GitHub The Download, 15 Jul 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: The Download: GitHub Copilot Goes GA, Deno Fresh, GPT-3 Code Explanations and more

Description

On this episode of The Download, Christina returns from some downtime with a random t-shirt and the latest developer and open source project news, including:

0:00 Intro
0:55 GitHub Copilot GA https://github.blog/2022-06-21-github-copilot-is-generally-available-to-all-developers/
2:05 Explaining Code with GPT-3 https://simonwillison.net/2022/Jul/9/gpt-3-explain-code/
3:35 Deno Fresh https://fresh.deno.dev/ and Fresh on GitHub https://github.com/denoland/fresh
4:33 Virtual Buddy https://github.com/insidegui/VirtualBuddy
5:55 **Christina's Pick of the Week:** Commodore 64 TensorFlow https://github.com/nickbild/tflite_c64 and YouTube video https://www.youtube.com/watch?v=D3Ii4mchgnA

Presented by: @ChristinaWarren

As always, feel free to leave us a comment below and don't forget to subscribe: http://bit.ly/subgithub

Thanks!

Connect with us.
Facebook: http://fb.com/github
Twitter: http://twitter.com/github
LinkedIn: http://linkedin.com/company/github

About GitHub
GitHub is the best place to share code with friends, co-workers, classmates, and complete strangers. Millions of people use GitHub to build amazing things together. For more info, go to http://github.com

A

Github copilot goes ga a fresh new web framework, the best way to try out a new version of mac os and a way to do machine learning on a 40 year old computer. All that and more on this episode of the download.

A

Welcome back to another episode of the download, I'm your host, christina warren senior developer, advocate at github, and this is a show that covers the biggest developer news and open source projects of the week. So please, like this video and subscribe to our channel, so that I can keep buying ridiculous shirts like the one I'm wearing right now that I don't even remember when I got it, but I also missed the old kanye so anyway, all right, so we've been gone for a couple of weeks.

A

The tldr is that I was out of the country and then everyone at github had the week off. So I'm sorry, we will try not to have it happen again, but we've got lots of stuff to catch up on, so let's just get into it all right. So the first thing that I have to talk about is that github copilot is now generally available to everyone. Yay and github copilot.

A

If you're not familiar, is a fantastic extension in vs code, visual studio, it's available for jetbrains, ides and e of them, and it acts as your pair programming buddy your co-pilot. If you will helping suggest the next line of code- and this isn't just like a single word or line, it- can actually suggest complete methods, boilerplate code unit tests, even complex algorithms, it's very cool I've been using github copilot for the past year and it's awesome.

A

It has drastically improved my productivity as a developer, and it's made it easier for me to try things out and break things which I love and copilot is now available to everyone. It's priced at 10 a month, um though, there's a 60-day trial, so you can test it out and verified students and maintainers of open source projects can get it for free, and this uh is for individual github accounts, but co-pilot for companies will be available later in the year.

A

If you have not had a chance to try out copilot, you really should because it's awesome more info is in the links in the show notes in the description down below good stuff and speaking of co-pilot. Co-Pilot is actually built um using a ml model called the openai codex, and that is a descendant of gpt3 and gpt3 for those that aren't familiar.

A

It's a language prediction model that can be used to produce human-like text and it's one of the most exciting advancements that we've seen in ai and so-called deep learning in a really really long time. Well, simon wilson did something really cool, and that was to showcase how he can use gpt-3 to explain how code works- and this is actually pretty similar to the explain feature that's available in copilot labs.

A

But what I love about, simon's blog post linked down below, is that he really breaks down what the model is doing, how it works and he uses open ai's playground interface for that and the results.

A

The results are amazing, amazingly convincing um his blog post is definitely worth a read, especially some of the tips and tricks that he points out about avoiding maybe leading questions to get the best results. I'm just going to quote simon here for something that he says about gbt3, because as cool as all the stuff is it's still important to keep this in mind. Simon, writes, once again, I'm reminded that tools like gpt3 should be classified as bicycles for the mind. You still have to know how to pedal and bicycles for the mind.

A

Indeed, that's a great steve job system and it's cool to see. This is very cool and it's actually great to see machine learning explaining stuff in human terms. But if you're curious about what's possible with gpt3 and its descendants, you should definitely check out simon's blog post. It is linked down below next, if you're looking for a new web framework and look who doesn't want yet another web framework, uh I found one. That's totally fresh.

A

In fact: that's its name fresh, so fresh is actually part of deno, which is a modern run time for javascript and typescript, and it's similar to node.js. In fact, it's actually from the creator of node.js, but there's still a lot of important differences and fresh is a web framework. That's designed for speed, reliability and simplicity, and it has some really cool features baked in, and so that includes you know like. It's has typescript support out of the box.

A

There's zero runtime overhead, so that means no javascript is shipped to the client by default, there's just in time rendering at the edge and no configuration is necessary, very cool stuff. I've got a link to the docs and the project page in the links in the show notes down below I've been really interested in checking out deno, and this is a great way to do it. I'm actually going to probably build my next web project trying this out just to see how it works.

A

Next up, it is july, and that means that the latest mac os developer in public betas are available to the public, so mac os 13 known as ventura, will be out in the fall, the developers and people who like to try new things can install the betas now, but if you've ever wanted to install a mac os beta without destroying your main machine, you know that that process isn't always easy.

A

So like yeah, of course, you can install it on an external drive or an apfs partition, and then you can boot from that, but for a lot of users that can be complicated and time consuming.

A

Fortunately, my pal gear may rambo has released a new oss mac os app called virtual buddy that creates virtual machines of any version of mac, os 12 or later for apple silicon, and the app is simple and easy to use and you can use it um either to open like a custom, ipsw file, you can download a mac os installer from a list of options.

A

You can even offer it a custom url to use and then virtualbody will just spin up a virtual machine and away you go this project's fantastic, rambo, put it on github, which we love, and I'm just really grateful that the process of testing new mac os versions has been greatly simplified this way, so that I won't be tempted to say, yellow and install it on my main machine, which I do not recommend you do if you value your sanity, so use virtualbody instead, anyway, uh links to the project are down below great job rambo and now it's time for my pick of the week all right.

A

So we love retro computing here at the download, and this was something that I saw recently and it just kind of blew my mind all right, so nick build created a project that he calls tensorflow lite for the commodore 64., and this allows him to run inferences using tensorflow lite for microcontrollers on a commodore 64..

A

Yes, the same commodore 64 that came out in 1982. wait what okay. So this is how it works um and nick is using something called tensorflow lite for microcontrollers, and this is an open source machine learning framework where the tensorflow model is built and trained on a host computer, and then he extracts that sequence of operations and model parameters, and then he parses that into a commodore, 64 basic and then those inferences are run on the commodore, 64.

A

and so nick had to modify some of the tensorflow lite micro source code to get all this working. The way you want it to, but the results are great, like you can see from this video. It's working. It's amazing! I'm gonna be honest. Some of this stuff, all right. A lot of it is like over my head, um but he did it, he did it well.

A

I think it's great, you can check out his github project and his youtube demonstration in the links below I just again like this is so cool great job nick, I love it. What uh what throwback computer? Would you like to see some this sort of thing on? You know like what what other ml inferences should we do the atari 800, maybe the zx spectrum. Let me know in the comments down below and also let me know your thoughts on any of our other stories and in general. What do you think of the show?

A

So once again, if you like, this episode, give it a like and subscribe to the github youtube channel for all your nerd needs. See you next time.