The Thomistic Institute Can a Machine Understand?: ChatGPT, Knowledge, and the Nature of Understanding – Prof. Tomás Bogardus
9 snips
Jan 12, 2026 In this insightful discussion, philosopher Tomás Bogardus explores whether machines can truly understand. With a background in metaphysics and epistemology, he delves into the distinctions between types of knowledge and the implications of large language models (LLMs) in our understanding of truth and belief. He argues that LLMs, while capable of producing language, lack genuine understanding and consciousness. Bogardus warns about the risks of over-reliance on technology potentially degrading human thinking skills while acknowledging LLMs' value when truth-focused.
AI Snips
Chapters
Books
Transcript
Episode notes
Knowledge Requires Truth-Explaining Grounds
- Tomás Bogardus defines propositional knowledge as believing a true proposition because it's true, not by luck or irrelevant reasons.
- Knowledge requires that the truth explains why you believe the proposition, ruling out lucky guesses and misleading evidence.
Understanding As Knowledge-Why
- Bogardus characterizes understanding as knowledge why: knowing that a correct explanation is true.
- Understanding is a specific kind of propositional knowledge tied to explanatory correctness.
How LLMs Predict Tokens
- Bogardus explains LLMs as large token-prediction systems that map text to numbers and compute next-token probabilities.
- He uses a carnival Plinko analogy: neurons act like pins tuned by training to steer outputs toward desired tokens.
