Google Sets Project Sparrow Free, Announces Open Se Cura as a Framework for Efficient, Scalable AI

Google has announced the opening of its Project Sparrow, now known as Open Se Cura, as the basis for the development of "secure, scalable, transparent, and efficient" artificial intelligence (AI) systems — with a complete dump of the source code to start.

"Our goal with Open Se Cura is to evolve a set of open source design tools and IP [Intellectual Property] libraries that will accelerate the development of full-stack systems with ML [Machine Learning] workloads through co-design and development," claims Google's Kai Yick, a member of the Google Research Cerebra Open Source team, of the project. "This will enable us to better center system designs around security, efficiency, and scalability, empowering the next generation of AI experiences."

The problem, as Google sees it, is that artificial intelligence is becoming more and more entwined in people's daily lives — but while the software side is evolving rapidly, the hardware is lagging. The result is poor efficiency and portability, with modern machine learning models requiring vast computational resources and being wholly unsuited to deployment to low-resource or battery-powered devices.

Project Open Se Cura aims to offer an open source framework to solve these issues, with input from a number of Google's partners: lowRISC, Antmicro, and VeriSilicon, who provided a transparent root-of-trust from the RISC-V-based OpenTitan project, the Renode system simulation toolchain, and IP design expertise respectively. The project has already shown results, Google claims: "we’ve used these tools for the first time to extend our IP library with secure ML capabilities," Yick writes, "and generated a proof of concept for a low-power AI system."

The company isn't looking to keep all the project's potential to itself, though, hence its evolution from the closed Project Sparrow to Open Se Cura. "Going forward, we’ll continue to evolve Open Se Cura in the open," Yick claims, "and seek to onboard additional partners, such as Cambridge University (for CHERI innovations) and the University of Michigan (for low-power and generative AI). We are excited to explore what we can build with these new tools and hope the open source community will join us and contribute."

The source code for Project Open Se Cura has been released on Google Source, under the Apache 2.0 license and other inherited licenses, with instructions on its use; a "technical deep dive" is scheduled for 9:10 AM PST on November 9 as part of the CHIPS Alliance Technology Update, with virtual attendance possible via the event website.