AI Confidential – Détails, épisodes et analyse

OpenAI's ChatGPT-4, released in March 2023, was noticeably more intelligent than ChatGPT-3. It reasoned better, answered questions more accurately, and could handle more complex tasks. But LLM advancements have since slowed down. And critics have called OpenAI's latest release — ChatGPT-5  — "overhyped and underwhelming."

In this episode, hosts Aaron Fulkerson and Mark Hinkle discuss:

• Whether LLMs are dead
• Why 95% of AI projects are failing
• How enterprises can harness AI without sacrificing data security

If you want to learn more about AI and how to use it responsibly, visit opaque.co

In the age of agentic AI, Kellie Romack knows deploying secure agents needs to be a baseline for enterprises, not an aspiration.

In this episode, host Aaron Fulkerson unpacks how confidential AI agents work and talks to Kellie Romack, CDIO at ServiceNow, about how her company is implementing confidential AI agents across its platform. Since partnering with OPAQUE, Microsoft Azure, and NVIDIA to deploy confidential AI agents, ServiceNow has seen average response times from its sales commission help desk speed up from four days to just eight seconds.

Kellie and Aaron also discuss:

• How ServiceNow, a defining enterprise software company, is approaching AI governance
• Advice for enterprise leaders building out agentic workflows
• Predictions on how companies can gain the most value from AI

If you want to learn more about AI and how to benefit from it responsibly, visit opaque.co

Special thanks to Now Productions for filming this interview at ServiceNow's headquarters in Santa Clara.

Hosts Aaron Fulkerson and Mark Hinkle sit down with Sriram Raghavan, Vice President of IBM Research AI, at the AllThingsOpen.ai conference. As an experienced industry expert, Sriram leads a global team of over 750 research scientists and engineers focused on advancing AI. 

Sriram and IBM believe that the future of AI is open source. That's why IBM recently contributed three AI projects — Docling, Data Prep Kit, and BeeAI — to the Linux Foundation. And while there is a lot of hype around LLMs right now, IBM is leaning into smaller "fit-for-purpose" models that give developers choice.

In the episode, Sriram also digs into:

• IBM's approach to open-source models
• What innovations he sees coming for enterprise AI
• How his team is approaching responsible, secure AI

From cloud computing to infrastructure-as-a-service, Reuven Cohen (AKA rUv) has been on the cutting edge of every major technology supercycle — and he's now one of the most influential people in the agentic AI space.

With the help of agents, rUv produced 10 million lines of usable code last year. That's 33,000 years of output from one human. rUv's approach proves one of two things: he's either an outlier or a leading indicator of what's to come.

In this episode, we talk to rUv about: 

• His approach to training autonomous agents
• The rapid advancements in agentic code development 
• Why LLMs are vulnerable to data poisoning
• Why Python is bad for agentic development

What does it take to ensure AI is safe, ethical, and resilient? In this season finale, Anthropic's CISO joins Aaron to discuss the critical intersection of innovation, data privacy, and sovereignty, offering an optimistic perspective on what lies ahead for the future of confidential AI. 

AI is full of potential, but what does it really take to make it work in complex environments? Will Grannis, CTO and VP at Google Cloud, shares his perspective on bridging AI innovation with practical application, including insights into the real challenges and rewards of implementing AI in today's top enterprises.

Can AI be both innovative and private? NVIDIA VP of Software Security Daniel Rohrer and Opaque Co-Founder and President Raluca Ada Popa discuss the advancements in privacy-preserving technology that enable organizations to push AI forward without compromising data sovereignty.

Generative AI is transforming industries—but how do we ensure it's safe? Teresa Tung, Senior Managing Director & Global Data Capability Lead at Accenture, dives into what companies need to make generative AI both impactful and secure, exploring the frameworks that bring safety to scale.

How do you build AI applications that are not only powerful but also trustworthy and easy to adopt? Join Mark Papermaster, CTO & EVP at AMD, and Mark Russinovich, CTO & Deputy CISO at Microsoft Azure, as they discuss the essential role of confidential computing in making AI seamless, secure, and accessible to businesses everywhere.

Welcome to AI Confidential, the podcast that explores the future of AI and data sovereignty.

In today's rapidly evolving tech landscape, AI is at the heart of transformation across industries. From finance to high tech, leaders are racing to harness its power while navigating complex challenges like data sovereignty, regulatory compliance, and privacy. But what if there was a way to solve these problems without sacrificing innovation?

On AI Confidential, join host Aaron Fulkerson as he dives deep with visionaries from AMD, Accenture, Google Cloud, Microsoft Azure, NVIDIA, Opaque Systems, and more to uncover how confidential computing is reshaping what's possible for AI. Together, they explore pressing questions, share groundbreaking use cases, and take bold steps in predicting AI's future.