text stringlengths 0 2.35k |
|---|
**Natalie Pistunovich:** So what would your ideal Copilot Go library look like? What would be a super-useful function/functions? |
**Alexey Palazhchenko:** I don't know... |
**Natalie Pistunovich:** I can imagine like replacing these reflections and so on would be -- |
**Alexey Palazhchenko:** So we were speaking about the machine learning model explaining this code. I wrote something like "explain this usage of concurrency". |
**Natalie Pistunovich:** \[36:16\] In the sense of what are potential outputs, or what are you missing when you're looking at it, or what exactly? |
**Alexey Palazhchenko:** So on many layers. For example, this function that taps these channels cannot tap that and that, and that channel is closed at this point, and then the big picture... And then the ultimate question - it should answer "Was it worth it a lot, or you should just write linear code there, without al... |
**Natalie Pistunovich:** So basically sort of a take one time -- |
**Alexey Palazhchenko:** Yeah. Actually, the last question would be the simplest one, I guess... It's just a constant answer, "No. Write simple code." |
**Natalie Pistunovich:** I guess that's part of the value proposition of Go, right? That it's simple. Last week we had - it was part of GopherCon - a game on Go Time where we had to guess what gophers answered. So there was a survey to some hundreds of participants of the conference, and one of the questions was "What ... |
**Alexey Palazhchenko:** Yeah, yeah. That's the reason why I chose Go in the first place. It would be my answer, too. |
**Natalie Pistunovich:** Go is used a lot, or more and more, not just for regular code, like web code, but also for infrastructure; like, different tools that are written in Go. So how can we improve that with AI? Let's see what we can do better. |
**Alexey Palazhchenko:** So let's imagine that we have an AI that actually can write the domain-specific language for our infrastructure. So for all those benefits of Go, Go is quite various language, and this error handling can -- I mean, writing business logic in Go is not the nicest way to use the language. It's gre... |
At the same time, existing solutions for integrating different other languages are also not great. |
So imagine that Copilot analyzes your business problem, your domain, and then creates a tailored, domain-specific language for your problem, implement it in Go, with virtual machine and just-in-time compilation. |
**Natalie Pistunovich:** Interesting... How would you name the startup that does that? \[laughter\] That sounds like such a useful pitch... |
**Alexey Palazhchenko:** Yeah. For that, we need to create another machine learning model, to come up with a name. |
**Natalie Pistunovich:** \[laughs\] Those are some interesting things. It would be interesting to listen to this show in like five years and see which of those already spun up to be a tool or a product. Or an open source project. Maybe somebody will go to Copilot and will write something, "Create for me a project that ... |
**Alexey Palazhchenko:** Yeah. And then you click a button, "Create Repository" and it already contains all the code in it. |
**Natalie Pistunovich:** And the license. |
**Alexey Palazhchenko:** Yeah. \[laughter\] |
**Natalie Pistunovich:** It's always important to have the right license. |
**Alexey Palazhchenko:** Yeah. So we'll be out of job. Completely out of job. You don't even have to pick up the license. |
**Natalie Pistunovich:** \[laughs\] Computers will be arguing with each other. Yeah, it brings us back to this AI judges, right? That seems to be like the logical next step of the computer industry. |
**Alexey Palazhchenko:** And then we all will be in Matrix. |
**Natalie Pistunovich:** Well, that is some interesting expected future for the development of our life as developers. Not to ask in our lifetime, but let's say in our professional career, which is hopefully for most people shorter than our lifetime, what things do you see that will be becoming a very common thing in a... |
**Alexey Palazhchenko:** \[40:11\] I think for machine learning you have to have some experience already. And if you don't have it by now, then in a few years it might be a problem to find a good job. And I guess -- yeah, I think Copilot will be much more common, in more editors, more languages, with different represen... |
**Natalie Pistunovich:** Would you be writing that on your CV? I'm asking because I saw CVs of people who write that they do Git as a skill, but I never -- I cannot remember, let's say, seeing a CV that mentions their favorite IDE. So like "I am good with Vim. I am good with VS Code." But would you say that Copilot is ... |
**Alexey Palazhchenko:** Actually, I saw quite a lot of CVs that mention a preferred editor. Like, "I have skills in IDE" for example. |
**Natalie Pistunovich:** Okay. Interesting. |
**Alexey Palazhchenko:** Yeah, people do write that. Maybe that's a Russian thing that people just write down all the keywords they know in English and everything else is in Russian... |
**Natalie Pistunovich:** Or you're coming very opinionated. "If you're not working with Vim, I don't want to work here." |
**Alexey Palazhchenko:** Yeah, that also happens, yeah. I saw that, too. I mean, with Copilot you don't have to have some special skills to use it, right? It doesn't have much configuration options. It's not like you could have a ton of knobs to tweak, and you are an expert of tweaking the machine learning model for yo... |
**Natalie Pistunovich:** \[laughs\] Copilot, yeah. |
**Alexey Palazhchenko:** Except Copilot developers, of course. |
**Natalie Pistunovich:** Yeah. What I can imagine is that sometimes when using the different OpenAI engines, like GPT-3, Codex, whatever, you have to write the prompt -- or you can write the prompt in different ways, and one way would yield you a better result than another, right? That's kind of the concept of prompt e... |
And back to this Advent of Code with Codex, I noticed that sometimes some of the days when I just copy-pasted the entire instruction of what is the Advent Code of today that you need to solve, for some languages it solved it perfectly from the first attempt, for some other languages, other days, were just not working. ... |
**Alexey Palazhchenko:** Yeah, that's definitely a useful skill to mention in your CV. You know how to relay things for Copilot. You're fast at tweaking words for Copilot. |
**Natalie Pistunovich:** I wonder if this is a similar category to "I'm really good in documenting my code." Would you write that in your CV? \[laughs\] |
**Alexey Palazhchenko:** Yeah, I'm also very quick at quitting Vim... |
**Natalie Pistunovich:** \[laughs\] "I can do that in only 7.5 seconds." \[laughs\] I guess that all falls under communication skills... But then maybe this is kind of the expanding my communication skills with my human teammates, and with my AI augmentation tools...? |
**Alexey Palazhchenko:** Yes, that would be interesting, if the skills should be significantly different and diverse. You speak with humans with one language, and with AI this different language. |
**Natalie Pistunovich:** There is a point to that, because the model was trained on English, or on natural language, and it was trained on code... But it was trained only on a subset, and it's always -- like, it will never be the same subset of the average person. It's not like the average person of one country has ano... |
**Alexey Palazhchenko:** \[44:22\] Yeah. And also imagine that you have to command AI to write the code for you, for it to work... But you don't want to command your co-workers, right? That would be a different language already on that level. |
**Natalie Pistunovich:** For sure. Interesting. Okay, so if we summarize kind of the tip around communication skills and how to present that next time you speak about your Copilot skills in a professional setup like a job interview, what would be your take-away? |
**Alexey Palazhchenko:** Yeah, so you could say that you have soft skills and hard skills. Soft skills is how you talk with people, and hard skills is how you talk with AI. \[laughs\] You can say that "Soft-hard skills" or "Hard-soft skills." |
**Natalie Pistunovich:** Yeah. Firmware skills, right? Firm skills. |
**Alexey Palazhchenko:** Yeah. \[laughter\] Yeah, separate section. LinkedIn will introduce it next year. |
**Natalie Pistunovich:** Yeah. More AI engines will be doing code, so it only makes sense. Well, that was an inspiring and interesting conversation... Time for some unpopular opinion. I hope you have one. |
**Jingle:** \[45:28\] to \[45:45\] |
**Natalie Pistunovich:** The nice thing about this tune is that even on episodes where Mat is not here with us, he is a little bit with us... Because he is the creator of this short tune, so... Hi, Mat. I hope you're feeling better. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.