February 9, 2023
It’s been a little while, I hope that you remember signing up for the JetsonHacks Newsletter.
January was a busy month. Both December and January saw a redesign of the JetsonHacks website. If you have not already, stop by and take a look around.
There are four new articles on the website:
- jtop: The Ultimate Tool for Monitoring NVIDIA Jetson Devices
- JetPack SDK 5.1 Release Now Live!
- Jetson Orin NX 16GB Module Now Available
- Seeed Studio reComputer J4012 – Jetson Orin NX
The big news here is that the Jetson Orin NX 16GB is now available. The NVIDIA partners are busily developing carrier boards for the Orin NX boards. NVIDIA has a fun web page that tells you about the carrier boards and configurations for the Orin. Some of these are available now, or will be soon.
The JetPack release fixes a few outstanding issues, and adds a couple of features. The main reason for the release is to support the Orin NX. Many JetPack releases coincide with hardware releases.
NVIDIA at CES
There were several companies that won awards for their Jetson-enabled products. Here’s the NVIDIA blog post about the winners. It’s worth wandering over and see what people are working on. The companies:
- John Deere
- Seoul Robotics
The JetsonHacks Blog
One of the new features of the JetsonHacks website is the new blog section. The blog section is for keeping track of what’s going on in the JetsonHacks world. As you know, JetsonHacks publishes articles with corresponding videos.
The videos require a large time investment to create. This tends to limit the amount of content for publishing. The idea of the blog is to add some accountability. It’s a little bit of the “Show Your Work” mentality. The blog tells you what I’m working on for JetsonHacks, along with some other ideas. It’s not a blog post everyday, but a few times a week.
Jim’s Shelf of Books, Ideas and Other Curiosities
Another new page on the website is “Jim’s Shelf of Books, Ideas and Other Curiosities“. I like collecting ideas. There are all sorts of ways to find ideas. There are a lot deep, well thought out ideas in books. When an author spends years writing and researching a subject, there is bound to be insight. So I’ve read a lot of books.
In the last 20 years, a lot of ideas began wandering out online. Some of the ideas are deep, but don’t tend to be completely formed. There are some people who have web sites containing a treasure troves of ideas. There are videos out there too.
The idea behind the shelf is that of a gathering spot for some of the ideas that I find attractive. Or that influence my thinking. Some of the ideas I first encountered many years ago, and they stuck. Others are new. For books, I’ll write a little review about the interesting parts. For the other resources, I hope to point out the good bits.
Right now there are three books, but that will grow over time.
Jetson Orin Nano
The Jetson Orin Nano should start shipping later next month in March. The Orin Nano slots into where the Xavier NX sat previously in the Jetson lineup. NVIDIA provides the carrier board design for the Jetson Nano and Jetson Xavier NX. My understanding is that to take full advantage of the Orin Nano, the design changes slightly. The partners will be able to do this fairly easily.
An interesting aspect of the Orin NX and Orin Nano is the elimination of eMMC memory on production modules. A NVMe SSD replaces the eMMC. This should fix one of the biggest complaints, lack of drive storage. A very large number of developers were having issues trying to fit their apps onboard. There is a cost tradeoff, but this should be a beneficial feature going forward.
I wrote up an experiment in a blog post. One of the skills/tools that the people will need to incorporate into their work flow is AI. AI is a very broad term. The one with the most noise around it right now is ChatGPT. You can read about how my experiment went in the blog post.
Alan Kay famously states that “Point of View is worth 80 IQ Points“
Let me share something here that you won’t see on the JetsonHacks website/YouTube channel.
In 1978 James Burke presented a TV Show “Connections” on the BBC. I believe that the show is available on DVD from Netflix if you have that service.
Connections is fabulous. It answers questions like “How did a test of gold’s purity in 500 B.C. lead to the invention of the atomic bomb?” Tying together different inventions, here are stories of how people make something new.
I vividly remember one episode which goes over many of the inventions that had to come together to form the automobile. A device intended to detect malaria (it didn’t) became the basis for the spark plug. Or how the perfume atomizer became a carburetor. The ideas and inventions sometimes span hundreds of years. But the sum is greater than the parts. The right person or group of people takes these and puts them together, magic happens.
A lot of the time it’s not who you would think would be the inventors. It’s disputed who made the first flying machine. But certainly you would not expect guys from a bicycle shop (the Wright brothers) to be in the mix. But history says otherwise. Many times it is the people you do not expect.
It turns out a lot of creators and inventors are people who are multi-discipline. This gives them the ability to take from here and there with a unique perspective on what may be possible.
A lot of people are saying to put the AI genie back in the bottle. Make it illegal. That’s nonsense, of course. Broadly, AI is computational statistics. How would you go about outlawing or have the government license it?
The underlying LLM for ChatGPT (GPT-1 to GPT-3.5) has been around for a while. It has been publicly available for people to work/play with. Of course the people at OpenAI have been using it for a long time. Upon release of the chat interface, OpenAI thought of it as an incremental improvement. Sam Altman, CEO of OpenAI, said they were taken a back by the tsunami of interest after the ChatGPT release.
This is one of those inflection points. The idea is at critical mass. It does not take much of an imagination to know that quite wonderful and totally unexpected things are coming. Or the end of the world as we know it. Depends on if you have your optimism glasses on.
It is clear to everyone. It’s not like crypto currency, NFTs, or block chain. When you explain those to someone, they look at you like you’re from another planet.
Not so with ChatGPT. Everyone knows what to do with it, almost intuitively. Need a paper written? Done. A program written? Done. A note of condolence? Done. And so on.
The Take Away
With that all said, that’s not my only take away from the story. All too often developers create a tool which is nearly useful. This is especially true of programmers. It’s what I call the “Johnny Genius” syndrome. Johnny is a genius, and he assumes that everyone else is too.
Something that Johnny creates works in the strict sense, but is not easy to use. Johnny might just take a little pride in that. The command line options, configuration files and their location, commands and such are all at Johnny’s fingertips. When mere mortals try to use it, they get lost. Quick.
When Johnny hears the criticism, he tends to fix things by adding more features. This complicates things further. The cycle repeats until someone steps in to put a proper user interface in place. Does this sound familiar in the Jetson world?
Note the parallel to the ChatGPT story. The underlying GPT model is not different. There are other large models like GPT. Yet the interface to the model is different. It’s the enabling part. It’s now easy to imagine this building entire industries.
The user interface is more important than a lot of people think.