Can a machine program by itself? The answer until not long ago was a resounding no, but that is not so true anymore, as demonstrated by GitHub Copilot, a system capable of generating code autonomously and thus helping developers save time and be more efficient.
The question, of course, is how far Copilot can go. At the moment, it is unknown how much its functionalities can be exploited until developers from all over the world use it and begin to exploit it to the fullest.
What is GitHub Copilot?
Trained in billions of lines of public code, GitHub Copilot puts the knowledge you need at your fingertips, saving you time and helping you stay focused. Source: GitHub.
It implements as an OpenAI engine with GPT3 a very advanced artificial intelligence implementation, which it generates with a series of suggestions when writing lines of code which allows generating large sections of code with functions that are repeated in computer systems.
How accurate will this platform be in predicting when writing code?
The most interesting thing about artificial intelligence is that as we use it, it learns from the trend in the writing of source code and its logic and generates more accurate predictions and code completion options.
Can all people with no programming knowledge use this platform?
At the moment it is aimed at people with knowledge in programming and a fairly advanced domain with respect to the area of technologies such as the development of complex computer systems as a novel solution for the technology industry.
Is it possible that this technology will replace programmers?
It must be made clear that this technology does not seek to replace programmers or developers, what it seeks is to lighten the work of programmers and focus their attention on business logic and not on technical details that waste time and thus be more productive by creating competitive software solutions in a very short period of time.
GPT-3 as a Written Language Writer
It is a new artificial intelligence model that allows the generation of written language. Thanks to this algorithm, the user only has to start writing a paragraph and the system itself takes care of completing the rest in the most coherent way possible. Its great potential is a sample of the possibilities that exist to reach a general artificial intelligence, capable of learning intellectual tasks like people.
Despite these limitations, there are other models similar to the GPT-3, but smaller, which already represent a real revolution. Its application can be seen in human language technologies (language processing, speech, translation, search or virtual assistants), in the generation of images or audio and even in the programming aids of ‘software’. This is the case of systems such as GPT-2 (OpenAI), BERT and T-5 (Google), ROBERTA (Facebook) or CTRL (Salesforce).
Even so, GPT-3 is presented as an alternative with great potential for the future. As it is a particularly good system for generating long texts that maintain a certain coherence between sentences, de Pablo predicts that “we will see more and more models of this type used to generate texts in a semi-automatic way. For example, for conversational assistants of all kinds, to help in the writing of news or other types of reports, and in the search for information ”.
A sample of how this technology works can be found in this essay published by The Guardian, which, as its editor comments, was prepared from the best fragments of eight articles generated with GPT-3 in order to capture the different styles and AI logs.
Some experiences on GitHub Copilot
In conditional blocks (if/else), Copilot is able to predict the log message that it wants to add based on the conditions of the block. It also highlights how the tool suggests a suitable name for a function without a single line of code having been written, only from a comment you write.
For him, the tool is not going to replace programmers far from it, but it is going to take away a lot of work and many searches on Stackoverflow, an online portal where many programmers resolve doubts.
For him, AI has a lot of potential and the code recommendations he has received are very good, especially those from the backend (that is, the programming related to making a web portal work correctly but that the navigator never sees).
It is still very green for User Interfaces and CSS code in the case of WEB programming.
Despite the good of the suggestions, you have to keep checking the code anyway because “sometimes it gets a bit messy assuming which variables you want to pass to some functions”. He helps a lot, but he doesn’t tune one hundred percent.
More opinions and experiences with this technology will continue to appear as time passes and more programmers begin to test it in their programming projects.
Is GitHub Copilot the only one implementing this new technology?
The answer is there are no platforms like Tabnine that implement artificial intelligence with GPT-3 for completion and code implementation suggestions, as explained with the following I have on their website “Tabnine’s powerful artificial intelligence assistant works directly in your IDE, for easy and seamless coding Tabnine AI studies publicly shared code using deep learning to predict and suggest time-saving code completions ”, Source: Tabnine.
How will the future be with these technologies that implement artificial intelligence.
Possibly humanity will achieve with these what it had not been able to achieve before, perhaps with this it will be achieved with this that technology evolves faster and in a very dynamic way, causing humanity to achieve what we cannot even imagine.