Agenda | Causal AI Conference 2024 (2024)

Speakers:

• Darko Matovski, Founding CEO, causaLens
• Tom Farrand, Director of Product, causaLens

Synopsis:

Causal AI represents a new framework for enterprise decision making- one which can represent the complex cause and effect interdependencies of an organisation. In this talk, Darko and Tom will paint a vision for the future of enterprise decision making, discussing both the real world successes which causaLens has seen to date in making this reality- as well as the challenges being tackled.

Speaker:

Sharath Gokula, Director of Data Science & Analytics, Sam's Club

Synopsis:

In retail, measuring technology investments in non-eCommerce contexts is incredibly hard. Many external influences, such as operational process changes, business decisions, customer preferences, and assortment changes, make isolating the true impact of technology investments difficult. However, our main focus should be to enable good and quick decision-making to prioritize the right projects and facilitate learning. With measurement, we are often forced to not just optimize for accuracy but also take a balanced approach and consider reusability, time to insight, and explainability. We need to look at causal approaches as a solution to a problem, not the problem itself.

Speaker:

Nalanda Mattia, Senior Director, Econometrics, Dun & Bradstreet

Synopsis:

Causal AI models can help lawmakers identify the most effective policies for achieving economic goals. Traditional causal methods like difference-in-difference have limited usefulness in evaluating policy effectiveness. This work with causaLens employs decisionOS and its causal AI methods to illustrate its effectiveness in the context of healthcare policy and its impact on pharmacies.

Cost control policies for prescription drugs are crucial for the resilience of firms within the medical system. However, the impact of these policies on pharmacies remains poorly understood. This study uses a synthetic control method to estimate the direct effect of cost control policies on the probability of pharmacies going out of business. By applying this method to a rich firm-level dataset aggregated at the state level, the research aims to find the optimal balance between cost control and pharmacy viability through understanding the causal dynamics of the regulation on the target pharmacies.

In this engaging fireside chat, Turing Prize Winner Judea Pearl will share his insights on the critical role of causal reasoning in shaping the future of artificial intelligence. Pearl will discuss the limitations of current AI approaches, including deep learning and Bayesian networks, arguing that a focus on causal inference is essential for creating truly intelligent systems. He will also share his views on the potential and limitations of large language models (LLMs) and their role in achieving artificial general intelligence (AGI).

Speaker:

Scott Evans, Principal Scientist, AI-Machine Learning, GE Vernova

Synopsis:

Causal inference is imperative if we are to leverage AI to solve the climate crisis. But if we are not intentional, training and use of increasingly energy-intensive AI models could make matters worse – by some predictions, up to 20% of world energy (and related carbon) could be directed to computational-related tasks within a decade. This talk will explore how causality is essential to Green AI applications using wind energy use case examples. We will then explore the optimization of what?, where?, when?, and how. Causal and Green AI models are produced and utilized to work together to provide breakthrough benefits for society.

At its root, Green AI is not just the domain of using Causal AI to combat the climate crisis, nor only the concept of conducting AI in energy-efficient and carbon-neutral ways, but rather the recognition that energy and learning are intrinsically linked and can be jointly optimized to maximize learning, minimize carbon, and even stabilize the grid.

Speakers:

• Naveed Sharif, Director of Data Science & Web Analytics, Kaiser Permanente
• Jeff Groesbeck, Staff Data Scientist, Kaiser Permanente

Synopsis:

In today’s fragmented healthcare landscape, the seamless integration of care delivery and insurance coverage stands out as a transformative model that promises to enhance patient care and redefine health management. At Kaiser Permanente, our unique, fully integrated healthcare model—encompassing care delivery and insurance—provides a fertile ground for causal inference, allowing us to connect dots that remain disparate in more traditional setups.

This presentation will delve into how our comprehensive and cohesive data environment enables us to apply causal inference techniques effectively to improve patient care and operational efficiency. I will showcase specific case studies from Kaiser Permanente, such as the impacts of digital engagement on preventive health behaviors and the optimization of treatment protocols across diverse patient demographics. These examples highlight our capability not only to analyze but also anticipate the needs of our patients, shaping a healthcare delivery system that is as proactive as it is reactive.

Speaker:

Maher Lahmer, Head of Data Science, Google

Synopsis:

Sellers and marketers interact with advertisers to offer services that vary from early onboarding, to sales and support. Understanding the true impact (causal effect) of these interactions on advertiser outcomes is critical to achieving their marketing objectives. This presentation dives into Google Ads' business environment and the strategic decisions driven by scientific measurement.

We'll explore our journey towards building causal AI solutions that empower science-driven decisions for impactful business operations at scale. We'll emphasize the importance of cross-disciplinary pollination in advancing causal inference and argue how Causal AI is uniquely positioned to improve AI transparency and solving real-world problems effectively.

Speaker:

Adam Plumpton, VP of Global Planning, Cisco

Synopsis:

In this fireside chat, Adam Plumpton, VP of Demand Planning at Cisco, will discuss how the company is revolutionizing its demand planning processes by embracing Causal AI. By identifying the underlying causal relationships that drive customer demand, Cisco is moving beyond mere predictions and taking proactive measures to influence outcomes.

Adam will share insights on how Causal AI is enabling Cisco to build a resilient supply chain, the challenges faced during implementation, and the surprising benefits discovered along the way. He will also discuss the practical application of Causal AI within Cisco's Supply Chain Operations Team and how it is transforming decision-making and organizational behavior.

We will take an hour break for lunch and networking in person.

Speakers:

• Abhi Mukerji, Senior Economist, Amazon
• Qing Wu, Director of Econometrics, Google
• Tilman Drerup, Director, Machine Learning, Instacart
• Wenjing Zheng, Tech Lead, Ecosystem Data Science

Synopsis:

Businesses increasingly leverage causal methods to uncover valuable insights, make data-driven decisions, and drive meaningful impact. However, the path to successful causal AI implementation is not without its challenges. In this panel discussion, we bring together experts from leading tech companies Google, Amazon, Instacart, and Roblox to share their successes, challenges, and hard-earned lessons in navigating the complexities of applying causal AI in real-world settings.

Speaker:

Eray Turkel, Senior Data Scientist, Google

Synopsis:

We constantly need to make decisions in business, choosing between interventions or making resource allocation decisions:

• Who should we try to convince to adopt our new product?
• How much should we spend on marketing for our new hardware launch?
• How should we decide which clients our salespeople should prioritize talking to?

Using machine learning to inform these decisions presents a great opportunity.

However, ML models are only helpful when used in conjunction with rigorous data collection and evaluation, and when we keep in mind their limitations.

Adopting a continuous measurement initiative based on data collection, experimentation, and the appropriate causal inference techniques, we can ensure high quality decisions, and minimize risk.

Speaker:

Max Sipos, CTO & Co-Founder, causaLens

Synopsis:

In this talk, we will discuss the interesting developments in the intersections of causal AI and large language models (LLMs). We will discuss the broader academic research and our own work on specific parts of this subfield which are of most interest to our enterprise customers.

First such interest is in using LLMs as domain knowledge copilots which extract causal information from textual corpuses. This domain knowledge can be seamlessly combined with algorithmic causal discovery in an interesting synergy. Secondly, we will discuss development of AI agents to make the work of causal AI model developers easier and more productive. Thirdly we will discuss how to use causal models to ground LLMs against hallucinations and how to bridge the world of causal models upon structured data with that of unstructured, free form, human text.

Speaker:

Guido Imbens, Nobel Prize Laureate and Economics Professor, Stanford University

Synopsis:

In many settings where researchers are interested in estimating causal effects there may be interference between units, where treating one unit affects outcomes for other units. This creates challenges for both the design of experiments and for the analyses of observational studies. In this talk I will discuss some recent developments in the design of experiments in online settings where concerns about spillovers or interference are common. I will also discuss some settings with observational studies where new methods have been proposed to deal with such complications.

Speaker:

Sanchin Raj, Principal, Data Science & Advanced Analytics, AT&T

Synopsis:

Darko Matovski, Founder & CEO, causaLens

We will be closing the event out and opening up the in-person event for networking drinks.