Thank you for attending Ai day!

The discussion focused on using the RIO (Reduce, Increase, Optimize) framework to identify and prioritize AI use cases in an organization. 

Several case studies were presented to illustrate how companies are using AI to reduce costs, increase productivity and efficiency, and optimize operations:

• Reducing costs: A data center optimized cooling systems using 3D simulations and machine learning models, reducing energy load.
• Increasing productivity: A retailer used computer vision AI to reduce theft at self-checkout, improving success rates.
• Optimizing operations: A utility company unified data sources and added automation to streamline operations and reduce high-priority incidents.

The speakers also covered some of WWT's internal use cases, including an AI chatbot, RFP assistant, and coding assistant to boost employee productivity and efficiency.

The key takeaway was the importance of focusing on the "why" behind AI use cases to drive meaningful business outcomes, rather than just implementing technology for technology's sake.

The discussion focused on the importance of having a well-defined data strategy and framework, rather than relying on hope. 

Key topics:

• Recognition that generative AI is rapidly changing the data landscape, emphasizing the need to start AI projects now, even if the organization is not at the highest level of data maturity.
• Challenges encountered in creating a synthetic data set, highlighting the complexities of working with unstructured and semi-structured data, and the critical role of thorough data cleaning.
• Demonstration of using tools like NVIDIA's NeMo Curator and Google Cloud's Vertex AI to quickly ingest, curate, and integrate data into a vector database, enabling rapid insights and interaction.
• Focused on the importance of data quality, governance, and leveraging current investments in data and technology, rather than acquiring new tools. The potential of generative AI and the need for continuous learning and adaptation in data strategy were also emphasized.

The discussion focused on the practical applications and integration of AI in the enterprise. It covered the evolution of AI capabilities, from recognition to advanced reasoning, and the importance of generative AI models.

Key topics:

• The conversation highlighted the omnipresence of AI in various software products and the need for enterprises to adopt a strategic approach to integrating AI into their workflows and processes.
• A significant portion of the discussion revolved around the concept of "agent AI" - the ability for AI systems to act independently, make decisions and collaborate with each other to achieve complex tasks.
• The presentation also provided several practical examples and demonstrations of using AI tools, such as ChatGPT, Copilot and Notebook.ai, for tasks like personalized education, sentiment analysis, role-playing and self-discovery.
• The importance of prompt engineering was emphasized as a key skill for effectively leveraging AI models and extracting maximum value from them.
• The discussion also touched on the integration of AI with other automation tools and the need to consider security and privacy implications when using AI in the enterprise.

Throughout the conversation, the focus was on encouraging attendees to actively experiment with AI tools, understand their capabilities, and explore practical applications that can enhance productivity and decision-making within their organizations.

The discussion focused on the development and testing of an AI and cybersecurity "enclave" or sandbox environment. The AI Security Enclave was created to help educate people on AI security and allow for efficient testing and validation of AI security solutions.

Learn more about The AI Security Enclave here. 

Key topics:

• The AI Security Enclave includes high-performance computing resources like servers, GPUs, and orchestration tools to enable flexible and scalable testing. It has both virtual and physical infrastructure set up to allow for secure testing of AI applications.
• Key activities in the enclave include testing security controls, evaluating vendor solutions, exploring responsible AI practices, and developing testing frameworks for things like prompt injection, data security and identity/access management.
• There was discussion around the challenges of segregating data sources and access controls when deploying internal vs external-facing AI systems, and the need for integrated security tooling to address the evolving AI security landscape.
• The overall goal of the AI Security Enclave is to help organizations develop the expertise and test the security of their AI systems in a low-risk environment before deploying them in production.

The discussion focused on the growing threat of deepfakes and misinformation, and the need for effective strategies to combat these challenges.

Key topics:

• It covered the impact of AI advancements on society and the various reactions from different groups, including the challenges faced by organizations without deepfake defense plans.
• The conversation highlighted the persistent issues with cybersecurity awareness and training, and the need to shift the focus towards augmenting human capabilities rather than relying solely on technology.
• Evaluating deepfake vendors and solutions was a key topic, emphasizing the importance of comprehensive coverage, real-world testing and vendor accountability.
• The approach discussed involved breaking down the problem into manageable parts, focusing on signals and human interaction to enhance deepfake detection.
• Building a comprehensive solution was also discussed, including components like border protection, session protocols and action-based responses, with an emphasis on continuous learning and adaptation.
• The importance of identity, trust and privacy was highlighted, exploring the use of advanced biometrics and the need for a balanced approach between security and user experience.

The conversation concluded with a call for industry collaboration, continuous innovation and the importance of staying vigilant in the face of evolving deepfake threats.

The presentation focuses on the importance of securing AI systems and their supply chains. It highlights the current industry trends, the vulnerabilities of AI models, and the need for robust AI governance and security measures.

Key topics:

• Emphasizes the priority of red teaming AI systems and ensuring data accessibility.
• Notes the acceleration of startup activity and the slow pace of established vendors in shipping code.
• Discusses the importance of securing AI systems through app and API security, and strong software supply chain processes.
• Highlights the risks of adversarial use of AI, including phishing, CAPTCHA breaking and deepfakes.
• Covers the need for comprehensive AI security governance, including data security, model security and compliance measures.
• Explains the significance of securing the software supply chain for AI/ML models.
• Identifies various vulnerabilities in AI models, such as data poisoning and model tampering.
• Introduces the AI Security Enclave in the AI Proving Ground, which supports AI security efforts and demonstrates expertise in testing innovative security solutions.

The discussion highlighted the importance of a comprehensive strategy and the right infrastructure in enabling the scalability and performance required for enterprise-level AI deployments.

Key topics:

• Challenges in choosing the right vector database solution to support large data volumes for chatbot applications.
• The need for a hybrid infrastructure approach, combining on-premises and cloud resources. - Comparing the performance and tradeoffs of Ethernet vs. InfiniBand networking for AI workloads.
• Using a sizing tool to recommend the appropriate infrastructure based on application requirements. - Addressing data silos and the importance of a unified data integration platform.
• The role of high-performance networking, RDMA, and network tuning for AI workloads.
• Considerations around storage metrics, checkpointing, and modern storage architectures.
• The significance of orchestration and workflow management in scaling AI deployments.
• Taking a holistic approach to AI architecture, integrating platform tools, algorithms, and services.

The conversation discussed the significant impact of AI on data center facilities and the need for close collaboration between IT and facilities teams. Key areas impacted include power, cooling, space and cabling, requiring upgrades and new strategies to handle the increased demands. Monitoring and management solutions also need to evolve to support the complexity of AI infrastructure. Overall, the discussion highlighted the importance of proactive planning and design to overcome the challenges posed by the accelerated growth of AI deployments.

Recommended reading: WWT Research | Facilities Infrastructure Priorities in the Age of AI

Key topics:

• AI is expected to have a significant impact on data center infrastructure, affecting power, cooling, space and cabling requirements.
• Facilities and IT teams need to collaborate closely to design and deploy AI solutions, as the infrastructure changes will be substantial.
• Power capacity and cooling strategies need to be re-evaluated to handle the increased loads from AI workloads.
• Cabinets and cable pathways will need to be larger and more robust to accommodate the higher power, cooling and cabling density.
• Monitoring and management solutions (DCIM) will need to be enhanced to handle the increased complexity of AI infrastructure.
• Challenges include long lead times for power and cooling equipment, skill shortages and the need for flexible and scalable designs.

The presentation discusses the current landscape of AI, focusing on the decision-making process between building AI solutions in-house versus buying them from external providers. It covers various aspects such as costs, operational challenges and strategic considerations for both approaches.

Key topics:

• The presentation starts with an overview of the AI landscape, highlighting the increasing importance and investment in AI technologies.
• It emphasizes the high costs associated with AI projects, particularly generative AI, and provides best practices for optimizing these costs.
• The presentation outlines the operational complexities and challenges in managing AI infrastructure, including data center power demands and the need for specialized expertise.
• It provides a detailed comparison of building AI solutions in-house versus buying them, discussing factors such as control over design, data security, cost savings and integration with existing systems.
• Various paths to AI service delivery are explored, including public cloud platform services, on-premises solutions and GPU as a service.