LOCI 2.0 detects emerging software anomalies and trends, provides guidance on the progress of project branches for seamless integration, and helps develop accurate, targeted tests, improving system quality, reliability, and compatibility
Aurora Labs announced a technical preview of the new extended version of LOCI (Line-of-Code Intelligence™), the first AI Advisor Engineer with advanced prompting capabilities. Aurora Labs is building a future where anyone can be an expert developer or tester and make reliable, high-quality software using a proprietary Large Code Language Model (LCLM). Designed with the insights and needs of software developers and embedded systems engineers in mind, LOCI enhances software reliability, quality, and predictive software maintenance across a wide range of platforms including cloud, mobile applications, HPC, and embedded systems. LOCI supports diverse hardware from suppliers such as Infineon, NVIDIA, NXP, Qualcomm, Renesas, Samsung, and STMicroelectronics. A special free early access offer for engineers will be available for the first batch of users before the general availability in 2024.
LOCI 2.0 extends the capabilities of GitHub Copilot. After coding with or without a Copilot-like tool, LOCI 2.0 advises on quality, reliability, and compatibility issues enabling the developers and testers to meet the defined KPIs. Deployed in highly demanding industries, such as Automotive, LOCI 2.0 features are compatible with the highest-quality processes for critical safety software projects.
LOCI is equipped with Aurora Labs LCLM, a Large Code Language model, that analyzes software artifacts and transforms complicated information into meaningful insights. Unlike existing Large Language Models, LOCI 2.0 tokenizers are more productive and efficient because they are x1000 smaller and have reinvented the vocabulary and pipeline training using only 6 GPUs. LOCI trained on real-time tracing of different systems-on-chip, starting in early 2017, and created billions of ASM tokens. LOCI has been trained on more than 5 billion lines of code in various languages, such as C and C++. The advanced model has an error prediction mechanism to ensure maximum model quality.
The cutting-edge AI Advisor brings transformative capabilities to software development. LOCI predicts the impact of project branching on software behavior, compares integration efforts of different pull requests, and detects dead code while providing evidence of SW behavior deviations for root-cause analysis, even without access to the source code. LOCI identifies cloned code across multiple repositories, creating a clear view of current and older systems for maintenance. This helps developers and testers understand system functionality and dependencies.
Also Read: Compass UOL Transforms Software Delivery Process with Amazon Q Integration
LOCI identifies the parts of the code that have changed to simplify the software update file creation and installation, ensuring seamless software updates. These features empower developers, testers, and IT managers to enhance software reliability, quality, and performance of over-the-air (OTA) updates. LOCI 2.0 creates software binary images and delta files for updates, improving system stability and compatibility, reducing data size, extending storage life, and minimizing device downtime for software updates. LOCI assists with recommendations for the risk management of SW updates by providing predictive insights into compatibility and performance degradation issues when introducing new and updated code or images.
The LOCI Advisor Engineer brings attention to deviations in the running software that are invaluable to intrusion detection systems. Suspicious software activity creates deviations in the software behavior that LOCI identifies using a behavioral software protection signature while continuously monitoring the software during the lifecycle of the project.
“Aurora Labs started this journey seven years ago, using natural language processing (NLP) to understand machine codes. We’ve reached a significant milestone by incorporating a transformer model, adding proprietary layers, and reinventing tokenizers and vocabulary. This sets the stage for a future where software engineering goes beyond traditional limits, improving capabilities and smoothly integrating with the latest technologies to boost software quality and reliability,” said Zohar Fox, CEO of Aurora Labs.
Thomas Schneid, Senior Director of Software, Partnership & Ecosystem Management at Infineon Technologies, reflected on their experience: “LOCI 2.0 is designed for software engineers and impressively supports complex software systems. Improving system quality, reliability, and compatibility are our main takeaways from the impactful data provided by LOCI 2.0. As a leader in automotive semiconductors, we are very happy to offer our AURIX™ TC4x family of microcontrollers (MCUs) with LOCI 2.0. This innovative solution will empower automotive manufacturers to deliver safer, more reliable vehicles and enhance performance throughout the vehicle’s lifetime.”
Source: PRNewswire
