Article/Ali Tao F(X) Team – Tianke
This article will introduce some applications of gpT3-based code generation/analysis, including generating front-end React code, Python code, SQL, mathematical formulas, and more. Analyze and discuss the usage scenarios with your friends.
The overview
First, give a list of apps and the original Twitter address (some need to cross the wall) :
-
Front-end layout generator
-
Generate design sketches from descriptions
-
Generate the React todo list from the description
-
Generate a React Component from the description
-
Generate a React Component from the component name
-
Translate natural language into shell commands
-
Explain the code
-
Generate formulas from descriptions
-
Generate SQL from the description
-
Code for the interview
-
Generating Python code
-
Generate machine learning code
Front-end layout generator
Front-end layout generator: Based on GPT3, generates arbitrary layouts from descriptions.
Some comments:
- Unbelievable. When are you gonna get a public access Demo?
- UX developers are dead.
Generate design sketches from descriptions
Generate designs from descriptions: A Figma plug-in called Designer allows you to generate designs from descriptions.
Some comments:
- I don’t buy it. It looks like a bunch of if statements.
- So, how long before we use robots as code farmers and designers?
Generate React Todolist from description
Generate React Todo List from description: In addition to generating layouts, you can also generate logical Todolists.
Some comments:
- How long before you get your API key?
- Cool. What do your training numbers look like?
Generate a React Component from the description
Generate a React Component from a description: Another example of a logical React Component.
Some comments:
- Someone must have written the code you describe, because I don’t believe there is an app with this level of language understanding.
- Why not use voice control?
Generate a React Component from the component name
Generate React Component from component name: The programmer simply writes a component name and the AI completes the rest of the code.
Some comments:
- Too cattle!
- Where can I play?
Translate natural language into shell commands
Translating natural language into shell commands
- I was surprised that it knew how to quit Vim.
- Q: Delete all files A: sudo rm -rf / Q: No no wait! No!!!! A: echo “Too late”
Explain the code
Explain the code: read the code and answer questions about it. For example, what is the type of this variable? What does this code do?
Some comments:
- Did you try to make it automatically generate documents?
- That’s amazing. Wonder how it works.
Generate formulas from descriptions
Comments from the description generated formula section:
- Next, the speech transfer formula
- Have you ever tried a formula to generate an English description?
Generate SQL from the description
Generate SQL from description section comments:
- 1. How to access this? 2. How about using slang?
- Game over for SQL
Code for the interview
Code interview: Treat the AI as a candidate for an interview. Give it questions to write code.
Some comments:
- Based on what data? Job interview dialogue?
- Interesting but useless
Generating Python code
Generating Python code
Some comments:
- Look nice
- Can I use it?
Generate machine learning code
Generate machine learning code: Input function description and generate Keras code
Some comments:
- The result is a very bad image classifier
- When is the Beta available?
conclusion
For now, these GPT3-based code generation/analysis applications are “toys” and few are open to the public. But these product designs are valuable, and when the technology matures, these products have the potential to transform our working lives.
