Last Updated on March 13, 2024 1:24 pm by Laszlo Szabo / NowadAIs | Published on March 13, 2024 by Juhasz “the Mage” Gabor
Will Cognition AI’s Devin Replace Software Engineers? – Key Notes
- Devin, Cognition AI’s first AI software engineer, excels in the SWE-Bench coding benchmark, completes practical engineering interviews, and real jobs on Upwork.
- Unlike GitHub’s Copilot, Devin independently creates code for entire projects, including debugging.
- Demonstrated by CEO Scott Wu, Devin can plan and execute complex tasks by learning from the internet and using standard developer tools in a sandboxed environment.
- It allows engineering teams to delegate tasks, focusing on creative aspects better suited for humans.
- Devin has completed tasks like creating a game and a website in under 20 minutes, showcasing its capability to handle both simple and complex projects efficiently.
Cognition AI’s Devin on Board
Today we’re excited to introduce Devin, the first AI software engineer.
Devin is the new state-of-the-art on the SWE-Bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on Upwork.
Devin is… pic.twitter.com/ladBicxEat
— Cognition (@cognition_labs) March 12, 2024
Cognition, a new startup utilizing generative Artificial Intelligence, has launched a tool that can independently create code for entire engineering projects, including its own AI models.
This tool, which goes by the name of Devin, goes beyond GitHub Inc.’s and Microsoft Corp.’s Copilot developer tool by being able to complete entire jobs on its own, rather than just assisting human coders.
In the video attached to the Twitter / X post introducing Devin, Cognition CEO Scott Wu demonstrates how users can observe the model in action, including its command line, code editor, and workflow as it completes coding and data research tasks assigned to it. Once Devin receives a request, it will search the internet for educational content to learn how to complete the task.
It can even debug any issues that arise during the process, but human users can intervene if necessary. The launch of Devin marks a great shift in the emerging AI-powered coding industry, promising to provide engineers with their own automated worker to complete entire projects, rather than just snippets or suggestions.
At the moment, Devin is only available in private preview and is only accessible to a select few journalists such as Bloomberg’s Ashlee Vance.
Capabilities of Devin
Wu told Bloomberg that teaching AI to code is a complex problem that requires the system to make sophisticated decisions and anticipate future steps. According to Wu, Devin can access standard developer tools like a code editor, browser, and shell, and can run them in a sandboxed environment to plan and carry out complex engineering tasks that involve thousands of decisions.
Human users simply provide a project outline through natural language prompts using a chatbot-style interface, and Devin handles everything else, writing its own code, troubleshooting, and providing real-time progress updates. If human users notice any issues, they can add a prompt and ask the AI to fix it. This allows engineering teams to delegate tasks to Devin and focus on more creative work that is better suited to human intelligence.
Vance tested Devin by asking it to create a simple Pong-style game and a website from scratch, both of which it completed in under 20 minutes.
It can also handle more complex tasks, although they may take longer. It’s ironic that Wu, a software engineer himself, is launching a generative AI tool that aims to replace human software engineers. Computer scientist Silas Alberti, who also had early access to Devin, told Bloomberg,
“It’s less like an assistant helping with code and more like a real worker doing its own thing.”
As the era of AI progresses, the phrase “learn to code,” once suggested as an alternative for those who lose their jobs to AI, is becoming more outdated.
The creators of Devin believe that it will eventually be able to perform many low-level coding tasks instead of human coders, and much more quickly.
Cognition has recently secured $21 million in funding through a Series A round led by Founders Fund. The startup is now looking to expand its capacity and offer early access to more users, and encourages interested companies to apply via email.
Frequently Asked Questions
- What sets Cognition AI’s Devin apart from other coding tools like GitHub’s Copilot?
- Unlike other tools that assist human coders, Devin independently completes entire engineering projects, including AI models and debugging.
- How does Cognition AI’s Devin work on a given project?
- Devin uses natural language prompts from users, plans and executes tasks using a suite of developer tools, and provides real-time updates and troubleshooting.
- What kind of tasks can Cognition AI’s Devin handle?
- Devin can perform a wide range of tasks from simple projects like a Pong-style game to more complex engineering tasks, adapting to various coding requirements.
- How does Cognition AI’s Devin ensure quality and accuracy in its coding projects?
- Devin learns from educational content online to tackle tasks and allows for human intervention to debug or refine the process, ensuring high-quality outputs.
- What is the future potential of Cognition AI’s Devin in the software engineering field?
- Devin aims to take over many low-level coding tasks from human coders, promising efficiency and the capacity to handle a broad spectrum of software engineering jobs.
