Quantcast
Channel: JetsonHacks
Viewing all articles
Browse latest Browse all 339

JetsonHacks Newsletter December, 2022

$
0
0

12-28-2022

My intention in starting the JetsonHacks Newsletter is threefold. First, cover NVIDIA Jetson events and announcements that are significant. This is to either amplify significant developments or announcements. Or to bring attention to offerings that may not fill an entire JetsonHacks video or article.

Second, add some of my own personal insight into computing environments. This can be hardware; Jetson in robotics for example. This can be software; deployment of machine learning vision models at the “edge”. Or it can be something a little more over-arching, like trends in computing.

Third, I want to start getting a better feel for the Jetson community at large. By understanding the needs, I can make articles and videos to better benefit everyone. 

The biggest news is the reintroduction of the Jetson Nano Developer Kit (4GB). USA availability has been pretty tight, though production has been rapidly ramping up. In Europe, checkout Silicon Highway , in Asia checkout Seeed Studio for availability. At the very least, you should be able to back order Nanos now and actually have them delivered in a reasonable time frame. 

Over the next couple of months of 2023, we’ll see the Jetson Orin NX and Orin Nanos begin to ship. The Orin range replaces the Xavier range. There are more models of Orins, and cover a wider range of capabilities than the Xaviers. All at similar (or better) pricing and more performance.

With that said, I’ll go completely afield. At the end of November, 2022 OpenAI introduced ChatGPT.

My take is that you should believe everyones opinion about ChatGPT. It is gobsmackingly jaw dropping. The initial corpus that the chat bot works against (GPT 3.5) is vast. The next iteration of the underlying model, GPT 4, is an order of magnitude bigger. 

The chat bot can generate and write code, songs, poetry, blog posts, scripts. And it can do it “in the style” of different human authors. It can summarize books, movies and articles. Its potential is out in the open for all to see. So much so that 1 million people signed up to use it in one week

Sign up and use it to get a better grasp on what it is about. It’s impressive in the extreme. It’s easy to see why business people believe that it’s the next gold rush. Look at this article: AI Homework by Stratechery. 

To be clear, this is unprecedented. To have that many people using a research prototype, and exploring its uses in that time span? Wowie.

But here’s the thing. ChatGPT is impressively correct or confidently wrong. Open AI is very clear about this. It’s way more human sounding than it has any right to be. We’ve all met people that are bullshitters, ChatGPT takes that up a notch. Like many bullshitters, it’s also likable. Here’s a good article, “ChatGPT is a bullshit generator. But it can still be amazingly useful“.

It easy to imagine how you can use ChatGPT for good. At the same time, it’s easy to see why it’s way evil. Two sides of the same coin.

Back in 1987 Apple made a video, “Knowledge Navigator“. It’s a quick 5 minute watch. At the time, an Apple Macintosh had 512K memory and a 800K floppy disk for storage. The Mac did not have a color screen. There was no web yet. And so on. Machines that were not as capable as many of todays microcontrollers. Yet Alan Kay, an Apple Fellow at the time, foresaw what computing would/could become. Here we are 35 years later, and it’s easy to see that most of it is here.

What Kay didn’t include is the idea of “fuzzy computing”. Machines that appear to work, but may not generate a correct answer. Or the idea of a correct answer given “a certain point of view”. Or ideas with no attribution or source material. This all seems like the source of science fiction nightmares, like HAL in 2001: A Space Odyssey. That’s alarming. 

Of course, the corpus of the underlying LLM influences the answers of the chat bot. You can easily imagine influencing the answers in a political context, social constructs, or straight propaganda. You can bet certain people will also want to control the spigot to control “misinformation”.

People expect that “The AI” is not biased towards any one perspective. It is not clear that people can build an underlying database that would reach that goal. Or want to. Especially given the behavior of some recent Silicon Valley companies.

Gary Marcus calls all this a “Jurassic Park Moment“. Like I said, believe everyones opinion. A great opportunity? Helping humanity? Unleashing the devil? Think about it, it’s really interesting.

Many people ask, “Will ChatGPT take my job?” People not like us do not understand that new technologies changes things. That is their purpose, but they are only a tool. The purpose of tools is to amplify. To make the wielder more powerful, more precise. To offer means of manipulation that are otherwise difficult or impossible.

Here’s an example from the distant past. When the first personal computers starting making their way into the business world in the early 1980s, many people worried that they would be out of a job. Typists, secretaries, clerks, management, and so on. For a scale of time, it took about 15 years before the most obvious casualty, the typewriter, went by the wayside. 

As you know, computer automation multiplied the number of documents produced. People began building spread sheets, PowerPoint presentations, and a myriad of different applications. None of this is an obvious consequence, except in hindsight. Nor is the innumerable number of companies and entire industries that sprang up in support. In a broad sense, specific tasks became even more specialized. Instead of extremely narrow definitions of a job (e.g. typist), people now face more variety in their work. They use computers as tools to enhance their skills. Everyone types now.

It is naive to think that over time you will always have the same job. Jobs change. Or they may be eliminated. That is not comforting. However, it is how you react to change that defines you. Ask any stoic. Being in technology, we are always aware that we need to keep up. Lest we be left behind.

——

Thanks for taking the time to read this newsletter. I am looking forward to doing more writing in 2023. The introduction of the new Jetson Orins is certain to open up quite a few new opportunities. There has been huge investments in machine learning models and tools recently. The Orins should be able to bring them to places where it would be difficult for most hardware to reach.

I hope your 2023 turns out really swell. Happy New Year!

Jim

The post JetsonHacks Newsletter December, 2022 appeared first on JetsonHacks.


Viewing all articles
Browse latest Browse all 339

Trending Articles