More and more students and full -fledged specialists use AI technologies to work with code. For example, recently, a teacher from the University of Duke, Noah Gift, having 25 years of experience in the field of software development, decided to radically change the program of his course. He switched with Python, one of the most popular languages for beginners, on Rust is a newer, powerful and complex option. This unexpected step was possible thanks to the appearance of Copilot – innovative AI instrument from github.
Copilot, created on the basis of the latest large language models from Openai, GPT-3.5 and GPT-4-this is an improved code for auto-filling code. It monitors pressing each key on the programmer’s computer, predicts his intentions and offers suitable fragments of the code for the task. The gift, having learned about Copilot from a friend from Microsoft, immediately saw his potential and decided that it would be useful for students to get acquainted with technology.
It is interesting to note that most of the students of computer sciences and programmers have already used Copilot. Despite some technical flaws, its effectiveness was recognized even by those who once expressed criticism. Included in a paid plugin for Microsoft’s Visual Studio, Copilot is the most advanced version of this technology, although not one of a kind.
In August, Meta released its free model for generating code – Code Llama, based on Llama 2 (GPT -4 Analogue). In the same month, Stable Ai, the company standing behind the Stable Diffusion images model, introduced Stablecode. There is also an Openai ChatGPT. It is presented as a chatbot and mainly works with the text, but also helps in writing and debugging programs.
These tools not only simplify the work, but also cause concerns about the theft of intellectual property and copyright. For example, some companies, including Apple, asked their employees not to use Copilot, fearing that IP and other confidential data will fall into the hands of competitors. In addition, Copilot has become the center of a court claim filed by IT specialists, dissatisfied with the fact that their code without permission is used to teach the model.
Despite legal disputes, it is expected that the influence of Copilot and similar tools can add up to 1.5 trillion dollars by 2030 to the global economy. These tools not only help existing developers to produce more products, but also reduce barriers for beginner programmers.
Thus, Copilot opens a new era in programming, where human synergy and artificial intelligence creates more complex and effective products. However, according to the GIFT, the successful use of technology still requires professional skills and knowledge in the field of IT.