SAN FRANCISCO – In a keynote address at IBM’s Think 2024 conference, IBM SVP and Director of Research Darío Gil outlined groundbreaking advancements in artificial intelligence (AI) that promise to transform how enterprises leverage technology. The event, held at the bustling Moscone Center, gathered industry leaders, tech enthusiasts, and innovators eager to hear about the future of AI from one of its most prominent voices.

“The future of AI is open,” declared Gil, emphasizing the importance of open-source innovation and collaborative efforts in the AI landscape. He urged businesses to adopt open strategies to maximize the potential of their AI systems, arguing that such an approach not only fosters innovation but also ensures flexibility and adaptability.

Embracing Open Source

“Open is about innovating together, not in isolation,” Gil said. By choosing open-source frameworks, companies can decide which models to use, what data to integrate, and how to adapt AI to their specific needs. Gil argued that this collaborative approach is essential for the evolution of AI to meet the diverse aspirations of various industries.

The strength of open source lies in its ability to foster a community-driven ecosystem where innovation can thrive unencumbered by proprietary constraints. Gil pointed to the success of IBM’s own Granite family of models, designed to handle tasks ranging from coding to time series analysis and geospatial data processing. These models, released under an Apache 2 license, provide users with unparalleled freedom to modify and improve the technology, ensuring it remains adaptable to their unique requirements.

“By leveraging open-source models, enterprises are not just passive consumers of technology; they become active contributors to a broader AI ecosystem,” Gil explained. This participatory approach accelerates innovation and ensures that AI advancements are grounded in real-world applications and challenges. The open-source community’s collaborative spirit also means that improvements and breakthroughs can be rapidly disseminated, benefiting all users.

Moreover, open-source frameworks offer a level of transparency and trust that is crucial in today’s data-driven world. Users can scrutinize the underlying code, understand the data used to train models and ensure compliance with regulatory and ethical standards. “Transparency is key to building trust in AI systems,” Gil emphasized. “When enterprises can see and understand what goes into their AI, they are more likely to embrace and deploy these technologies confidently.”

IBM’s commitment to open source is further exemplified by its contributions to major projects and partnerships within the community. The company’s involvement in the AI Alliance, launched in collaboration with Meta, brings together nearly 100 institutions, including leading universities, startups, and large-scale enterprises. This alliance aims to advance AI in a way that reflects the diversity and complexity of global societies, fostering inclusive and beneficial innovations for all.

In summary, embracing open source is not just a strategic choice for IBM; it is a fundamental philosophy that drives the company’s approach to AI. By championing open-source models and methodologies, IBM is positioning itself at the forefront of AI innovation, ensuring that the technology evolves in a way that is transparent, collaborative, and aligned with the needs of businesses and society. As Gil succinctly put it, “The future of AI is open, and together, we can build a more innovative and equitable world.”

Foundation Models: The Bedrock of AI

Foundation models have emerged as the cornerstone of modern AI, underpinning the transformative capabilities that are revolutionizing industries across the globe. In his keynote, Darío Gil underscored the significance of these models, emphasizing their role in encoding vast amounts of data and knowledge into highly capable AI systems. “The power of foundation models lies in their ability to represent and process data in previously unimaginable ways,” Gil noted. “They enable us to capture the complexity and nuance of human knowledge, making it accessible and actionable.”

One of the key advantages of foundation models is their scalability. These models can be trained on enormous datasets, incorporating a wide array of information from different domains. This scalability not only enhances their performance but also allows them to be applied to a variety of use cases. Gil highlighted IBM’s Granite family of models as a prime example, showcasing their versatility in handling tasks from natural language processing to coding and geospatial analysis. “These models are designed to be adaptable, ensuring that they can meet the diverse needs of enterprises,” he said.

The integration of multimodal data is another critical feature of foundation models. By combining information from text, images, audio, and other data types, these models can create richer and more accurate representations of the world. This capability is particularly valuable in applications such as autonomous vehicles, healthcare diagnostics, and financial analysis, where understanding the context and relationships between different data types is essential. “Multimodality is a game-changer,” Gil asserted. “It allows us to build AI systems that can understand and interact with the world in more sophisticated ways.”

Furthermore, foundation models are instrumental in democratizing AI. Providing a robust and flexible base enables organizations of all sizes to leverage advanced AI capabilities without requiring extensive in-house expertise. This democratization is facilitated by open-source initiatives, which make these powerful tools accessible to a broader audience. As exemplified by the Granite models, IBM’s commitment to open source ensures that AI’s benefits are widely shared, fostering innovation and inclusivity. “Open-source foundation models are leveling the playing field,” Gil remarked. “They empower companies to innovate and compete on a global scale.”

The potential of foundation models extends beyond current applications, promising to drive future advancements in AI. As these models evolve, they will unlock new possibilities and address increasingly complex challenges. Gil called on enterprises to actively engage in this evolution by contributing their data and expertise to enhance the models further. “The future of AI is a collaborative journey,” he said. “By working together, we can push the boundaries of what is possible and create AI systems that are more powerful, reliable, and beneficial for all.”

Foundation models represent a fundamental shift in AI technology, providing the bedrock upon which future innovations will be built. Their scalability, multimodal capabilities, and democratizing impact make them indispensable tools for enterprises seeking to harness the full potential of AI. As Gil eloquently put it, “Foundation models are not just technological advancements; they are enablers of a new era of human ingenuity and progress.”

A New Methodology: Instruct Lab

To revolutionize how enterprises interact with AI, IBM Research introduced a groundbreaking methodology called Instruct Lab. This innovative approach allows businesses to enhance their AI models incrementally, adding new skills and knowledge progressively, much like human learning. “Instruct Lab is a game-changer in the realm of AI development,” Darío Gil declared. “It enables us to teach AI in a more natural, human-like way, which is crucial for developing specialized capabilities efficiently.”

Instruct Lab stands out for its ability to integrate new information without starting from scratch, making the process both time and cost-efficient. Using a base model as a starting point, enterprises can introduce specific domain knowledge and skills, allowing the model to evolve and improve continuously. This approach contrasts sharply with traditional fine-tuning methods that often require creating multiple specialized models for different tasks. “With Instruct Lab, we can build upon a solid foundation, adding layers of expertise without losing the generality and robustness of the original model,” Gil explained.

One of the key features of Instruct Lab is its use of a teacher model to generate synthetic data, which is then used to train the AI. This process ensures that the model can learn from a broad range of examples, enhancing its ability to understand and respond to various scenarios. “Synthetic data generation is a powerful tool in our methodology,” Gil noted. “It allows us to scale the training process efficiently, providing the model with the diversity of experiences needed to perform well in real-world applications.”

The methodology also emphasizes transparency and control, ensuring that enterprises have full visibility into the training process and the data being used. This transparency is crucial for maintaining trust and ensuring the security of enterprise data. “Instruct Lab is designed with enterprise needs in mind,” Gil emphasized. “We prioritize transparency and control, allowing businesses to understand and trust the AI systems they are developing.”

The impact of the Instruct Lab is already evident in IBM’s own projects. For instance, the development of the IBM Watson X Code Assistant for Z demonstrated the methodology’s effectiveness. By applying Instruct Lab, IBM was able to significantly enhance the model’s understanding of COBOL, a critical language for mainframe applications. “In just one week, we achieved results that surpassed months of traditional fine-tuning,” Gil shared. “This showcases the incredible potential of Instruct Lab to accelerate AI development and deliver superior performance.”

The introduction of Instruct Lab represents a significant step forward in AI technology, providing enterprises with a robust and flexible tool for continuous improvement. As businesses increasingly rely on AI to drive innovation and efficiency, methodologies like Instruct Lab will be essential for staying ahead of the curve. “Instruct Lab embodies our commitment to empowering enterprises with cutting-edge AI capabilities,” Gil concluded. “It is a testament to our dedication to advancing AI in ways that are both practical and transformative.”

Scaling AI in Enterprises

Scaling AI in enterprises is not just about deploying advanced algorithms; it’s about integrating these technologies seamlessly into the fabric of the business to drive meaningful impact. Darío Gil emphasized the transformative potential of AI when it’s scaled correctly within enterprises. “The real power of AI comes from its ability to enhance every aspect of an organization,” he stated. “From optimizing supply chains to personalizing customer interactions, the possibilities are limitless when AI is effectively scaled.”

One of the critical challenges in scaling AI is ensuring that the technology is accessible and usable across various departments and functions within an organization. IBM’s approach addresses this by providing robust tools and frameworks that allow businesses to customize AI models to their specific needs. “We recognize that every enterprise has unique requirements,” Gil noted. “Our solutions are designed to be flexible and adaptable, enabling companies to tailor AI to their particular contexts and goals.”

Moreover, scaling AI requires a strong foundation of data management and governance. Enterprises must be able to trust the data that feeds their AI models, ensuring it is accurate, secure, and used ethically. IBM places a strong emphasis on data governance as a cornerstone of its AI strategy. “Data is the lifeblood of AI,” Gil explained. “Without proper governance and management, the insights derived from AI could be flawed. We provide comprehensive tools to help enterprises manage their data effectively, ensuring that their AI initiatives are built on a solid foundation.”

To truly scale AI, enterprises must also invest in the continuous training and development of their workforce. AI is not a set-it-and-forget-it solution; it requires ongoing learning and adaptation. IBM supports this through its extensive training programs and resources, helping organizations develop the skills needed to harness the full potential of AI. “Human expertise is essential in driving AI success,” Gil said. “We are committed to empowering our clients with the knowledge and skills they need to excel in an AI-driven world.”

Additionally, IBM’s focus on open-source models plays a crucial role in scaling AI. By leveraging open-source technologies, enterprises can benefit from a collaborative approach to AI development, accessing a wealth of community-driven innovations and best practices. “The open-source community is a vital component of AI advancement,” Gil highlighted. “It fosters a spirit of collaboration and continuous improvement, essential for scaling AI effectively across enterprises.”

As enterprises navigate the complexities of scaling AI, IBM’s comprehensive approach—spanning advanced technologies, robust data management, continuous learning, and open-source collaboration—provides a clear pathway to success. “Scaling AI is a journey,” Gil concluded. “It’s about creating a sustainable, adaptable framework that grows with the enterprise, driving innovation and competitive advantage at every step.”

Looking Ahead

As IBM continues to push the boundaries of AI, the future holds immense potential for enterprises willing to embrace these transformative technologies. Darío Gil’s vision for AI is one where innovation and collaboration drive progress, ensuring that AI serves not just as a tool for efficiency but as a catalyst for groundbreaking advancements across industries.

One of the key areas of focus for IBM moving forward is the integration of AI with other cutting-edge technologies, such as quantum computing and blockchain. “The convergence of AI with quantum computing can unlock new levels of problem-solving capabilities that were previously unimaginable,” Gil noted. “By combining the strengths of these technologies, we can tackle some of the most complex challenges facing humanity, from climate change to healthcare.”

IBM is also committed to ensuring that AI development remains ethical and inclusive. The company is actively working on initiatives to address biases in AI models and to promote transparency and accountability in AI systems. “As we look ahead, it’s crucial that we build AI that is fair, transparent, and respects the values of our society,” Gil emphasized. “We are dedicated to leading the charge in creating ethical AI frameworks that benefit everyone.”

In enterprise applications, IBM plans to expand its portfolio of AI-driven solutions, providing businesses with even more tools to enhance their operations and drive innovation. The company’s continued investment in research and development ensures its clients have access to the latest advancements in AI technology. “Our goal is to empower enterprises to leverage AI in ways that were previously thought impossible,” Gil said. “We are constantly exploring new frontiers and developing solutions that will keep our clients at the forefront of their industries.”

Moreover, IBM’s commitment to open-source AI models will play a significant role in the future of AI development. By fostering a collaborative environment, IBM aims to accelerate the pace of innovation and ensure that AI technology evolves in a way that is beneficial for all stakeholders. “The future of AI is one that is built on collaboration and shared knowledge,” Gil stated. “By embracing open-source principles, we can create a thriving ecosystem where everyone has the opportunity to contribute and benefit from AI advancements.”

As the landscape of AI continues to evolve, IBM remains steadfast in its mission to drive technological progress while addressing its ethical and societal implications. “The road ahead is full of exciting possibilities,” Gil concluded. We are committed to leading the way in AI innovation, ensuring that our advancements serve the greater good and pave the way for a better future for all.”

With a forward-looking approach that combines technological excellence, ethical considerations, and a collaborative spirit, IBM is well-positioned to shape the future of AI and drive meaningful change across the globe. As enterprises prepare to navigate this dynamic landscape, they can look to IBM for guidance, support, and innovative solutions to help them thrive in the age of AI.