News

Causal Reasoning Still a Challenge for AI, and the Key to Its Future

• By John K. Waters
• 05/31/2024

"Causality and causal reasoning are hallmarks of human intelligence," declared Dr. Darko Matovski during his keynote at the GenAI Summit in San Francisco this week. "From birth, children learn about cause and effect before even speaking their mother tongue. This fundamental aspect of human understanding remains a challenge for AI."

Matovski is co-founder and CEO of causaLens, maker of the decisionOS decision-support solution, which was designed to empower AI systems to reason about cause and effect (causal not "casual"), helping businesses analyze and visualize data to assess the impact of decisions on a unified interface.

Matovski criticized the current state of AI and statistics, noting that while statisticians emphasize the difference between correlation and causation, they often fail to provide methods for identifying causality from data. "Because of this gap, machine learning, including deep learning, is built on correlations," he said. "All the latest AI innovations are essentially correlations at scale."

The keynote explored how AI could be augmented to reason more like humans, particularly in enterprise decision making. "With causal reasoning, we don't just predict the future, we shape it," Matovski said, adding a quote from a speech by Abraham Lincoln: "The best way to predict the future is to create it." Matovski argued that causal AI can transform enterprises by enabling them to ask and answer questions that directly impact future outcomes.

He also described the evolution of enterprise AI, beginning with technologies that analyze past data, such as Salesforce's Tableau AI and Google Cloud's Looker, and moving to predictive AI that forecasts future events. But he underscored the limitations of current predictive AI, which assumes no agency in altering outcomes. "Causal AI unlocks the ability to ask, 'What should I do today to change tomorrow?'" he said.

Matovski provided some practical examples of the benefits of causal AI, such as improving customer retention and optimizing sales. Predictive AI can identify the probability of customer churn, but causal AI can suggest actionable strategies to prevent it. "Imagine knowing the specific actions to take to retain customers or increase sales," he said. "That’s the power of causal AI."

He also addressed the challenges of integrating large language models (LLMs) with causal reasoning. Although LLMs excel in pattern recognition and generating predictions, they lack the capacity for causal reasoning, he said. "LLMs are great for intuitive, fast thinking, but they struggle with the analytical, logical reasoning required for causal questions," he explained.

Highlighting the importance of causal grounding, Matovski introduced a new method, causal RAG (Retrieval-Augmented Generation), to enhance LLMs with causal reasoning. This approach allows LLMs to use causal models, knowledge graphs, and vector databases to provide grounded answers to complex questions, he said.

Matovski showcased real-world applications where leading organizations, such as Netflix and Spotify, utilize causal AI to personalize user experiences. He emphasized that his company aims to make this advanced technology accessible to all enterprises, enabling them to leverage causal AI without extensive resources.

Matovski concluded his talk with his company's vision of a future where enterprises can describe and optimize their entire business operations through causal models. "We want to enable businesses to ask fundamental questions like 'Why are revenues down?' and 'How can I fix it?' and receive actionable insights," he said. "This is the future of enterprise AI."

This year's GenAI Summit, underway this week in San Francisco's Palace of Fine Arts, is the second annual event organized by GPT Dao, a global generative AI community. According to event organizers, this year's summit drew an estimated 10,000 attendees and 300 exhibitors. The list of exhibitors at this year's conference includes Microsoft, IBM, and Amazon. (A complete list is available on the conference website.)

("Dao" stands for Decentralized Autonomous Organization. Dao's operate based on smart contracts executed on a decentralized network, typically a blockchain, which allows for decentralized decision making and control over the organization’s assets and operations.)

GPT Dao provides a range of services to its community, including Web3 and AI project incubation, GPT investment research education, and AI infrastructure services. It also provides a platform for community governance designed to allow members to propose, discuss, and vote on changes, new features, and/or initiatives. It also provides a decentralized crowdfunding platform that allows anyone to invest in or contribute to a project, regardless of their location or financial status.