Best of the Pod: Reid Hoffman on How AI Is Answering Our Biggest Questions

Quick Overview

Reid Hoffman argues that philosophy is more crucial for entrepreneurship than an MBA because it teaches one to think crisply about possibilities and theories of human nature, which is vital when considering how new technologies like AI modify human existence and create new ecosystems. Hoffman views the intersection of AI and philosophy as a rich area, especially concerning perennial questions like truth and knowledge, and suggests that LLMs, functioning through next-token prediction, currently align more with late Wittgensteinian, pragmatic language games, even as developers try to ground them in essentialist truth conditions.

Key Points: Hoffman asserts that a background in philosophy is "more important for entrepreneurship than an MBA" because it forces clear thinking about possibilities and theories of human nature. AI prompts deep philosophical questions about how it might change "what it means to be human," enhancing creativity and intelligence. Hoffman critiques disciplinary "arianism" in academia, advocating for multi-disciplinary blending, noting that science evolved from philosophy by developing more specific theories. He explains that thought experiments like the trolley problem are often misused by forcing an intuition to derive a principle, ignoring the practical human response like trying to break the trolley. Current LLM function, predicting the next token, aligns with a late Wittgensteinian, pragmatic view of truth as social convention or what works, contrasting with earlier essentialist AI approaches. The process of refining LLMs involves grounding their pragmatic base in more essentialist characteristics to reduce hallucination and improve truth-telling, reflecting a Hegelian thesis (essentialism) and antithesis (nominalism) leading to synthesis. Reasoning is being integrated into models by training them on data sets that contain crisp reasoning patterns, such as computer code and textbooks, and exploring computational philosophy based on thinkers like Lakatos and Popper.