On 29 June 2021, GitHub took the world by surprise by launching their “AI Pair Programmer” Github Copilot, a tool that helps out with any kind of coding as it is written. Backed by Microsoft, this collaboration between GitHub and OpenAI relies on Codex, an AI engine built similar to GPT-3. While still in its infancy, the Copilot is compatible with a plethora of programming languages like Ruby, JavaScript, Python and TypeScript. In addition, the AI tool showcases its versatility by being available as an extension for Visual Studio Code, JetBrains and Neovim.
The anatomy of Copilot
GitHub Copilot is designed along the lines of its predecessor Natural Language Semantic Code Search, which suggested code examples based on simple descriptions in the English language. The Copilot, however, has the capability of generating entire multi-line programmes as well as documentation and test runs based on the context given by the user.
Mark Pesce, an award-winning columnist at Cosmos Magazine and a renowned name in the world of AI, reveals his experience in testing the newly launched GitHub Copilot and quoted: “I wrote a single line of comment that stated my requirements and – abracadabra! – the computer provided a complete solution.“
Now, let us understand the workings of Copilot through a simple example:
If I want to assign a certain text to a file name, all that I have to do is type: ‘write text file to fname’. The AI reads the syntax of the sentence and comes up with a suggestive line of code in grey that meets my requirement. One press of the tab button and the suggestive code is accepted and inserted into the body. Once the suggested code is accepted, the AI then builds on it and provides a variety of options with respect to the context of the test function.
With the launch of GitHub Copilot, programmers breathe a sigh of relief as the tool helps them furnish the code that they want to create with exceptional speed. Can the same AI tool help coders who look to create pieces of art?
AI on the digital canvas
Jeff (@ippsketch), the director of R&D in a medical device company, who has a penchant for using code to create masterpieces, showcases his work with Copilot on Twitter as he talks about filling the digital canvas with colour with minimal code.
Although a bit rudimentary, Jeff demonstrates the potential of GitHub Copilot to land some interesting brush strokes in the form of a high-resolution image.
Dylan Beattie, an international keynote speaker and director of Ursatile, explains in detail the art of code along with some examples.
GitHub Copilot: Here to change the game
Github Copilot is making a buzz with its receptive user interface and making headlines in the world of gaming. Coders are seen using the AI tool to create primitive games without writing a single line of code!
ThePrimeagen, a digital personality with Twitch, YouTube streams as well as name in GitHub, Linecode, Instagram and Twitter, describes his experience in GitHub Copilot writing the entire code of a game from scratch! Primeagen was so impressed that he quoted: “Yes the dang AI MAKES A GAME!!! I could not believe it and the ending just blew me away.”
Github Copilot is a real smooth talker!
The most tricky form of art is the art of socialising, and the GitHub Copilot seems to have that trick taken care of. Ido, a full stack developer from DagsHub, talks about his conversation with Github Copilot, where he spoke English instead of code, and the result was beyond his imagination! Ido was impressed that the AI tool could do non-code related tasks as well.
Ido further tested its ingenuity by opening a blank file and placing a simple question in the English language. While the AI tool does not respond to a simple sentence, given the proper context, it starts talking back!
GitHub Copilot is yet far from perfect, and studies suggest that there are areas of vulnerability when it suggests code. However, with a little fine-tuning and training of the OpenAI codex, the day is not far when this amazing AI tool can create masterpieces in a jiffy.
