GitHub GitHub Universe 2022 - AI, 21 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Level up with Copilot - Universe 2022

Description

Presented by: Rizel Scarlett

If you’ve tried GitHub Copilot, you know that it can feel like magic. The tool fills out entire blocks of code with shocking accuracy—though knowing exactly where to correct it requires skill. In this session, Rizèl Scarlett shares how GitHub Copilot can help you level up your software engineering career.

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

Hello: everyone today we're going to talk about how to level up with GitHub copilot I, love, audience interaction. So I'm going to ask you all a couple of questions, even if you're at home, you can give me an emoji or a thumbs up or a thumbs down the other day, I tweeted out I wrote this tweet with copilot. Do you think that was true or false? Give me a thumbs up if you think it was true or a thumbs down.

A

If you think it's false, my next question to you all is: do you think you can use GitHub co-pilot to cultivate your engineering career at any level?

A

Give me a thumbs up again, if you think it's true or a thumbs down for false, if you think I'm lying, let me tell you a couple of the things we're going to cover in today's talk, we're going to talk about what is GitHub copilot we're going to talk about how to use GitHub copilot we're going to explore if you can use it to level up your career I'll talk about how it's personally helped me and we'll talk about.

A

If you can actually send tweets with copilot before we get into that, I will introduce myself because you're probably like who is this amazing person talking to you? My name is Roselle Scarlett I'm, a junior developer Advocate at GitHub. Prior to that I was a software engineer for about three years. I am currently an advisor and I was a director at an organization called G-Code which introduces women of color and non-binary people of color to code, I'm, really passionate about empowering people through technology, but I'm also passionate or some might say, addicted to social media.

A

So, if you're on the interwebs as well, you can go ahead and give me a follow at black girl bites on Twitter, Instagram, dev.2 and hashnode, or on Twitch I'm. Actually, black girl bites one, because I got myself locked out of my account. I forgot my password, so let's jump into it. What is GitHub co-pilot copilot is an AI peer programmer that helps you write code faster with less work and it's not magic. I know when co-pilot first came out, a lot of people were nervous that it was going to steal their jobs.

A

That can't happen y'all, it's not a sentient being. It can't read your mind, although sometimes it kind of feels like it, but it does need the programmer's brain to actually work and I will always be honest with you. uh Copilot doesn't always spit out perfectly well written or up-to-date code if you're still having a hard time envisioning. What this looks like it is pretty similar to the smart compose feature in Gmail or other Google products where sometimes when you're typing you'll see autocomplete Happening Here.

A

In the example, we have someone writing to Jacqueline about Taco Tuesday. They say: hey, Jacqueline I haven't seen you in a while and then smart compose, jumps in and says and I hope, you're doing well that user experience is pretty similar with GitHub co-pilot, except it works for code under the hood. What's happening is GitHub. Co-Pilot is drawing context from your comments and your code and it's suggesting individual lines and whole functions instantly. Github co-pilot is actually powered by something called open. Ai codex now I'll be honest with y'all for a long time.

A

I was repeating that without really knowing what openai codex was so I decided to do a little bit of research before talking to you first off openai is a research lab an AI research lab to be specific and I already knew this because they developed the cool Dolly, which is an AI image generation tool, but I still didn't know what codex was and I dug a little deeper and I found out. It's a model that open AI developed that translates natural language into code.

A

Openai codex is actually a version of something called gpt3 which stands for generative pre-trained, Transformer 3., and that uses deep learning to produce human-like text. Gpt3 is actually used by some of your favorite apps out there like Duolingo uses it for grammar correction. Codex is pretty similar to gpt3, except it's been fine-tuned for programming, so I'm going to go ahead and like play around with some of openai's examples. By going to beta.openai.com examples, they have like an API playground, so you can get a better idea of what it is.

A

So here I'm going to choose their movie to Emoji converter. Basically, it takes movie titles and converts it into emojis. So you can see in the first example for Back to the Future. It gives us a young man and old man, a car and a clock and then for Star Wars I want to see what it's going to give us. I'm gonna hit, submit and see what gives us so it gave us like a star and a Starship.

A

um Now I want to try out the Devil Wears Prada, that's one of my favorite movies. Let's see what it gives us all right, so, once I hit submit, it gave me high heels, lipstick and someone dancing in a dress. This was pretty fun, so I'm gonna play around with some of their other examples too. This one is a mood to color converter. So what you do is you give it a description of a color and it generates a hex code for you in this first example.

A

It says a blue sky at dusk, I'm, gonna copy and paste that hex code, because I don't memorize hex code, colors and I'm gonna paste it into Google and see what it gives us cool. It gave us like a little Dusky blue. So it's pretty on point in my mind, uh I want to try out some more, maybe a basketball. Hopefully it gives us the right, orange I hope you know what I'm talking about like not too bright of an orange or not too, like burnt orange.

A

Let's see so it generated a hex code for us. I'm gonna go ahead and copy that and paste it into Google, and let's see what it gives us all right. That looks pretty basketball lee orange to me uh and the last one I want to try is a sunflower sunflowers. Have this golden yellow I think they are so beautiful? uh Let's see what it gives us, I'm gonna hit submit once I get the hex color and paste it into Google, and let's see what happens all right. That is a pretty golden like yellow to me.

A

I love. It super fun now, you're, probably like okay. These are cool, but what about the code? Don't worry, I am getting to that. So in this example, I'm gonna go ahead and write a comment that says, write a binary, search algorithm and, let's see what happens, let's see what codex does so I press submit and it actually finishes my comment and writes the whole function, along with other comments in between it.

A

Just explaining what's going on within the code base, I think that's pretty cool all right so now that we understand what openai codex is and what copilot is using to to generate this predictive code, let's figure out about some of the features within copilot.

A

First off we have predictive code, that's the one I've been talking about the most, that's the one you're probably most familiar with, and in this screenshot you can see an example of I wrote a comment: I started off the first line of the function and then what it did is it auto completed the rest of the function, for me, isn't that amazing? Now the this is another feature where it allows you to view more suggestions and I.

A

Think, maybe many people don't know about it, I sure did it so I used to use it and get like a little bit frustrated with copilot, because it wasn't always suggesting the desired suggestions for me.

A

So the other day I was talking to the product manager and I was like you know what would be cool if we can view more suggestions- and he told me that already exists and I was a little embarrassed, but pretty happy to to learn that that's how it works all I had to do is actually hover over the initial suggestion and then I would get to see previous suggestions next suggestions and, if I clicked, open, GitHub co-pilot I can see all the suggestions that GitHub co-pilot was going to give me another feature within copilot is actually within the co-pilot Labs extension.

A

Now this extension was developed by GitHub. Next, it's an experimental extension that comes with copilot access. The first thing that it allows you to do is Translate a code snippet from one programming language to another. So if you can see in the screenshot, I actually took a code snippet of some JavaScript and I prompted it to translate it into Python and I. Don't know that much python y'all but pythonistas tell me if that looks correct to you. They also have a feature where you can get an explanation of.

A

What's going on within a code snippet so again, I've highlighted a code. Snippet I prompted it to explain the code for me, and this screenshot looks a little small, so I'll zoom in for you, but here you can see it's it's um stepping up through what's happening within the code, all right so maybe you're like this is so cool I want to install it like right now. Well, if you have copilot access, you're able to install the extension in Visual Studio code, jetbrains and neovim, and now you probably like these are really cool features.

A

But how is this useful for me? You called this talk level up with copilot and I still haven't seen. How take a moment to think of some of your software engineering goals throughout your career. I'll share mine with you.

A

I always want to be able to code faster and cleaner I want to pick up an issue and give it back so quickly that my engineering manager, whoever is supervising me, is like wow Roselle is super smart, and then everyone looks at my code and they're like wow, and it's clever and clean too I also want to be able to write good documentation. I want to be a good Mentor. A mentee and I want to gain understanding for New Concepts quickly.

A

These have always been my goals as a software engineer, since I became one in 2018. and co-pilot seems to have been helping me to accomplish these goals since I've been using it. It continues to help me code faster and cleaner. It continues to help me write, good documentation, be a better mentor and gain context for New Concepts quickly and I'll. Show you how first off, in terms of coding, faster and cleaner co-pilot, is really really great at picking up patterns.

A

So if I'm writing a unit test or creating an API or using an API after I write the first function, copilot kinda gets the gist. It's also really good. With syntax now, I don't know about you, but I'm, never going to memorize, regex I, think it's really powerful, but to memorize all those little digits is a little much so I'm, usually Googling or going into a regex playground to find the answers. The cool thing about copilot is I. Don't have to do that anymore. Copilot will generate the regex for me.

A

Also, it's really really great with comments so to get the best results from copilot. You need to write clear, understandable comments, and this benefits both you and other readers, because if an AI can understand your code, imagine how well a human will I'll share a personal experience with y'all I feel like I write much code and I said this talk to one of my friends who is a non-engineer and he was like. Oh my gosh, don't say that don't tell people you write messy code, but I'm sure y'all could relate with me.

A

I usually write message code and then go ahead and refactor it. So I was a little nervous when I first started at GitHub, and we collaborated with the Microsoft team and they wanted me to build a vs code. Extension I was like oh my gosh. Now everyone's gonna see my really bad code but turns out I got this email that said code is gorgeous love it so easy to read. I was really confused. I've never received a compliment like that in my life about my code turns out that I think using copilot actually helped me.

A

I wrote really clear comments for co-pilot to be able to generate the functions I needed and it resulted in this. Everyone has different strategies when writing code or using different, tooling I'll share mine with you when I use, GitHub cool pilot I'll turn it off when I'm laying down the initial structure of a project and then I'll turn it back on once I have a pattern going. This is just is what works for me now. I'll show you all a couple examples of patterns, Mill Mike Dev on Twitter, actually sent me this example.

A

When I wrote a blog post, where they're, essentially translating the words answer and question and date and all they had to do- is put in the language codes for like French and Japanese and what it did is it actually translated them for Mill miked up without Mel mcdev having to do it? Alright. Here's an example of me using regex, so I wrote a comment that says: validate a phone number and I start writing out the function, and it's able to actually do all the regex for me.

A

I didn't believe it so I ran it to see if it would work and it came out to be true now this is my favorite part about GitHub copilot that it's really helpful for those who speak English as a second language. Already code is kind of hard to remember, like the syntax of it all and it's pretty American English Centric. So, for example, the CSS property color uses an O, whereas people who speak British, English or Canadian English may be used to spelling it with a.

A

U and I think co-pilot's super helpful, which is like helping you to remember all of that. Syntax. Now I discovered that GitHub co-pilot can understand other spoken languages when I was sharing GitHub co-pilot with my friend Ramon, he told me, do you think GitHub copilot can understand Spanish and I was like I, don't know, so we ended up writing a comment in Spanish that essentially says import the necessary libraries needed for this program and it imported Tweety for me, which is the exact library that I needed all right now, co-pilot can be great for documentation.

A

Please keep in mind again that co-pilot is fine-tuned for programming and not necessarily Pros so be be wary of that. But I think it's really great if you want to add in code Snippets to your documentation and it speeds up the process. Here are some examples. My friend Hayden Barnes actually was writing some install requirements, and here all he writes, is on Zeus and already knows the command, and that is needed to install whatever he's creating on Zeus.

A

Also a Microsoft developer, Advocate tweeted and admitted that actually half of the static web app CLI documentation is written using copilot I'm, assuming that's a lot of the code Snippets all right. Let's talk about being a better Mentor, mentee, I'm gonna admit something else to y'all.

A

I am I, get really nervous when I am peer programming with somebody and I'm, the driver and I'm learning from my mentor I get super nervous that I'm going to make a lot of typos and syntax issues, and that just makes me make more typos and I stop focusing on the actual logic of the program. Having co-pilot allows me to focus Less on the syntax, because it's going to auto complete it for me and focus more on the logic.

A

Also, on the flip side, when I'm a mentor, I normally teach people HTML, CSS JavaScript, they go on and learn more and then they go on to get a job, and sometimes they come back and they're like okay. I need help with this language that you don't know like C, sharp and I'm like uh don't know how to help, but what's been helpful to me, is using copilot labs to take a couple of code. Snippets get an explanation of what's going on and then I'll be able to help them. That way, all right.

A

Let's talk about gaining context for New Concepts: this is gonna sound, a little unconventional, but I've used it to study for interviews going through things like leak code, especially when I'm on Elite code, hard and I get stuck. That's an easy way for me to start like watching, Netflix or cuddling up in bed, because I'm like this is too hard I give up. But what co-pilots helped me to do is gain momentum.

A

Sometimes I'll turn co-pilot on it'll answer it for me: I'll delete the answer, deactivate, co-pilot and then I'll challenge myself to be able to respond to it on my own. Now this might sound weird or like cheating but I, don't think so. In fact, I learned that chess players actually play against an AI to help them level up. So if chess players can do it, why can't software Engineers?

A

It's also great, if you're navigating a new code base, so if you're new to a company or an open source project- and you see a piece of code that you're like what is going on here- you can actually just highlight it and get an explanation of what's Happening. Also, it's been helpful for me as a developer, Advocate doing demos. So a lot of my job is doing demos with like tools and programming languages that I'm not getting that much experience with as I did when I was a software engineer.

A

So it's kind of sped up the process and I'll give you an example. The other day I worked with the Twitter devrel team to do a twitch stream and they wanted me to show Twitter's API V2 in Python I, don't know python, so that freaked me out that other people were going to be watching me write in a language. I didn't know: I created a demo just within 30 minutes of me just writing in Python, and then there's just me on Twitter, like celebrating with a gif I.

A

Don't know if you all know um that video where Usher goes watch this now that you've seen all of the cool things that co-pilot can do. Let's use copilot to actually write a tweet. So in this next video or GIF I'm going to show you how I wrote comments and co-pilot does the rest. So here I just write a comment that says import the Tweety library and it Imports it for me, after that, uh it already knows: I want to import a config file with my credentials and after that it knows.

A

I want to authenticate so I'm just doing a whole bunch of tabbing at this point, so I'm just going to keep tabbing until it finishes it for me, and actually it's a little bit off-centered and I know python. Here is about like being well formatted, so I'm going to take some time to format it really quickly, uh once I do that, I can get back into what I was doing before next thing.

A

I want to do is create a variable with like the tweet that I want to send out and copilot's trying to help me out with that which is cool, I think I want to say something like hi I'm at GitHub universe and I sent this tweet. What co-pilot? Let's see what happens.

A

A

Was able to generate the variable for me inside of the quotes with all the things I wanted now I want to send the Tweet copilot knows that I'm just going to press enter and tab, and then, after that, I want to print it and I press enter and tab as well. Let's go ahead and see if this actually works so I'm going to run it in my console. I'll press enter and, let's see what happens.

A

Cool the response came back, fine, I, don't know it seemed too easy. Let me go check my Twitter to actually see if this got tweeted out so I'm scrolling down and it looks like 10 seconds or eight seconds ago. I actually did write a tweet I'll delete it because I've done this demo millions of times my timeline is probably tired of me.

A

Tweeting this out now I actually decided to try this in a different language, just because I thought this was so cool, but I wanted to try more so I brought my friend in Ramon who actually speaks Spanish to help me out and here's a video of us trying this out on stream wanna like prompted to write the to create a variable that like says this, but in Spanish I, don't know: okay, I'll stop highlighting it.

A

So people can see okay, um so yeah, let's see what we got here, we go, it says hi, it even says writing a tweet using python or something. Is that what it says it says: hello, I am a tweet sent from python all generated. This is Bonkers.

A

I can't believe this works, as you can see, it worked and we're super excited. I, always love to geek out about co-pilot and show share it with my friends. uh I just think co-pilot's so powerful the fact that it actually generated a Spanish tweet for us. Let's take it a step further sending a tweet with cool pilot sounds fun, but it's not really practical. I can easily tweet from like my mobile device or my web app, so I thought. Maybe I could take it a step further and send a tweet every time.

A

I label a good first issue: I have open source on the mind with hacktoberfest and working at GitHub, so I decided. Let me see what I can do to make open source easier for um for new contributors to to discover I made a GitHub action workflow that essentially runs the file that we just created with GitHub co-pilot every time she was labeled good first issue.

A

So, as you can see here, I write if the label has a good first issue then grab the URL of that issue and then go ahead and run the file and pass in all the the secret credentials that we need and the results show up with me saying. This is a good first issue, with a link to that actual like issue there, which is kind of cool right, so I think we were able to prove that. Yes, you can send tweets with co-pilot and you can use it to level up your career.

A

So I want to end by just highlighting a couple things for you: I actually wrote a blog post on github's blog called Eight things. You didn't know you could do with GitHub co-pilot, which add an even more examples beyond what I've shared in this talk, so go ahead and check that one out, if you want to learn more and go ahead and scan these QR codes, if you want to link to the slides or if you want to connect with me, go ahead and scan that as well.

A

Well again, my name is black girl bites on social media and I want to thank GitHub Universe for allowing me to speak and I want to. Thank you all for listening, bye.