Intel’s recent struggles highlight the highly competitive nature of the silicon ecosystem, but those struggles have not deterred the chip giant from its long-standing approach in embracing open-source ethos that allow for deep collaboration with rivals, efforts that are becoming increasingly targeted as the segment rides the current artificial intelligence (AI)-powered development wave.

Melissa Evers, VP and GM at Intel, explained in an interview with SDxCentral that some of the company’s most recent open-source work has been through the Open Platform for Enterprise AI (OPEA) project. That project, which is part of the Linux Foundation, is targeted at providing an open-source framework of composable building blocks for generative AI (genAI) systems, including large language models (LLMs), data stores, and prompt engines.

“What we see happening with generative AI is a similar storming that has occurred in open source in the past, where there's a lot of innovation that's happening very, very rapidly, and folks are stifled,” Evers said. “The innovation and adoption of all of these technologies is stifled due to the complexity and pace of innovation.”

Evers specifically cited that accelerated innovation pace with new the launch of new models, new vector database projects, and new APIs.

“What we see and hear from our customers is that the pace of innovation as well as the complexity of constructing generative AI RAG [retrieval-augmented generation] pipelines is challenging and requires specific levels of expertise that many customers don't have,” Evers said. “The talent is very expensive, it's hard to find, and there's a risk of obsolescence once you build something.”

Evers did add that OPEA is geared toward more “generic implementations,” which leaves the door open for other ways to target more “specific verticals” where there is more “complexity and a lot of nuance that can be applied.”

Evers explained that these include medical use cases in regard to private and security; industrial use cases in regards to “integration of both vision and text formats of data and being able to assimilate both.”

“There's some really interesting further innovation and refinement that I believe will happen in the open-source ecosystem to service or vertical markets,” Evers said.

However, that innovation and refinement can lack a proper open-source platform channel to help drive a cohesive approach. Evers cited similar work being done by EdgeX Foundry for retail use cases and the AI-focused retrieval-augmented generation (RAG) pipeline framework.

“If you merged the two, you could look at some really interesting combined use cases with various stages of RAG feeding or vision models feeding RAG pipelines for refinement,” Evers explained.

Open source communities have been working through these types of mergers in order to simplify their increasingly complex project structures.

Intel’s OpenSSF view Intel has a more long-term view on the Open Source Security Foundation (OpenSSF). That Linux Foundation-based group was formed in 2020, and is focused on developing security tools and practices for open source application development without bias toward a specific ecosystem or vendor.

“We have taken kind of a long view with regard to our OpenSSF engagement,” Evers said of Intel’s efforts with that group. Evers explained that Intel has employees that are part of OpenSSF’s governing board and touted the group’s efforts around solidifying a secure supply chain “not only in creating the framework for that development, but also engagement with various government institutions and helping to advance the role of open-source security globally.”

Ryan Ware, director of open source security at Intel, recently penned a blog post on Intel’s efforts in using OpenSSF’s Scorecard framework that highlighted potential challenges with those efforts when it comes to the open source world.

“Given the massive scale of Intel’s open-source ecosystem, it’s not as straightforward as running OpenSSF Scorecard on a repository, it can be a lot more challenging than expected,” Ware wrote of one of those challenges. “Intel’s challenge was not just about the quantity but also understanding the diversity and uniqueness of each repository, across our many GitHub organizations.”

Open source for AI collaboration and trust That diversity is becoming increasingly complex in the world of AI, which efforts like OPEA show will require deep collaboration.

Evers explained that a lot of Intel’s collaboration in the space is around next-generation workloads. This includes work on AI workloads “and being able to have finer-grained telemetry at the GPU level. That’s a need if you are running an Intel GPU or an AMD GPU or a Nvidia GPU.”

These efforts involve downstream work with those rivals to provide that solid, open-source base that they can all then use to build on top of in providing differentiated upstream services to end users.

“We're going to do the extra step of enabling that within the upstream communities to ensure that when it is pulled downstream, folks can harness the value that is rendered in our silicon, and that we, because it's open source, it's done transparently,” Evers said. “But, in the same context, this is the work that our competitors do as well to drive differentiation of their capabilities.”

The timing of this AI-focused open-source work is crucial as the market continues to froth over AI opportunities and further highlights the need for nailing down the basics of AI trust and security at the lowest levels of the hardware stack.

“In the context of AI with regard to trust reliability, we do face challenges as an industry and Intel sees those and faces those too, and we believe that we have a role to play, a responsibility in the context of the future of those open ecosystems and ensuring the power that they can render can be built upon and utilized, because we are actively addressing some of these systemic concerns and issues that the industry faces,” Evers said.