- In a new study, researchers tasked an AI-powered tech company with developing 70 different programs.
- They found that AI can develop software in 7 minutes, at an average cost of less than $1.
- AI bots were assigned roles and were able to speak, make logical decisions and troubleshoot bugs.
AI chatbots like OpenAI’s ChatGPT can help software companies operate in a fast, cost-effective manner with minimal human intervention, a new study has found.
The findings come after researchers published another study in which AI agents powered by large language models were able to run a virtual city on their own.
In a recent paper, a team of researchers from Brown University and several Chinese universities conducted an experiment to see if AI bots powered by the 3.5 model version of ChatGPT could complete the software development process without prior training.
To test this, the researchers created a fictional software-development company called ChatDev. Based on the waterfall model — a sequential approach to building software — the company was chronologically divided into four distinct phases: designing, coding, testing, and documentation.
From there, the researchers assigned specific roles to the AI bots by indicating “critical details” that describe “assigned tasks and roles, communication protocols, termination criteria and limitations.”
Once the researchers assigned their roles to the AI bots, each bot was allocated to their respective stages. For example, ChatDev’s “CEO” and “CTO” worked under the “Designing” stage, and “Programmer” and “Art Designer” worked under the “Coding” stage.
During each stage, AI workers chat with each other with minimal human input to complete specific parts of the software-development process—from deciding which programming language to use to identifying bugs in the code—until the software is complete.
The researchers ran the experiment in different software scenarios and applied a series of analyzes to see how long it took ChatDev to complete each type of software and how much each would cost.
The researchers, for example, tasked ChatDev with designing a “basic Gomoku game,” an abstract strategy board game also known as “Five in a Row.”
At the designing stage, the CEO asked the CTO to propose a solid programming language that would “meet the new user demand”, to which the CTO responded with Python. In turn, the CEO said, “Great!” and explained that the programming language’s “simplicity and readability make it a popular choice for both novice and experienced developers.”
After the CTO replied “let’s get started”, ChatDev moved to the coding stage, where the CTO asked the programmer to write the file, then the programmer asked the designer to give the software a “beautiful graphical user interface”. The chat chain is repeated at each stage until the software is developed.
After assigning ChatDev 70 different tasks, the study found that the AI-powered company could complete the entire software development process “for less than a dollar in seven minutes” — all while “identifying possibilities and troubleshooting” with its “memory” and “self-reflection.” Vulnerability through capabilities.
The study reported that 86.66% of the generated software systems were executed flawlessly.
“Our experimental results demonstrate the efficiency and cost-effectiveness of the automated software development process driven by CHADTEV,” the researchers wrote in the paper.
The researchers did not immediately respond to Insider’s request for comment ahead of publication.
The study findings highlight one of the many ways that powerful generative AI technologies like ChatGPT can perform specific job functions. Since the AI chatbot came out last November, workers across industries have used it to save time and increase productivity.
Coders, in particular, may find generative AI tools beneficial to their personal and professional lives. Berlin-based coder Daniel Dippold used ChatGPT to develop a program that helped him find apartments and discovered that Amazon employees were using ChatGPT for software development.
Still, the study isn’t perfect: the researchers identified limitations, such as errors and biases in the language model, that can cause problems in the creation of software. Still, the researchers said the findings “could potentially help junior programmers or engineers in the real world”.
NOW WATCH: Top Videos from Insider Inc.