Recently, the technology startup Cognition Labs introduced Devin, the world’s first software engineering AI tool.
This artificial intelligence (AI) tool has advanced reasoning capabilities and can automatically take on numerous software development tasks. This is considered a new milestone in artificial intelligence, yet it has also raised concerns among many software engineers.
As we know, language models like ChatGPT 4 can write programming code, but Devin—a new AI tool from Cognition Labs—can do much more.
This is not just a simple code-writing tool; it acts as a virtual software engineer, capable of expanding its capabilities to create other AI models.
While this new tool has impressed many scientists, it has also sparked numerous concerns within the software programming industry. Will we still need programmers in the future?
This tool can ensure the work of a team of programmers. (Image source: Trust my science).
Cognition Labs has tested its product in real-world tasks.
In one test, engineers assigned this AI tool with the task of setting up, debugging, and testing a computer vision model.
Ensuring the Role of the Entire Team
Devin can perform the work of an entire software development team. It ensures completion of all phases, from planning to determining requirements and deploying software designs.
The engineering team has integrated project management and business analysis skills, allowing Devin to understand goals and plan necessary steps to complete projects.
To accomplish its tasks, Devin creates “sub-assistants” that take on parts of the work; they interact with various development tools (terminals, code editors, browsers) to write code, test, and debug software.
Additionally, Devin has the ability to learn quickly and adapt to new technologies. All it needs to do is read blog posts to understand how a tool works and then utilize it.
Creating AI and Solving Problems
In a testing video, Devin replicated Meta’s Llama language model, read the README file to understand how to configure it, and then automatically initiated a training cycle.
In this test, Devin successfully removed or corrected non-functional issues. This demonstrates its capability to independently solve programming problems.
In another test, Cognition Labs aimed to assess Devin’s performance in addressing deeper programming errors.
The engineering team used open-source projects found on GitHub, and the tool resolved nearly 14% of identified error sets without any external assistance.
What Does This Technology Mean for Developers?
The emergence of Devin and the rapid advancement of AI in general are undoubtedly causing concerns among software developers.
However, at this point, companies cannot fully entrust their software and application development to AI, especially those that require high reliability and security.
The idea of allowing AI to “intrude” into entire codebases and make automatic edits will undoubtedly raise waves of concern regarding security aspects.
For example, changes made by AI could introduce bugs or security vulnerabilities, particularly if the AI’s decision-making process is not transparent.
Therefore, we will still need time and reliable evidence for enterprise customers to fully trust this type of technology.
