This week I learned more about automated code generation and AI programming assistants. This tool can generate well-written code based on prompts written in plain English. Today, AI programming assistant tools based on deep learning are finally becoming viable. I believe this will be a game-changer for programmers, making them more efficient.
Naturally, when powerful new tools are developed to reduce human workload, workers get worried that they’ll lose their jobs to machines. Yet throughout history, machines have phased out grunt work, and human labor has evolved in step. AI programming assistants make it faster for developers to translate their ideas into well-written code.
Another reason AI programmers won’t replace human developers any time soon is that a developer’s job is much more than just coding. For example, developers must create design specifications and work with customers, supervisors, and colleagues.
This week, I talked about two AI programming assistants in my class discussion – Tabnine and GitHub Copilot. Tabnine is less ambitious on the AI code generation side of things and has an autocomplete feature similar to IntelliSense. While GitHub Copilot isn’t actually available yet (I’m on the waiting list), I have to say I’m more impressed with the technology behind GitHub Copilot. It’s based on OpenAI Codex, which uses the GPT-3 language generation model. I’ll include a link at the bottom for more info.
Once it’s released to the public, I’m really looking forward to seeing GitHub Copilot in action. Today I installed Tabnine for VS Code, and I plan on trying it out tomorrow when I do some programming in Python and JavaScript for one of my classes. I’ll probably mention Tabnine in next week’s post if I like it enough.
https://openai.com/blog/gpt-3-apps/
