As artificial intelligence (AI) continues to develop, so does the fear of it Thousands of jobs will be replaced through automated systems. This has already happened in various guilds and, according to the results of current studies, the nightmare now seems to have finally reached the software developers.
Scientists at Brown University in Providence, Rhode Island, concluded that ChatGPT has the in its 3.5 model Ability to develop software without the need for training, making it able to run a development company with almost no human intervention.
An “interview for the manager position” at ChatGPT
For research purposes, the scientists founded ChatDev, a virtual development company on which the AI would work. To give you a more realistic approach, The company was divided into four parts: Design, coding, testing and documentation.
Functions were assigned for each phase and position, with one common instruction: the AIs should regularly report what they were working on. This way Communication was achieved between theEmployees” from different areas. While one AI group dedicated itself to design, the other used the reported information to advance in the coding section.
When this chain runs, Human intervention was minimal, as only the language to be used had to be clarified. The AIs responded to specific requests in an agile and organized manner. In one of the exercises they were asked to design a board game called Gomokua Japanese version of the “three in a row“O”Cat game“.
The process was completed in just seven minutes., but not without leaving an extensive internal chat line between the various “employees in each area”. For example, leading AI asked designers to do tasks like “a nice graphical user interface“.
After 70 similar tests, the results were analyzed: 86.66% of the software developed by AIs was able to run successfully and all this with a minimal investment outlay of around one euro per task during the changeover.
“Our experimental results demonstrate the effectiveness and cost-effectiveness of the automated software development process supported by ChatDev.” reads in the studio.
Months ago, researchers at Stanford University, California, conducted another experiment to measure internal communication between AIs by placing them in a controlled environment: 25 avatars were designed for a virtual city and they were allowed to interact.
The results were also overwhelming, as were the avatars They achieved an organized society in a short time, from electing the president to organizing celebrations like Valentine’s Day to a schedule of activities. All without a single human interaction. This study served as a precedent to better understand the level at which AI imitates human behavior.
Despite the results of both studies The researchers noted important limitations: There are biases in linguistic models that hinder internal communication between AIs, which inevitably leads to errors in the final results.
The intention is to use this research as Support tools for software developers but many speculate and fear that they will be replaced once AIs reach a level of precision equal to or exceeding that of humans, as appears to be the case in other disciplines.