Introduction to Devin AI
In the midst of the revolution of artificial intelligence and machine learning models, Devin AI, a software engineer trained by developers from Cognition Labs, emerges. The developers hope that Devin will be a good companion for those working in the field, taking on the burden of solving problems and routine programming of applications, and searching for solutions to frequently recurring problems. To this end, they have equipped him with high capabilities, which we will present to you in this post based on the reports and videos published by the developer company.
Devin's User Interface and Functionality
Devin's user interface consists of four main screens, between which he constantly navigates to produce the desired program. The first is the shell, which displays the command sentences that Devin executes as a small window through which we can peek into his artificial brain. In this window, we see the sequence of his decisions and his processing of problems written in machine language. This is followed immediately by the browser, which allows him to connect to the internet to open links that we provide him with, or to conduct a search for answers to his questions, or even to access learning resources to read, analyze, and learn from. In almost all of the videos shown, Devin appears browsing websites, reading their data, looking at the codes used in them, and drawing inspiration from them either for a solution to his problem or a way to implement his idea! Next is the editor window, where Devin writes the code, runs it, reviews and edits the errors, and even finds solutions to any programming problem given to him. This window is his compiler. Finally, there is the planner window, where Devin summarizes his plans for solving the problem, how he expects each experiment to help him, and how he will deal with potential exceptions.
Devin's Problem-Solving and Learning Capabilities
I assume that the previous paragraphs have presented an overview of the capabilities of this artificial intelligence. The first is self-learning to solve problems that he was not trained on during the development period and his ability to learn like a real human from any piece of information that he is provided with. The next is his ability to build any required software with a user interface and database that matches the given text instructions. He even goes beyond that to his ability to train artificial intelligence models that follow him and perform his tasks for him! Devin receives the assigned task through the command window, reads it, and analyzes it, which is shown on the shell screen. Did you provide him with a specific link? Does Devin know how to do what you asked? Devin uses the explorer window to deal with these two questions. You can see how he reads the pages and learns from them through the command screen, which shows him constantly switching between it and all the other screens. Once the learning process is over, the software engineer goes to the planning screen and there he puts in plain language how he plans to write the program before he starts it. He returns to that screen whenever he encounters an error or exception to keep track of his attempts. For example, you see him recording the possible causes of the error, then adding how he plans to detect each one, and later recording the result he got from his experiment. Not only that, but according to his ability to solve problems, Devin was able to solve GITHUB problems with an efficiency that exceeds all announced artificial intelligence models by about 13%, without any human intervention or written commands, and without even being given the text of the problem literally. Give him the link to the site and he knows what to do!
Devin's Performance and Comparison with Other AI Models
According to the company's announcement, Devin outperforms all released ChatGpt and Gemini models in terms of his ability to face and solve problems. However, he is not yet ready for use, and we have no proof of his actual capabilities other than through video clips that his developers filmed themselves and displayed on their website. Unlike Google, which displays the development and production mechanism through its Google for Developers website, Cognition Labs keeps these details, so as much as the announcement of Devin's capabilities is impressive, we can't help but be skeptical of his actual capabilities and performance once he is added to the competitive arena.
Skepticism and Ethical Considerations
The company and critics do not deny that Devin, with all his impressive capabilities and expected promising start, will not replace software engineers and the huge number of tasks they perform. However, at best, he is expected to create a space for professionals to direct their energies and capabilities towards creativity and innovation, and leave the routine work of building applications and searching for solutions to their virtual companion Devin! We do not know what ethical boundaries Devin will be subject to, as is the case with all the amazing developments that are happening at a frightening speed today. We do not know yet whether Devin is a bitter enemy or a dear friend.