In an era where Artificial Intelligence (AI) is increasingly integral to technological advancements, making it accessible and equitable is paramount. Open source AI tools emerges as a pivotal strategy to achieve this, as highlighted in a discussion with Marshal Correia, Vice President and General Manager, Red Hat India /South Asia. Open source development thrives on principles of cross-industry collaboration, transparency, and community-driven innovation. This approach has historically driven significant technological breakthroughs, from blockchain and cloud computing to the Internet of Things (IoT) and machine learning (ML). Correia accentuated that an open, transparent, and responsible AI can address global challenges, such as revolutionizing healthcare, enhancing environmental sustainability, and improving education. By democratizing AI, open-source communities are working towards greater accessibility, sustainability, and trust. In this interview, Correia elaborates on how Red Hat’s collaborative efforts and innovative initiatives like InstructLab are making AI more inclusive and equitable for everyone.
TAM: How do open-sourcing AI tools contribute to making AI more accessible and equitable?
Marshal Correia: Open source thrives on principles of cross-industry collaboration, transparency, meritocracy, and community-oriented development. It has driven numerous technological advancements, including blockchain, cloud computing, Kubernetes, IoT, AI, and ML. AI offers immense opportunity for open research and innovation. An open, transparent, and responsible AI can address many global challenges like revolutionising healthcare, risk mitigation in FSI, enhancing environmental sustainability, improving education and many more. Its data processing capabilities enable innovative solutions, driving positive change across various sectors. Open source community through various projects is already working to bring greater accessibility, more sustainability, and enhanced security and trust to AI for the future so that everyone can benefit from AI and everyone should be able to access and contribute to it. With Red Hat’s enterprise-ready AI platforms like Red Hat Enterprise Linux AI, Red Hat OpenShift AI, InstructLab, organisations build, develop and use AI-enabled applications to operate more efficiently, drive deeper customer engagement and deliver better products and services.
TAM: What opportunities do you see for collaboration between Red Hat and other organisations in the AI space to further democratise AI?
Marshal Correia: AI solutions are not “one size fits all,” and no single vendor can meet every customer’s needs alone. Red Hat’s AI vision relies on the support and active collaboration of our global partner ecosystem. Our dynamic partner ecosystem, encompassing digital labour, infrastructure, conversational AI, and more, is well-equipped to address the entire range of AI use cases for our customers.
Red Hat’s collaborative relationships with hardware and software partners enable flexible AI model deployment, assisting customers in building comprehensive AI solutions tailored to their unique use cases. This includes everything from data acquisition and preparation to monitoring, maintenance, and hardware acceleration. For instance, we are working with hardware vendors and GPU providers like AMD, Intel, NVIDIA, and DELL to offer customers more options for GPU resources and AI-optimised hardware. We are also partnering with several ISVs such as Elastic, Run:ai and Stability AI to make it easier for customers to access essential AI capabilities directly from their Red Hat OpenShift AI console. Additionally, Red Hat’s extensive catalogue of certified AI/ML partner applications can be natively integrated within Red Hat OpenShift, providing complementary or extended capabilities on Red Hat OpenShift AI. In collaboration with IBM Research, we are open-sourcing several models for both language and code assistance.
TAM: What motivated Red Hat to announce InstructLab, and how do you envision this move impacting the AI landscape?
Marshal Correia: At Red Hat, we believe that AI innovation should not be confined to large organisations with extensive resources, such as massive Graphic Processing Unit (GPU) farms or extensive data science teams. This belief drives our AI product strategy. Just as we have leveraged the power of open source for Linux, Kubernetes, and hybrid cloud computing for enterprises, we are now applying the same principles to AI. Our strategy goes beyond merely providing the infrastructure for AI-enabled applications; we aim to embed the power of community and open source into the models themselves.
InstructLab, our new joint open-source community project with IBM, is designed to democratise AI contributions. It offers a cost-effective solution for improving the alignment of Large Language Models (LLMs) and opens the door for individuals with minimal machine learning experience to make meaningful contributions. Instead of forking an LLM—which involves creating a separate and independently developed version that leads to a closed environment with limited contributions—InstructLab enables anyone globally to add knowledge and skills. These contributions can be integrated into future model releases. In other words, you don’t need to be a data scientist to contribute to InstructLab. Domain and subject matter experts, as well as data scientists, can use InstructLab to make contributions that benefit everyone. This approach is incredibly powerful for both the community and enterprises.
TAM: In what ways does InstructLab enable individuals, regardless of their technical background, to contribute to and train large language models?
Marshal Correia: InstructLab is all about making large language models (LLMs) more accessible, empowering people from all walks of life to contribute, regardless of their technical background. But how does InstructLab differ from traditional LLM training? Traditionally, training LLMs involved complex coding and manipulating massive datasets, but InstructLab flips the script by using a method called Large-scale alignment for ChatBots (LAB). Instead of writing code, users provide clear instructions and examples for the LLM to learn from. This natural language interface allows anyone comfortable with giving instructions to contribute training data, whether they’re artists describing styles or historians providing factual examples. InstructLab doesn’t require expertise in AI/ML; contributions can be specific and focused, enabling individuals with niche knowledge to add valuable insights. The platform fosters a collaborative environment where diverse perspectives shape the LLM, leading to more well-rounded models. InstructLab removes technical hurdles, empowering everyone to be a teacher for the next generation of Open source LLMsI. By providing clear instructions and examples, people can contribute to developing powerful LLMs without extensive coding knowledge, opening doors for a more inclusive and diverse AI landscape where everyone has a voice in shaping the future of intelligent machines.
TAM: What are the most significant challenges Red Hat has faced in democratising AI through open-source initiatives like Instruct Lab?
Marshal Correia: While InstructLab represents a significant step forward in democratizing AI, we are actively addressing several challenges. Ensuring the quality and fairness of the data used to train AI models is a major hurdle, as bias in training data can result in biassed AI outputs. While InstructLab’s broad contributor base enhances diversity, we need robust mechanisms to ensure the data remains accurate and unbiased. Building a large, diverse community is exciting but requires effective governance, including clear guidelines for contributions, moderation processes to address potential issues, and a framework to ensure everyone feels valued and respected. Open source offers transparency but also means potential vulnerabilities can be more easily identified; hence, we are establishing clear security best practices and collaborating with the community to proactively identify and address potential security risks. Defining and measuring success in a democratized AI environment is another challenge, as it’s not just about technical metrics but also the inclusivity and impact of the AI models being developed. We are exploring new ways to measure the success of open-source AI projects that consider broader social and ethical considerations. These are just some of the challenges we face, but at Red Hat, we believe the potential benefits of democratizing AI far outweigh them. We are committed to continuous improvement and working with the community to ensure that open-source AI is not only accessible but also responsible, secure, and beneficial for everyone.
TAM: How is Red Hat addressing potential ethical and security concerns associated with open-sourcing AI technologies?
Marshal Correia: Our approach is rooted in the core principles of open source: collaboration, transparency, and community-oriented development. We are committed to responsible AI development, ensuring fairness, accountability, and transparency throughout the AI lifecycle. Through cross-industry collaborations like the AI Alliance with IBM, we work to define open-source AI governance and best practices on an industry scale.
Security is paramount in our approach. We implement robust measures to minimise vulnerabilities in our AI models and encourage community vigilance in reporting concerns. Transparency and explainability are key to building trust, which is why we are dedicated to developing AI models that provide clear insights into their decision-making processes.
Our commitment to transparency extends to the AI supply chain. We leverage projects like Sigstore to verify open-source code integrity and pioneer an AI Bill of Materials, ensuring that our AI technologies are not only innovative but also ethically sound and secure. The AI landscape is rapidly evolving, and so is our approach. We are committed to continuous improvement, actively researching new techniques and best practices. A crucial aspect of this is ensuring a secure AI supply chain, providing a clear understanding of AI components’ origins and handling.
At Red Hat, we believe that responsibly open-sourcing AI can address global challenges across various sectors. Our enterprise-ready AI platforms enable organisations to build and use AI-enabled applications efficiently, driving deeper customer engagement and delivering better products and services. By prioritising ethical development, robust security measures, and supply chain integrity, we work to ensure that the benefits of AI are accessible and equitable for all, fostering innovation while maintaining the highest standards of ethics and security.
