Digest

This episode of Smart Talks with IBM delves into the world of open AI, featuring Maryam Asuri, Director of Product Management and Head of Product for IBM's WatsonX.ai. The conversation explores the concept of "open" in the context of AI, highlighting how openness is driving innovation and reshaping industries. Asuri discusses the use of foundation models, particularly IBM's Granite, in enterprise AI applications. She explains the transformative power of generative AI and how it's shifting the market from experimentation to production. The episode also addresses the challenges of large models, such as increased latency and cost, and the emerging trend of smaller, customized models. Asuri introduces Granite, IBM's family of foundation models, which are trained from scratch and offered with indemnification. She details the different flavors of Granite models, including language, code, time series, and geospatial models, and provides examples of their real-world applications in customer care and content generation. The conversation explores the concept of foundation models and their potential risks due to lack of transparency in training data. Asuri emphasizes the importance of client protection and the need for trustworthy governance in AI models. She explains how Granite addresses these concerns through its training process and the Watson X platform. The episode delves into IBM's multi-model, multi-deployment strategy for AI, emphasizing the use of both IBM-developed models like Granite and third-party models. Asuri highlights the importance of offering customers a wide range of models and the ability to bring their own models to the platform. She discusses the future of AI, predicting a shift towards generative AI that can reason, plan, act, and reflect. She highlights the development of function calling capabilities in Granite models, enabling them to interact with external tools and perform complex tasks. Asuri provides advice for buyers and users of AI tools, emphasizing the importance of identifying the specific area where AI can benefit, assessing data availability and quality, finding a trusted partner, evaluating regulations, and applying design thinking principles. She advises project managers on how to navigate the fast-paced world of AI innovation, emphasizing the need for agility, experimentation, continuous learning, and rapid adoption of new ideas. The episode concludes with Asuri summarizing the key features of Granite, highlighting its openness, trustworthiness, and targeted approach for enterprise applications. She emphasizes the importance of smaller, domain-specific models for businesses and the value of Granite's customization capabilities. In a speed round, Asuri shares her vision for the future of AI, predicting that it will become invisible, simple, and seamlessly integrated into our lives.

Outlines

00:01:58

Open AI: Reshaping Industries and Redefining Possibilities

This episode explores the concept of "open" in the context of AI, focusing on how openness is driving innovation and reshaping industries. The conversation features Maryam Asuri, Director of Product Management and Head of Product for IBM's WatsonX.ai, who discusses the use of foundation models, particularly IBM's Granite, in enterprise AI applications.

00:04:13

The Rise of Generative AI and its Impact on Enterprise

Maryam Asuri shares her journey at IBM and her focus on AI application development. She discusses the transformative power of generative AI and how it's shifting the market from experimentation to production. She highlights the challenges of large models, such as increased latency and cost, and the emerging trend of smaller, customized models.

00:07:40

IBM's Granite: A Family of Foundation Models

Maryam explains the concept of Granite, IBM's family of foundation models, which are trained from scratch and offered with indemnification. She details the different flavors of Granite models, including language, code, time series, and geospatial models, and provides examples of their real-world applications in customer care and content generation.

00:12:57

The Future of AI: Agentic Workflows and the Power of Reasoning

The conversation explores the concept of foundation models and their potential risks due to lack of transparency in training data. Maryam emphasizes the importance of client protection and the need for trustworthy governance in AI models. She explains how Granite addresses these concerns through its training process and the Watson X platform. Maryam discusses the future of AI, predicting a shift towards generative AI that can reason, plan, act, and reflect. She highlights the development of function calling capabilities in Granite models, enabling them to interact with external tools and perform complex tasks.

Keywords

Q&A

• What are the key challenges enterprises face when moving from experimenting with generative AI to production?

  Enterprises face challenges such as increased latency, cost, and energy consumption when scaling up large generative AI models. This has led to a shift towards smaller, customized models that can deliver high performance at a fraction of the cost.

• How does IBM's Granite address the concerns of transparency and trust in foundation models?

  Granite is trained from scratch with a focus on client protection and trustworthy governance. IBM takes responsibility for the models and provides visibility into the training data and processes, addressing concerns about potential risks in highly regulated industries.

• What are the key considerations for buyers and users of AI tools and solutions?

  Buyers should consider the specific area where AI can benefit, assess data availability and quality, find a trusted partner, evaluate regulations, and apply design thinking principles to ensure AI is used effectively and ethically.

• What is the future of AI according to Maryam Asuri?

  Maryam believes the future of AI will be characterized by invisible, simple, and seamlessly integrated AI systems that are blended into our lives. She envisions a future where AI is so ubiquitous and user-friendly that it becomes an essential and largely invisible part of our daily routines.

Show Notes