Using OpenAI language models, Co-Pilot, and Co-Pilot Labs for code suggestions and generation

This chapter discusses the different language models offered by OpenAI, including GPT 3.5 and GPT 4, and their applications in testing and writing specs. It also introduces Co-Pilot Labs by GitHub, which provides tools for explaining and translating code.

Transcript

chevron_right

Play full episode

chevron_right

Transcript

Episode notes

Welcome to Ruby for All Season 2! This was the last episode we recorded before the break that we wanted to share.

Today’s episode of Ruby for All is mix of personal experiences, tech discussions, and the ever-evolving world of artificial intelligence. Andrew and Julie dive deep into their utilization of ChatGPT, an artificial intelligence tool that assists with tasks such as writing performance reviews and explaining code functions, and they envision a future where the integration of AI into workflows becomes a standard. Andrew, a long-time AI enthusiast, brings up the topic of OpenAI’s Whisper, and the exciting possibility of creating AI-driven podcast episodes. They share their experiments with AI personas and how giving a ‘character’ to the AI can improve its output. Andrew and Julie also have a conversation sharing their different uses of ChatGPT. Join us as we continue exploring the fascinating world of Ruby and AI. Hit download to hear more!

[00:01:18] Julie and Andrew discuss their uses of ChatGPT. Julie uses it for writing performance reviews, while Andrew is a heavy user of the tool and its API.

[00:04:00] Andrew mentions the Whisper from OpenAI and the possibility of using it to recreate podcasts episodes purely from AI, using the hosts’ past audio data.

[00:05:37] Andrew asks Julie about her experience with Copilot, he tells us what it does, and he finds it most helpful for writing tests.

[00:09:23] We hear about the use of “brushes” in Copilot labs, a GitHub project. Brushes can explain code, translate it into other languages, generate tests, and more. He also tells us about how he uses Warp.

[00:11:19] Julie and Andrew agree that ChatGPT and Copilot are useful for learning and improving code, especially for those early in their careers. However, the AI can sometimes provide inaccurate information.

[00:14:42] Andrew thinks using the AI to teach oneself how to code could be effective but cautions against relying solely on copy-pasting generated code.

[00:16:07] Julie wonders if ChatGPT or AI is going to replace tech jobs, and Andrew doesn’t believe it will. He does mention new professions like “prompt engineering” have emerged with the arrival of AI, with high earning standards.

[00:17:58] We hear the difference between few-shot prompting and one-shot prompting in ChatGPT. Julie has experimented with asking ChatGPT to act a knowledgeable senior software engineer, and Andrew mentions the new “browsing mode” in GPT-4.

[00:22:19] Andrew uses the AI for making chapter notes for the podcast, saving him a lot of time.

[00:24:10] Julie uses AI to help her understand complex concepts like Redux, Sagas, and React. Andrew tells us about an app he uses called Writers Brew, which has a feature to convert summaries into bullet points, and Reader from Readwise, to analyze and summarize articles, generate flashcards, and thought-provoking questions.

[00:26:11] Andrew goes through some recent questions he’s asked ChatGPT recently including explaining complex code functions, creating CSS selectors, and creating titles for code snippets.

[00:27:38] Julie shares a shortcut she uses that Andrew wasn’t aware of, she asks ChatGPT to explain code an follows up with clarifying questions, and she uses it to develop SMART goals for advancing at work.

[00:28:31] Andrew notes that ChatGPT is from OpenAI and utilizes the GPT-4 or GPT-3.5. language models, there’s also Google Bard, and other language models such as, Hugging Face and AI Playground that you can explore.

Panelists:

Andrew Mason

Julie J.

Sponsors:

Links: