The Causal AI Conference 2022

The Causal AI Conference 2022:
The Causal AI Community Today and Beyond
Max Sipos
CTO, Co-Founder, causaLens

Conventional AI approaches, such as generative language models, generate data based on correlations. Causal AI instead, attempts to discover the accurate data-generating process which underpins the mechanism of cause and effect in the world. This is necessary for the accurate application of AI via scenario modelling and interventional and counterfactual reasoning in complex systems, such as modern economies. Deploying Causal AI in the enterprise is fraught with difficulties that the community has started to address today. In this talk, we will first discuss the necessary interfaces that permit Causal AI systems to cooperate in the enterprise. We will also discuss how to put Causal AI into the hands of the business user in the most efficient way possible, through developing suitable user interfaces and advances in explainable and robust AI. Finally, we will discuss the future challenges ahead of us, and where Causal AI will fit in the rapidly evolving AI landscape.

The Causal AI Conference 22
Panel Discussion - Why there’s a lack of trust in AI and how can we fix it?

Speakers:
Puneet Gupta, Andre Franca, Nicholas Chia & Stephen Pritchard

Learn more and join the community here:
https://www.causalaiconference.com/

00:00 - Welcome
01:37 - Panel Introductions
08:17 - Why there’s a lack of trust in AI and how can we fix it?
38:02 - Q&A

The Causal AI Conference 2022:
An Emerging Solution to Harmonize Various Causal Discovery Methods
Nima Safaei
Senior Data Scientist, Scotia Bank

Explainability is one of the most desired properties of AI systems; without which the AI systems cannot be trusted in high-risk fields. Causal Inference (CI) is a vital tool for producing more insightful explainability. However, one major shortfall in the current CI literature is the lack of a unique definition for causality; resulting in many different methods such as pairwise dependency tests, statistical conditional tests, structural models, and graph-based models. One major barrier to the use of CI for explainability in AI applications is that the various CI methods usually result in different causal graphs with different inter-connectedness and density; specifically given a high-dimension feature space.

In this talk, Nima will address this challenge from the perspective of the financial services industry, walk through the associated complexities, and outline the possible methods to find a solution that harmonizes various CI methods.

Learn more and join the community here:
https://www.causalaiconference.com/

00:00 - Welcome
00:16 - An Emerging Solution to Harmonize Various Causal Discovery Methods
25:23 - Q&A

The Causal AI Conference 2022:
Modelling and Decision-Making in an Ever-Changing World
Annie Hou
SVP - Global Head of AI & Behavioural Sciences, MRM

Over 2500 years ago Heraclitus said that “the only constant in life is change”. While this is now a clichéd quote painted on walls (and used in conference abstracts), these words could not reflect reality more so than in our current times. The frequency of black swan events has risen dramatically over the past few decades. Yet, one of the core assumptions of many ML models is that future events can be captured in historical data. Coupled with the explosion of big data and data-driven processes, many decision-makers have become accustomed to ‘predictive models’ that provide singular solutions in spite of a constantly changing world.

In this talk, Annie will discuss the need for models that allow for testing interventions and counterfactuals (the 2nd and 3rd rungs on Judea Pearl’s Ladder of Causality) in decision-making. She will share her experience building these decision-support tools:

- Building tools for data science and decision-makers with innovative technologies
- Getting clients to use these tools and how they respond
- Understanding the potential value of adopting Causal AI for marketing

Learn more and join the community here:
https://www.causalaiconference.com/

00:00 - Welcome
00:48 - Modelling and Decision-Making in an Ever-Changing World
27:59 - Q&A

Causal AI Conference 2022:
Understanding Sales Drivers - Causal AI at Nestlé
Jordi Mur
Analyst, Emerging Technologies Nestlé Innovation, Nestlé

The food sector is highly complex and competitive. In such a data-rich environment, Nestle IT Innovation supports the development and adoption of Next AI B2B tools across the business, from edge AI to causal inference, to quantum computing, to name a few.

In this talk, Jordi will illustrate use cases and learnings from our experience bringing causal AI to Nestlé.

Learn more and join the community here:
https://www.causalaiconference.com/

00:00 - Welcome
00:46 - Understanding Sales Drivers: Causal AI at Nestlé
22:27 - Q&A

The Causal AI Conference 2022:
Causal AI in the Energy Industry - Lessons learned at TotalEnergies
Antoine Bertoncello
Head of Next Generation AI, TotalEnergies

One increasingly popular approach is to use machine learning algorithms to predict the future behaviour of a system (e.g., failure of equipment, the energy produced by renewable assets…) based on correlations found in the data. However, a prediction is sometimes insufficient, and one may want to know what will happen if a specific variable is changed to minimize the failure rate or optimize production. This step is particularly difficult because it requires going beyond correlations and instead inferring a causal relationship between variable A and outcome B. The concept of causality is, in general, answered through experiments, such as randomized control trials, experimental design, and simulation. However, in industrial settings, experiments are not feasible, and engineers and data scientists often only have access to observational data on the impacts on productivity.

Epidemiologists, economists, and computer scientists have recently developed a range of statistical tools to go beyond correlation to discover and infer causal relationships. However, their use remains scarce in the energy world. In the last three years, TotalEnergies launched a research project to assess the use of causal inference in the company. In this talk, Antoine will describe a selection of the business cases that were identified, the approaches used, as well as the new ones we are developing with their academic partners.

Learn more and join the community here:
https://www.causalaiconference.com/

00:00 - Welcome
00:50 - Causal AI in the Energy Industry: Lessons learned at TotalEnergies
23:50 - Q&A

The Causal AI Conference 2022:
Establishing Partnerships for the Learning of Causal Relationships in Medicine
Nicholas Chia
Associate Professor, Mayo Clinic

The talk will introduce the problems with ‘vanilla AI’ (non-causal AI), providing some thought-provoking examples. Nick will then discuss the implications for ethics and fairness and present his cutting-edge work on “The Unreasonable Effectiveness of Inverse Reinforcement Learning in Advancing Cancer Research”.

“Can we use causal inference methods to understand the molecular basis of cancer?”

Nick will deliberate on this important question and attempt to answer it through interesting examples.

Several of the applications of Causality to medicine and healthcare will be presented, including exciting work carried out with NASA.

Learn more and join the community here:
https://www.causalaiconference.com/

00:00 - Welcome
01:05 - Establishing Partnerships for the Learning of Causal Relationships in Medicine
23:40 - Q&A

The Causal AI Conference 2022:
Actual Causality: A Survey
Joe Halpern
Professor - Computer Science, Cornell University

What does it mean that event C “actually caused” event E?

The problem of defining actual causation goes beyond mere philosophical speculation. For example, in many legal arguments, it is precisely what needs to be established in order to determine responsibility. What exactly was the actual cause of the car accident or the medical problem?

The philosophy literature has been struggling with the problem of defining causality since the days of Hume, in the 1700s. Many of the definitions have been couched in terms of counterfactuals. C is a cause of E if, had C not happened, then E would not have happened.

In 2001, Judea Pearl and I introduced a new definition of the actual cause, using Pearl’s notion of structural equations to model counterfactuals. The definition has been revised twice since then, extended to deal with notions like “responsibility” and “blame”, and applied in databases and program verification.

In this talk, Joe will survey the last 15 years of work, including collaborations with Judea Pearl, Hana Chockler, and Chris Hitchcock.

Learn more and join the community here:
https://www.causalaiconference.com/

00:00 - Welcome
00:15 - Actual Causality: A Survey
44:35 - Q&A

The Causal AI Conference
Welcome to the Causal AI Community: Scaling the Global Adoption of Trustworthy AI
Darko Matovski
CEO, Co-Founder, causaLens

This talk celebrates the establishment and rapid growth of the Causal AI community and introduces The Causal AI Conference as part of a more excellent vision to nurture and support the community.

Darko will discuss the emergence of Causal AI, and the conditions that have come together to create the right environment for Causal AI to thrive. How these have led to the Causal AI revolution, with exploding interest from academia, start-ups, investors and big tech alike.

He will share his experience with adopting Causal AI throughout different sectors, introduce the current state of the art of Causal AI in practical applications, and present his vision for the future role of Causal AI in our community and society.

Learn more and join the community here:
https://www.causalaiconference.com/