell stories of each photo.

That's really fun.

We have some meetups
coming up, too, as well.

We go on factory tours during those.

There's some catered food and drinks.

We play some games.

We get to chat with the community,
chat with locals.

It's a good time.
The next tour, I believe, is in December.

Is that right?

We have our December December
and January meetups posted on meetup.

Com.

What would you say is
your favorite part on the tour?

When we get to the Blue Laser.

Why the Blue Laser?

I can't tell you.

Hang on.
Secrets.

You have to go on the tour to find out.

Okay, well, I'm going to
schedule a tour after this podcast.

How do I schedule a tour
after this podcast?

You have to email me,
emma, E-M-M-M-A@system70.

Com.
6.

Com directly, and you can schedule it, or
you can come on one of our group tours.

I will email you to get those secrets.

Okay.
All right, let's get into hardware news.

We talked in-depth about Thelio ASTRA last
month, and there have been some

benchmarking results released by Foronix.

Those show some promising results,
and you can check it out on foronix.

Com if you want to get some details there.

Yeah, we featured Thelio ASTRA at ATO.

People were excited to talk to us
about our new Arm 64 workstation.

It's an Ampere processor up to 128

cores, up to 6000 data graphics.

So beast of a machine, beautiful looking.

We were right next to the arm booth
at the conference, too.

So that was funny.

Sort of a funny coincidence.

And yeah.

Thaleoastra has really
been making the rounds.

We're going to be at CoupaCon coming up
pretty soon here, and there'll be

four Thaleo Astras at different booths
doing different demos at each booth.

That's cool.

We'll have more details
about that posted on our blog.

Another newcomer
to the Thaleo desktop line is Thaleo

Spark, our AMD-powered starter desktop.

It is an energy-efficient work of art
that fits perfectly on any desk.

You may remember Spark from
the first Thaleo Head music video.

It was on our minds, and it still is.

Now it's an AMD 8000 G-Series system and

lower starting price at, I think, 799.

Beautiful design, good starter desktop.

If you're starting something new
or just want something fast

that fits within budget.

New desktops are great,
but what about desktop environments?

What's new in cosmic?

You got to talk about
the environment, Emma.

So cosmic is a vast
environment of possibility.

We are now in Alpha 3.

Very exciting.

We have some changes to cosmic
files, a lot of smoothing over

with the preview mode and gallery mode.

Preview mode is no longer
going to be on by default,

and there's a new footer widget.

So When you click on a file, the button
isn't going to change locations.

It's just a lot neater.

Then Cosmic Store now has
a made for cosmic section.

You're going to see some nice cosmic apps
made by you guys, mostly.

Things like a tasks manager
for your to-do list, a weather app.

I think I saw a Pokédex in there.

Someone is out there speaking my language.

I'm just excited to see how long that list
is going to be by official release.

It's pretty cool.

Then lastly, we have screen reader
support in initial testable state.

It's not going to be perfect,
but we're interested in hearing

your feedback in the first testing
of accessibility in Cosmic.

Yeah.

If you want to give it a try,
go ahead and pop on over to system76.

Com/cosmic,
and you can download the Alpha ISO.

I can't forget to mention we
have a lot of bug fixes in Alpha 3.

Thanks to you guys.

You're doing great work
on these contributions here.

It's really feeling like
a polished beta to me at this point,

and we appreciate the support.

I'm so excited for our guest this month.

Who is our guest this month?

We have Shantanu from Traslin Robotics
to talk about using System76 hardware

and open source to create cutting edge
robotics solutions for research labs

and educational institutions.

That sounds awesome, and
I think we're going to learn a lot.

So let's get into it.

Thanks for joining us today, Shantanu.

It's very exciting to have you
on our podcast today.

This is Shantanu from Tracin Robotics.

Can you introduce yourself and tell us
a little bit about what you do at Tracin?

Hi.

Thank you so much for having
me on this podcast.

My name is Shantanu, and
I am a machine learning specialist

here at Tracin Robotics.

I work with the Aloha Kids.

We develop Aloha Kids as a research
product for engineers and researchers so

that they can easily set up their
experiment and get up and running with any

task that they have in terms of machine
learning, especially imitation learning.

So my goal here at Trouzen Robotics is
to make it as easy as possible for people

and researchers to get started
with machine learning,

may it be in terms of setting up their
hardware, software and making it

as easy as possible to use so
that they don't have any difficulty

setting up all these critical
functionalities in machine learning.

So, yeah, that's my job here.

Could you talk a little bit
about the Aloha kits and what those are?

Yeah.

So the Aloha kits started as
a collaboration between Stanford

University and Truss and Robotics.

Our researchers at Stanford University
were looking for a low-cost, affordable

product that they can use to develop
machine learning for their research.

So they collaborated with Truss
and Robotics to create these products

that are today known as the Aloha Kids.

Essentially, these are a leader-follower
arm pair in which you will be having

a Widow X arm and a Viper X arm.

It's essentially used for
teaching a robot how to do stuff

using bimanual manipulation.
Education.

As a human, it's easier for us to
look at something and learn it quickly.

Maybe it might be opening
up the door, doing something

as easy as tying a shoelace.

But for robots,
this is a complicated task.

In order to teach a robot how to do stuff,
there are multiple different ways,

like reinforcement learning,
writing heavy code that tells a robot

what traject it has to do.

But as an easier solution, Imitation

learning is something which is a go-to

in terms of machine learning landscape.

What people do is they use the leader arms
to tell the robot how something

has to be done, like tying a shoelace.

Once you collect enough episodes,
like maybe 50 or 100 episodes

of this task, you can train it
using machine learning algorithm.

Initially, it was act, which is
action-chunking transformers.

Once the robot knows how to
solve this problem, it will be able

to do it autonomously without any help.

There are multiple papers
that are published recently in conference

going on at Munich, Germany.

There are papers which are also
deemed as the foundation models

for the growth in the machine learning
landscape that we see right now.

I can keep talking about this
all night, all day, but I guess

that answers your question in terms
of what the Aloha kit can be used for.

It's basically like an all-in-one
kit for machine learning?

Yes.

With the robot arm?
Yes.

Cool.

Some of our listeners might be roboticists
or enthusiasts about robotics.

What in the industry
excites you most about robotics today?

For anyone in the robotics industry
or even those people who are not in

the robotics industry right now, people
know the hype about humanoid robots.

There has been a rise in the companies
that are producing humanoid robots

like Figur, Tesla, and Google.

I know that these It seemed like
the future is near, but the amount of work

that goes behind these robots to perform
simple tasks, the way I said it,

tying a shoelace,
it's top of my mind because we just saw

the release of a paper called as
Aloha Unleashed, which was

published by Google DeepMind.

If you see the core of the paper
is they used the Aloha kit

that we provide, and they achieved
a very simple task of tying a shoelace.

But for a normal person,
this is a very easy task,

and no one knows how complex it is.

But when you read into the paper,
you will see that to train the robot to

do that task, it was so complicated that
it required immense amount of GPU power.

It required immense amount
of training episodes.

It took about 11 days for
it to train to do that simple task.

The most exciting part for me is we
see these robots that are going

to work in the industry, in factories
and other places, and they'll be soon

collaborating with humans.

But the fundamentals, training
these robots to do something, starts

at very basic, that's the Aloha Kids.

That's me being a part of it and me
being able to contribute to the community

in a way where it becomes easier
for people to use these robots

is something very exciting for me.

That's awesome.

Can you tell us a little bit about
the Interbotx open source repositories?

Yeah.

So Interbotx is why we
call it open source.

We don't have any control roll
over the software in terms

of we don't sell the software.

We just are a hardware company.

The reason why we create the software
is that people should not

face any difficulties
while interfacing their robots

with the computers they are going to use.

They need something which is very sleek,
easy to use, and they don't

spend a lot of time debugging stuff.

What intervallex does is it gives you
a middleware framework, which is a layer

of ROSS, which is an existing robotics
operating system framework, which

is widely popular among roboticists.

We use this communication protocol or
the middleware framework,

build upon it using our Inovatex software
and make it as easy as possible for people

to calibrate their robots,
run their robots, debug if anything goes

wrong in terms of maybe servos,
not picking up the current correctly or

the commands correctly or
servos dying in the middle of operation.

We provide elaborate workflows
that will help them to debug anything at

any point in time and get up and running
with their robots in about an hour after

they have set up the robot correctly.

Yeah, that's the software that we offer.

It's pretty easy to use.

It uses something which is called
ROS, which our robotists

are highly familiar with.

We offer C++ as well as Python binding,
so it makes it easier for you as a user

to choose whatever you want.

And because we support Python as
well, people who build upon it can easily

integrate machine learning concepts
or deep learning concepts into

their projects without worrying about
the software that run behind.

It sounds like we're vastly reducing
the barrier entry for robotics here.

That's awesome.

Have you seen any standout projects yet
from the community?

What are your highlights, would you say?

There are multiple things that are going
out in the community right now

because as you know, there is
a conference on reinforcement learning.

That's the highly anticipated conference
in terms of reinforcement learning

at Robotics that's going
on right now at Munich.

When a conference is close by and people
will be releasing all their work

that they have done throughout the years.

We have noticed that at least five
of the top papers in this conference

use the AloHA kits, use our hardware,
our software to some extent.

There are big companies like
Google DeepMind just least the paper

called as Aloha Unleashed,
which I was talking about before this.

We also have seen a company called as Pai,

which is physical intelligence, that just

announced its own foundation model
that uses Aloha Kit again.

We have seen, I guess, there are
three papers from a researcher from UC,

Berkeley, along with researchers
that are leading Google DeepMind as well.

The core of this paper is to
train something, to train a robot to do a

very simple task, you need a lot of data,
and you to collect a lot of data sets.

To reduce the amount of data
that you collect, people have come

up with projects that let you
use any data on any other robot.

It's something called
as a cross-training platforms.

There are a lot of very interesting
papers, interesting researches

that are going on in the industry.

In terms of what I was saying,
the research papers that are coming out,

and they are coming out so frequently,
and I guess to some extent,

it's because of the way we are reducing
the barrier to entry with our products,

making it as easy as possible
to set up your robots.

We also provide with a community
that helps you publish your data sets,

publish your trained models to a platform
where you can pick the data set

from and train your own models.

So what happens is researchers now
don't have to spend their time

on setting up a hardware platform.

They can invest more time
in coming up with new ideas, new models,

and scale up their research
to an extent that they don't have to

worry about anything in the background.

I guess that's where cross
and robotics helps the researchers.

That's our motto to help researchers
or help innovators innovate.

That goes a lot with our model
of engineering the computer, too,

is just trying to get all the fussing
around out of the way,

all the installing drivers,
moving windows around, things like that,

to just get people focused on what
they actually want to work on.

Now, you're using the Serval workstation
for the Aloha stationery, too.

What led to the decision for the Serval?

As you said, your model
and our model is similar.

The reason is when I I
started using this workstation.

To begin with any project, you need
to install a lot of dependencies, a lot

of different things to interface
your robots with your workstation.

I worked with different...

When I was a student, I worked
with different laptops, and I have seen

other people work with different laptops
and the difficulties that they face while

installing simple things like Linux,
Cuda drivers, or any other thing.

But with this workstation, when I started,
it was within an hour, I was able

to set up all the necessary dependencies.

It took an hour because downloading
NVIDIA code takes an hour.

It wasn't the system
doing something weird.

But yeah, it was a very smooth and error
It was a free experience because most

of the hardware is compatible
with the NVIDIA drivers that you need.

It was as simple as running a command line
argument, like seeing a pseudo

install NVIDIA server.

So It was that simple.

The Linux operation and
the Linux setup was very easy.

I guess that's the reason why we are
partners with System76 to get our laptops

so that the users that we give these
laptops to, the researchers,

don't have to face any other difficulties
that you might face when you get a laptop

which doesn't support the necessary
drivers or the hardware interfaces.

I guess that's the core reason
why we go with system 76.

Awesome.
We're happy to hear that it's going well.

I heard the MeerCAT was
going to be in use soon.

Do you know about that project coming up?
Yes.

The mini PC?
Yeah.

As you see, the Aloha stationery
is a bimanual robotic system.

We had noticed that in the research
industry, people are also

working with single-arm devices.

They don't need two arms
to achieve or complete a task.

So what we noticed was people
required one R interface.

And if we are providing a cheaper
alternative to the Aloha, we also need

something that is as compact as possible
so that people can move around the robots

easily, record their episodes.

And we were looking for a system
that can be useful for this.

And then we saw the System76 MeerKAT
Mini, and it had super processing powers

and interfaces.

And Again, because it is System76,
Linux drivers, Cuda drivers,

and everything was super easy to install.

Given that and how portable
it is, we are soon going to launch

a product called as Aloha Solo.

That that will be using the Meer cat mini.

That's really awesome.
Do you have a timeline for that yet?

It should be out by December first,
officially.

Okay, cool.

Are there any other things that
you'd like to add about trust in robotics

or about the robotics Linux industry
and Linux in general?

Yeah, I'm happy about this partnership.

We are so satisfied that we
have partners like you.

The great parts being the customer
support, the customization that you guys

provide, and how easy it is to get
something resolved in terms of anything

that we need apart from the product,
the support in terms of getting a product

in time, sending it to the customer,
having this beautiful a stream line

process where the customer orders
the product, you get it in time,

and we are able to get it
to the customer in time.

There is absolutely no error in between
because we are able to get the software

and the hardware integrated with System76
with so much ease that it feels

like it's one product itself.

It doesn't look like there are
two different companies working on it.

It just feels like it's
a one full company.

I guess that's the best part
about working with you guys.

It's definitely what we go for here,
the the whole experience.

Well, thank you so much, Shantanu.

Thank you so much.

All right, let's play a game.

A game?

What's the game?

I'm going to play a sound, and it's a
sound that you would hear in our factory,

and you have to guess what the sound is.

Okay.

I haven't won one of these yet,
so let's see how this goes.

Okay.

It can be anything-It's not
another USB stick, is it?

No.
Okay.

It could be anything that you
encounter at the System 76 factory.

All right.

Here we go.
Anything I can encounter.

Any guesses?

Mostly, I just heard Nathaniel and Aaron,
so it feels like when we used to

put the switches and keys in by hand.

I didn't really hear any
machinery in there.

There were some fans, but...

You're totally right.

You win.
Oh, really?

It win.

When we used to install keycaps
by hand on the launch keyboard.

Oh, my God.
I got one.

You win.

Yay.

All right.
Well, that's it.

That's all we We have for you
this month on the transmission log.

So thanks for joining us,
and we'll see you next month.

We will see you next month,
and thanks for listening.

Three, two, one.

This has been the System

76 transmission log.

For more inspiration, check out the
website and follow us on social media.

On your descent back to Earth,
please keep your hands and feet

inside the transport beam at all times.

Captain, sign off in transmission.