Meet Devin, the world’s first AI software engineer that writes, debugs, and deploys code – Moneycontrol

author
2 minutes, 53 seconds Read
Devin

In a demonstration shared by the company, Devin showcases its ability to add a print statement when encountering an error, swiftly fixing the issue.

US-based startup Cognition has unveiled Devin, an artificial intelligence (AI) software engineer that promises to revolutionise the way code is written, debugged, and deployed.

The AI assistant is being hailed as the world’s first AI of its kind because it can take a simple prompt and turn it into a functioning website or software programme.

Story continues below Advertisement

The Founders Fund-backed startup claims that the AI assistant 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.

SWE-bench is an evaluation framework that encompasses 2,294 software engineering problems extracted from authentic GitHub issues and their corresponding pull requests within 12 popular Python repositories.

Let’s dive into the capabilities of this latest AI model:

What is Devin?

Devin is a software development assistant that surpasses mere code completion and suggestions. Unlike large language models (LLMs) such as OpenAI’s ChatGPT or Google’s Gemini, Devin can autonomously handle the entire process of building and launching full-fledged software applications.

Also read: What powers ChatGPT and Bard? A look at LLMs or large language models

Story continues below Advertisement

Equipped with its own command line, code editor, and browser, Devin operates independently, systematically implementing software by breaking down tasks step by step.

What is Devin capable of?

Devin can plan and execute complex engineering projects, involving thousands of decisions. It learns from its experiences, recalls relevant context, and even fixes its own mistakes.

Devin

In a demonstration shared by the company, Devin showcases its ability to add a print statement when encountering an error, swiftly fixing the issue.

A print statement is a line of code used to output messages to the console, helping programmers pinpoint where an issue arises within the code. Devin can also actively collaborate with the user. It reports on its progress in real time, accepts feedback, and works collaboratively with the user on design choices as needed.

The AI assistant can also learn unfamiliar technologies quickly after reading an article. It is even capable of training and fine-tuning large language models by drawing instructions from research repositories.

What about performance?

While the company has not yet released a detailed technical report, it claims that Devin has superior coding skills, solving a significantly higher percentage of coding problems compared to other AI models. Evaluated on the challenging SWE-bench coding benchmark, Devin successfully resolves real-world GitHub issues found in open-source projects.

Devin2

It demonstrated a remarkable 13.86 percent issue resolution rate, surpassing competitors such as Anthropic’s Claude 2 at 4.8 percent and GPT-4 at 1.8 percent.

Who are the startup’s backers?

Founded by Scott Wu, Steven Hao, and Walden Yan, the AI startup has raised $21 million in a Series A funding round led by Peter Thiel’s venture capital firm Founders Fund. Other investors include former Twitter executive Elad Gil, VC firm Conviction founder Sarah Guo, among others.

The 10-member founding team, all of whom hold gold medals from the International Olympiad for Informatics, have worked at the cutting edge of applied AI with companies like Cursor, Scale AI, Lunchclub, Modal, Google DeepMind, Waymo, and Nuro.

Event alert: Moneycontrol and CNBC TV18 is hosting the ultimate event on artificial intelligence, bringing together entrepreneurs, ecosystem enablers, policymakers, industry leaders, and innovators on March 27 in Pune. Click here to register and gain access to the AI Alliance Pune Chapter.

This post was originally published on this site

Similar Posts