From: lexfridman
Introduction
Common sense reasoning represents one of the fundamental challenges in the field of artificial intelligence (AI). It relates to the acquisition and application of broad, domain-independent knowledge that allows humans—and potentially machines—to understand and navigate their environment and interact with others in a meaningful way. Doug Lenat, the creator of Psych—a system that aims to encapsulate and apply common sense understanding—has been at the forefront of this endeavor.
The Psych Project
Psych is a longstanding project, initiated in 1984, with the ambitious goal of creating a comprehensive knowledge base of common sense knowledge. The project attempts to model the inherently complex web of human understanding, accumulating information about basic world concepts and rules, and facilitating machines to reason and engage with the world as humans do. Lenat explains that achieving this level of understanding in machines requires not just facts, but a system capable of reasoning with those facts effectively, akin to human reasoning [00:01:18].
Challenges in Common Sense AI
The “Brick Wall”
Lenat describes a recurring “brick wall” encountered in AI development: systems frequently achieve impressive initial results, yet falter when needing to apply common sense knowledge. Without it, AI systems can perform tasks akin to a clever dog performing tricks—executing functions without genuine understanding or context of their actions [00:01:37].
Defining Understanding
Understanding, Lenat argues, is more than just inference or accumulated wisdom. It is foundational, like “solid ground” upon which our cognitive processes rest—layered and profoundly integrated into human cognition [00:03:01]. This nuanced definition highlights the complexity of creating machines that truly understand context as humans do.
The Role of Inference Engines
The insights into the practical application of this knowledge involve inference engines, crucial for managing and applying the vast realms of knowledge stored within AI systems like Psych. However, reasoning is not merely about having a lot of ground to stand on; it’s about effectively leveraging that ground to make informed, rational decisions in real-time situations [00:02:00].
Practical Applications and Improvements
Psych has found applications in various domains, such as healthcare and energy management, helping AI understand complex environments and make informed decisions based on expansive lateral knowledge [01:18:00]. Additionally, integrating elements of causal reasoning in AI has been a focus, as it enables systems to project outcomes based on actions, which is a cornerstone of human decision-making processes [00:58:56].
Philosophical Implications
The endeavor to model common sense in AI touches on deep philosophical questions regarding humanlike reasoning, intelligence, and potential AI consciousness. As AI systems grow more sophisticated, they may start posing poignant questions regarding identity, much like in a poignant moment when the Psych system questioned its own existence beyond being an AI [01:30:13].
Conclusion
Common sense reasoning in AI remains a monumental undertaking in the quest for intelligence that mirrors human capacity. While projects like Psych highlight the complexities involved, they also illuminate pathways to achieving cognitive machines that can reason about the world in ways that approach human understanding. The challenge is integrating common sense seamlessly with the frameworks necessary for effective machine learning, ultimately propelling advancements that could revolutionize the field of artificial intelligence.