From: lexfridman
In a conversation with Vladimir Vapnik, co-inventor of support vector machines and a foundational figure in statistical learning, the intricate relationship between philosophy and the nature of intelligence was explored. This discussion delved into the philosophical underpinnings of artificial intelligence (AI) and the distinction between engineering and understanding intelligence.
Two Paths: Engineering vs. Understanding Intelligence
The discourse on intelligence often contrasts the engineering of intelligence with the understanding of its essence. Vapnik highlighted that engineering is the imitation of human activity, creating devices that mimic human behavior. On the other hand, understanding intelligence is a complex philosophical inquiry into what intelligence truly means and involves [00:03:11].
Predicates and Intelligence
Vladimir Vapnik drew connections from philosophical ideas to intelligence through the concept of predicates, inspired by Vladimir Propp’s work on folklore narratives. Propp identified structural units, or predicates, that define story narratives. Vapnik suggested that predicates might be universal ideas that could simplify understanding intelligence by breaking down human behavior into fundamental units [00:04:49].
Predicates in the context of AI represent the ideas that help construct the “invariants” of intelligence, potentially offering a foundation for understanding complex human behaviors and tasks involving AI [00:20:01].
Turing’s Influence and the Imitation Game
The conversation revisited Alan Turing’s contribution to AI with the imitation game, which laid the groundwork for machine reasoning and imitation over understanding. Vapnik acknowledged Turing’s profound impact, noting that while creating intelligent machines is paramount, a true understanding of intelligence eludes us. This distinction reflects the philosophical depth required to traverse from mere imitation to genuine comprehension [00:05:18].
The Plato Connection
Philosophical foundations trace back to Plato’s theory of forms, positing that a world of pure ideas exists beyond material reality. Vapnik draws from this lineage, suggesting that intelligence resides in the realm of ideas, and the task of understanding intelligence involves projecting these abstract ideas into the comprehensible world [00:08:11].
Weak Convergence and Learning
A key concept discussed was weak convergence, which Vapnik regards as central to forming an ‘admissible set of functions’ that define intelligence. Weak convergence, in contrast to strong convergence, concerns the overall property of functions rather than specific inputs, thereby serving as a philosophical and mathematical lens through which intelligence can be defined and understood better [01:18:02].
The Role of Philosophy
Philosophy influences machine learning by fostering an understanding of life and abstract ideas, serving as the bedrock for transforming such abstract notions into tangible intelligence [01:25:02]. For Vapnik, philosophy is intrinsic to understanding intelligence, as it aligns with utilizing and formulating abstract predicates that can encompass a wide array of intelligible functions.
Conclusion
The discussion with Vladimir Vapnik emphasized the philosophical and mathematical synthesis required to approach the nature of intelligence. It brought to light the intricate dance between creating machines that act intelligently and truly understanding the essence of intelligence — an endeavor as much philosophical as it is scientific. This dual emphasis reflects ongoing exploration in the fields of AI and philosophy, considering the deeper implications of machine reasoning and the abstraction of human cognitive processes.