AI Summit West schedule

Using Machine learning to run a Billion-dollar business is a complex endeavor. This talk will present a real-world use case of using Predictive Machine Learning to predict supply, demand, and inventory levels in a 3-sided marketplace and its impact on end users' experience.

The users will walk away with an appreciation of why forecasting is a critical and difficult problem to solve in delivery marketplaces and the algorithmic sophistication needed to solve it.

In an age where AI is redefining our world, the mastery of data storage, processing, and security at an immense scale is not just a technical feat – it's a necessity.

Today's sophisticated AI models are powered by extensive datasets, often escalating to petabyte levels and beyond. The challenge is multifaceted: How do we scale these storage systems efficiently? How do we maintain cost-effectiveness? And critically, how do we ensure robust data security in this vast digital expanse?

This talk will offer a comprehensive exploration of state-of-the-art strategies and vital considerations for constructing and protecting these massive data repositories.

Attendees will gain valuable insights into balancing large-scale data management with operational cost-efficiency, developer agility, and stringent security measures.

Join me to navigate the intricate dynamics of 'Data@Scale' – where size, speed, and security converge.

• What patterns emerge in the types of changes made during iterative revisions? 

• Examining the nature of revisions provides insights into whether certain patterns or trends exist, such as consistent adjustments in structure, language, or content. Understanding these patterns helps uncover the priorities and focus areas for writers during the revision process. 

• How does feedback influence iterative revision patterns, and what role does collaboration play in the writing process? 

• How can generative AI strategically enhance creativity, innovation, and productivity across the enterprise? 
• What safeguards ensure the ethical and responsible use of generative AI across different business units? 
• In which domains does generative AI show the most promise, and how can businesses tailor its implementation for maximum value? 

• How does self-supervised deep learning improve feature representation learning in Automated Speech Recognition (ASR)?
• What self-supervised tasks are effective in pretraining deep neural networks for ASR, and how do they impact performance?
• How does self-supervised learning address challenges in low-resource ASR settings?
  
  

We will focus on three pivotal market trends in skills technology: the essential role of AI literacy in business success, bridging the technical skills gap, and proving the ROI of upskilling in AI, Data, Cloud and Security. At the heart of our discussion is how we can accelerate project delivery through improving how we assess, configure, and improve how we address team skills.

• How do models contribute to the strategic goals and competitiveness of a business? 
• What are the key considerations in managing and maintaining models as valuable business assets over time? 
• How does the interpretability and explainability of models influence their adoption and acceptance within the business and by stakeholders? 

This talk delves into the critical task of ensuring that AI systems resonate with and prioritize human values. The speaker explores various methodologies and frameworks aimed at achieving this alignment, emphasizing the significance of ethical considerations and societal impact. The talk underscores the need for collaborative efforts among researchers, developers, and policymakers to establish guidelines and standards that prioritize the well-being and interests of humanity.

• Understand how to build trust and decrease fear around integrating AI across your enterprise
• Best practices in reducing risk around AI in the enterprise

This talk delves into the infrastructure behind training large models at Pinterest. It covers three essential components: training hardware, compute orchestration, and ML application development SDKs. Throughout the talk, we will highlight Pinterest's use cases, showcasing the practical applications and impact of the above three pillars on dev velocity and model training efficiency. We will also talk about how Pinterest uses AWS UltraCluster to enable efficient large scale model training.

This session will review the vulnerabilities of NLP models, unveiling two privacy attacks aimed at extracting instances of the training data from a trained model. Subsequently, the session will explore measures to safeguard language models against these threats, ensuring a responsible and secure integration of large language models.

Causal frameworks are popular now as businesses insist on data-driven decisions. They can provide clear cause-and-effect relationships between interventions and observed outcomes.

 One of the challenges in designing these frameworks is scaling across different scenarios. In this talk, we present the design of an AI/ML based counterfactual estimation framework.

At the core of this framework is a causal model that learns relationships between users' actions over time. We will discuss this using an example of a time series data. 

Next in the talk, we will discuss how a counterfactual can be estimated for any policy change. If the immediate effects of a policy change can be formulated, we can use our counterfactual estimation framework to predict the long term causal impact of the same. 

We will also go over some big-data challenges we encountered and the solutions we applied in this framework.

AI is becoming a dominate force in product development, with AI embedded in all aspects of product design.  Consumers are expecting more from end products and services, with AI generated intelligence and insights driving new experiences.  However, this paradigm presents new challenges, especially pertaining to the trust and context of AI generated information.  This presentation outlines three keys to incorporating AI in product development, ultimately helping to deliver new AI driven capabilities with speed and confidence. 

• How is unified ML data pipeline designed for real-time features, and what ensures efficiency throughout training to serving? 
• What challenges arise in building and maintaining a real-time ML pipeline, and how are they addressed for sustained performance? 
• How does the unified pipeline facilitate collaboration among stakeholders for effective development and deployment of real-time features? 

• How is accountability ensured in AI systems, emphasizing transparency and traceability in decision-making? 
• What methods address biases in AI algorithms and integrate fairness and ethics into system development and deployment? 
• How do organizations measure the impact of AI decisions on diverse groups and ensure equitable outcomes in AI applications? 

• How can organizations integrate ethical considerations across the AI development lifecycle, from data collection to deployment? 
• How do businesses effectively communicate their commitment to ethical AI practices to stakeholders, fostering trust and transparency? 
• What role do interdisciplinary teams play in building an ethical AI practice, and how can organizations foster a culture that values diverse perspectives? 

• How can generative AI strategically enhance creativity, innovation, and productivity across the enterprise? 
• What safeguards ensure the ethical and responsible use of generative AI across different business units? 
• In which domains does generative AI show the most promise, and how can businesses tailor its implementation for maximum value? 

The World Economic Forum says AI will eliminate 85 million jobs in the next three years… and create 97 million new ones. Now is the time for leaders to understand how to embrace generative AI to deliver better service experiences. In the decade ahead, those with an understanding of what’s possible with AI will vastly outperform the competition.

• How to train virtual agents on your content to have natural language conversations with employees
• How to use large language models (LLMs) to coach live agents to resolve more issues the right way the first time
• How to mitigate the risk of AI hallucinations and biased answers
• How to ensure generative AI apps are auditable and their answers are explainable

Join us for this thought-provoking talk that will challenge the way you think about AI and help you prepare for the future of work.