Google unveils OS for AI devices

Google has unveiled the KataOS operating system for embedded systems on RISC-V chips that process AI workloads.

In building the OS, the developers used the seL4 security‑focused microkernel. It is open source, but not Linux-based.

[on seL4] mathematically proven security with guaranteed confidentiality, integrity and availability,\” Google said.

They added that KataOS applications cannot breach the kernel’s hardware protections, which they described as providing reliable protection for the system components.

KataOS is almost entirely implemented in Rust, which provides a solid starting point for software security by eliminating entire classes of bugs such as \”one-after-another\” and buffer overflows,\” the developers said.

Google also released a reference version of Sparrow. The project remains under development.

We are pleased to share some initial details and invite others to collaborate on the platform so that we can all build intelligent ambient systems with security built in by default,\” Google said.

The developers noted that the OS is not intended for desktops or smartphones, but for the Internet of Things and smart homes. One of the project’s main aims is to create a secure system for embedded or peripheral devices that process data on-device or in the cloud.

The devices they assemble can expose personal data such as people’s images and voice recordings to malware,\” AmbiML specialists noted.

The GitHub repository contains frameworks for Rust, an alternative root server, and a modification to the seL4 kernel to reclaim memory used by the root server.

Sparrow includes a logically secure root of trust built with OpenTitan on the RISC-V architecture. However, in the first version we are targeting the standard 64-bit ARM platform running in a QEMU simulation,\” Google notes.

Back in July 2021, OpenAI released programming language for developing neural networks.

Follow ForkLog AI on Telegram: ForkLog AI — all the news from the world of AI!