Apple just published three new research papers detailing how it’s using artificial intelligence in key areas of software development and coding. This research shows that Apple is building artificial intelligence agents to automate processes that have always been time-consuming, costly and prone to human error.
In a new piece of research, Apple is focusing on one of the biggest bottlenecks in software development: quality assurance (QE). Traditionally, quality assurance engineers spend 30-40% of their time manually writing tests and automation scripts. To solve this problem, Apple has developed a multi-layer framework called “Agentic RAG”.
Instead of an engineer, the system uses six artificial intelligence agents, each with a specific task: one agent is responsible for ensuring compliance with regulations; Another agent measures previous tests to learn patterns. A third agent creates new tests based on current methodologies. The fourth agent manages conflicts and conflicts, and the other two agents establish communication between modules and systems.
The results of this approach are stunning. This system has been able to increase the accuracy from 65% to 94.8%, reduce the work time by 85% and improve the quality of defect identification by 35%.
Apple research to use artificial intelligence in development and bug hunting
Apple’s second investigation focuses on another issue: fixing bugs in the code. For this purpose, researchers have created a special training environment called “SWE-Gym”.
It’s the “gym club” for AI agents, with 2,438 real software engineering tasks pulled directly from GitHub issue reports in 11 popular Python repositories. In this environment, the AI agent must learn to solve these real-world problems using the available tools. This process allows the model to improve its debugging capabilities through trial and error.
The results show that the language models trained with this method managed to solve 72.5% of the tasks correctly, which is a very strong result and has great potential to increase the productivity of developers.
Apple’s third article is very interesting; Apple has explained how it wants to predict bugs before the development process starts, rather than finding them. This research introduces a new and complex model called “ADE-QVAET”. By combining advanced optimization techniques and quantum transformer models, this model learns to identify patterns that cause software bugs.
Altogether, these three articles show that Apple’s focus in the field of artificial intelligence is not only on capabilities such as Apple Intelligence, but that the company is seriously using artificial intelligence to improve and accelerate its internal engineering processes. While it’s unclear whether these technologies will make their way into developer products like Xcode, the possibility doesn’t seem far-fetched.
Latest Passing over countries : Spain | Dominica | United Arab Emirates
RCO NEWS
