What’s it like to try and build your own deep learning workstation? Is it worth it in terms of money, effort, and maintenance? Then once built, what’s the best way to utilize it? Chris and Daniel dig into questions today as they talk about Daniel’s recent workstation build. He built a workstation for his NLP and Speech work with two GPUs, and it has been serving him well (minus a few things he would change if he did it again).
You likely already saw this, but I don’t even care because I have to link to it because it is so freakin’ cool!
We’ve never been shy about borrowing a good idea. Which brings us to Raspberry Pi 400: it’s a faster, cooler 4GB Raspberry Pi 4, integrated into a compact keyboard. Priced at just $70 for the computer on its own, or $100 for a ready-to-go kit, if you’re looking for an affordable PC for day-to-day use this is the Raspberry Pi for you.
Steven Fuchs loves his Sonos, but…
While it is the radio of the future, our most common usage is as the radio of the past. We tend to tune it to one station and leave it there. By far, our most common interactions with the system are changing the volume and pausing/playing the music, often creating scrambles to find a phone to turn down the volume in order to answer a different phone. What we needed was an analog interface to this digital system that was always at arms reach.
Hackers gonna hack. Steven reached for Elixir and scratched his own itch with this very cool little hardware project. Here’s a demo video of it in action.
The Analog Terminal Bell isn’t for sale, but that isn’t going to stop Aaron “Tenderlove” Patterson from trying to sell it to you in this epic infomercial. “It even works with /dev/urandom” 😆
The Nerves Keyboard project is a small group of enthusiasts using the IoT tooling from the Nerves project to build a mechanical keyboard that can be programmed and customized with Elixir. The work happens in the open and is currently moving towards the hardware stage. This is a quick getting-started tutorial.
Max Braun thinks today’s webcams are boring, so he brought back a classic. Max took an Apple iSight and retrofitted it with a $5 Raspberry Pi Zero, which “fits the iSight’s dimensions almost perfectly.”
The PiSight actually works like you’d expect it to. Just plug in the USB cable and the camera will show up in your video conferencing app of choice. The image quality is quite good, possibly better than the built-in camera of today’s MacBooks.
The best part is you can do this too because Max made all the plans available as open source.
Just in case you’re not completely taken aback by the absurdity of this project and are now considering building your very own PiSight, rest assured that I’m making everything available as open source.
The GitHub repo has a list of parts and where to get them, the 3D-print-ready model of the frame, and the source code. I’m thinking it should be possible to get the total cost down to under $150. I had to spend a bit more than that because I needed to experiment and opted for higher-end materials.
An ambitious attempt to create an open source device for reading. But why?
As a society, we need an open source device for reading. Books are among the most important documents of our culture, yet the most popular and widespread devices we have for reading — the Kobo, the Nook, the Kindle and even the iPad — are closed devices, operating as small moving parts in a set of giant closed platforms whose owners’ interests are not always aligned with readers’.
It’s still early days, but the project got a boost of support by winning Hackaday’s Take Flight with Feather contest in January.
Find yourself stuck indoors during a pandemic? Why not build an open source settop box and connect to the only microcontroller powered video streaming service?
If you haven’t heard of the ESP32, it’s a low-cost/low-power SoC with Wi-Fi and Bluetooth integrated. So this is like your own little open source Amazon Fire Stick or Chromecast.
OK so maybe you don’t use a standing desk… like I do. OR maybe you do, but you don’t use the IKEA Bekant desk in particular… like I do. STILL you can appreciate how hacker it is that someone built their own drop-in controller to add memory positions to the Bekant… right?
I wanted to have memory positions for easily switching between various work positions. I also didn’t want to be limited to just 2 positions. However, as I went through the process, I realized the hardest part was designing the enclosure. 3D Printing is a great option, but lacks that professional look, and limits the availability to those with printers. Additionally, getting custom membrane buttons that would look good was also extremely expensive. Simple push-buttons would take away from the look of the desk.
By targeting the factory enclosure, it keeps the original look and robustness, while adding functionality.
See it in action right here.
According to my research among programmers, 43% are still using monitors with pixel per inch density less than 150…
Why is this a problem? Because the only way to get good letters is by spending more pixels per letter. That simple. In the past, the displays’ pixel count was small, so we learned to live with that and even invented some very clever tricks to make our lives better.
Nikita goes on to share more details of how text looks on a low-resolution display vs a retina display. I’d love to see a follow up poll of the 43% using 150 PPI or less monitors on “why” they haven’t made the move to retina yet.
On the heels of NVIDIA’s latest announcements, Daniel and Chris explore how the new NVIDIA Ampere architecture evolves the high-performance computing (HPC) landscape for artificial intelligence. After investigating the new specifications of the NVIDIA A100 Tensor Core GPU, Chris and Daniel turn their attention to the data center with the NVIDIA DGX A100, and then finish their journey at “the edge” with the NVIDIA EGX A100 and the NVIDIA Jetson Xavier NX.
The role of a father plays a pivotal role in a child’s life. Ian Bernstein is a former Founder of Sphero and is now the Founder and Head of Product of Misty Robotics — they’re building the first programmable robot for the home and business. It’s called Misty II. The journey of building Misty II started when Ian was 5 years old and his dad bought him an Apple IIe.
Daniel and Chris have a fascinating discussion with Anna Goldie and Azalia Mirhoseini from Google Brain about the use of reinforcement learning for chip floor planning - or placement - in which many new designs are generated, and then evaluated, to find an optimal component layout. Anna and Azalia also describe the use of graph convolutional neural networks in their approach.
John Cassel from The New Stack lays out the quiet-yet-effective push toward open source hardware. We first heard about RISC-V from Ron Evans on Go Time. He was very excited about its potential, saying:
it’s an open source set of silicon designs, so that you can build your own custom chips the same way that we’ve been able to build our own custom operating systems; either pieces of Linux to create their own Linux distros - we’ll be able to do the same exact things with custom silicon
Why a rotary cellphone? Because in a finicky, annoying, touchscreen world of hyperconnected people using phones they have no control over or understanding of, I wanted something that would be entirely mine, personal, and absolutely tactile, while also giving me an excuse for not texting.
This reminds me a bit of the finding your analog conversation we had at the end of The Changelog #378. What’s super cool about this phone is that it is super functional!
So it’s not just a show-and-tell piece… My intent is to use it as my primary phone. It fits in a pocket.; It’s reasonably compact; calling the people I most often call is faster than with my old phone, and the battery lasts almost 24 hours.
We partnered with Red Hat to promote Season 4 of Command Line Heroes — a podcast about the people who transform technology from the command line up. Season 4 is all about hardware that changed the game. We’re featuring episode 1 from season 4 — called “Minicomputers: The soul of an old machine.” This is the story of Minicomputers and how they paved the way for the personal computers that could fit in a bag and, eventually, the phones in our pockets.
Learn more and subscribe at redhat.com/commandlineheroes.
I’ve had never really come into contact with hardware programming, working mostly in python or C#, until a friend of mine asked me for some help with programming a simple controller for RGB strips using Arduino Nanos.
We’d, of course, fail spectacularly.
Not only did our hardware not work quite like intended and a few Nanos died in the process(but that’s a story for another time), but I actually learned a lot from this and similar projects.
And I want to tell you some of my mistakes, what I learned by making them and how to prevent them.
George Hilliard is an embedded system engineer who has designed a cheap business card that runs Linux and has a USB port. In this blog, he talks about the logistics and design of the card.
Daytripper is a laser tripwire that can, upon triggering:
- Hide all your windows
- Lock your computer
- Execute a custom script to do whatever you want!
Pick one up on Tindie.
CutiePi is a good name for this device. It sure is cute!
We believe in open source, and we believe people should have control over the technology they use. Everything you see here is open source – schematics, PCB, drivers, firmware, UI, everything.
It’s still early (no pricing, for example), but they’re shooting for a release before 2019 is out.
It astounds me how much bang you can get for 35 bucks today.
Colin Billings is the founder and CEO of Orro where they’ve built the first truly intelligent home lighting system. It knows when you’re in the room, and adjusts the lights automatically for you. But Colin’s path to starting this company wasn’t a straight line. Like most innovative products, Orro has an interesting beginning — after-all, they’re going up against the giants.
Google, Intel, and others have recently been targeting AI at the edge with things like Coral and the Neural Compute Stick, but NVIDIA is taking things a step farther. They just announced the Jetson Nano, which is a $99 computer with 472 GFLOPS of compute performance, an integrated NVIDIA GPU, and a Raspberry Pi form factor. According to NVIDIA:
The compute performance, compact footprint, and flexibility of Jetson Nano brings endless possibilities to developers for creating AI-powered devices and embedded systems.
And it’s not only for inference (which is the main target of things like Intel’s NCS). The Jetson Nano can also handle AI model training:
since Jetson Nano can run the full training frameworks like TensorFlow, PyTorch, and Caffe, it’s also able to re-train with transfer learning for those who may not have access to another dedicated training machine and are willing to wait longer for results.
Check it out! You can pre-order now.
These IPUs (Intelligence Processing Units — a term new to me) with visual design by Pentagram for Graphcore are really pretty. Also, I think the tech may be cool but it’s a bit over my head so maybe you can tell me?
Here is their brief spiel:
Our IPU systems are designed to lower the cost of accelerating AI applications in cloud and
enterprise datacenters to increase the performance of both training and inference by up to
100x compared to the fastest systems today.