A study conducted by Brown University has surprised the technology community by assigning an artificial intelligence model, ChatGPT 3.5, the role of CEO in a virtual software development company called CHAT DEV. This experiment was divided into four key phases: design, coding, testing and documentation.
The result was impressive because artificial intelligence managed to develop a board game called Gomoku in just seven minutes and for around one euro. In addition, 86.66% of the projects developed were executed without errors, demonstrating the efficiency and precision of AI in this context.
The importance of this experiment lies in the impact that artificial intelligence can have on software development. AI’s ability to run a software company, make logical decisions, assign tasks, and correct errors with little human intervention marks a milestone in the evolution of software development.
In addition to speed and lower costs, this experiment also shows how AI can be a valuable tool for identifying and remediating vulnerabilities in software, which is critical to cybersecurity.
The ChatGPT model 3.5 is an extended version of the GPT model series developed by OpenAI. It features a transformer architecture and millions of trainable parameters that allow it to perform natural language processing tasks with remarkable accuracy and efficiency. Thanks to improvements in context understanding, text generation and decision making, this model is capable of solving more complex problems.
The technological core of ChatGPT 3.5 is based on technologies such as Transformer architecture, supervised training, mechanized support, multilingual features, APIs and SDKs. Together, these technologies make ChatGPT 3.5 a powerful tool not only for natural language processing tasks, but also for project management in software development.
The experiment was designed by researchers at Brown University who founded a virtual software development company called CHAT DEV. This company simulated a real-world environment with defined organizational structures and specific roles assigned to different AI instances. Through a chat system, ChatGPT instances collaborated in real time to make decisions, troubleshoot errors, and coordinate the software development process.
This revolutionary experiment demonstrates the potential of artificial intelligence in the software development industry and opens new perspectives for automation and efficiency in development cycles.