University:
Boston University
Course:
CAS PH 630 | American Philosophy
Academic year:

2023
Views:

23

Pages:

15
Author:

Annabel Combs

Revisiting the Turing Test: Humans, Machines, and Phraseology Table of Contents Wittgenstein and Turing Turing 1936 The Turing Test Turing 1939-1954 We Live in Turing’s World Main Claims 1. Turing’s “Imitation Game” (1950), the “Turing Test”, was intended to be a social experiment in the evolution of human-to-human “phraseology” in the presence of machines: a language-game, in Wittgenstein’s sense. 2. Turing regarded“intelligence” as a response-dependent, emotional, and hence social concept, like “freedom” or “agency” (Turing 1948). 3. For Turing it is human-to-human interaction in the presence of computational technology that matters, not merely human-machine interaction (Turing 1948). Claims I Reject 1. The Turing Test answers the question of whether an individual consciousness or sentience might be had by a machine. 2. The Turing Test’s primary point is epistemological: can machines fool humans by masquerading as human beings? 3. Turing was a computational mechanist and/or a functionalist about the human mind. 4. Wittgenstein and Turing were mutually “alien” to one another: on opposite sides of a philosophical dichotomy between methods of ordinary language and methods of formal logic (Monk). My Argument • Wittgenstein and Turing shared a matrix of foundational philosophical ideas about the nature of logic. • They also discussed the nature, limits, and foundations of logic over many years. • They drew from one another, as they both recognized. • We have here a philosophical confluence of ideas forged over many years, not a conflict. Wittgenstein → Turing • Turing focussed on taking what we say and do with words seriously, and on the limits of formal methods, not only their power. • Everyday language, including our “typings” of objects as they occur naturally in science and everyday life, are an evolving framework or technology. Turing stressed everyday human conversation and “phraseology” as foundational, the social and cultural dimensions of computations and algorithms. Turing and Philosophy: Undergraduate 1931-4, Cambridge University 1. March 1933 (age 21) Turing read Russell’s Introduction to Mathematical Philosophy, encountering the notion of philosophical “phraseology” and Wittgenstein’s idea that logic is tautological. He began studying logic seriously. 2. Fall 1933-1935 Wittgenstein cancelled his “Philosophy for Mathematicians”, creating a brouhaha: the group was too large. He dictated The Blue and Brown Books to a small group of mathematics students instead. 3. Turing spoke to the Moral Sciences Club in December 1933, arguing that “the purely logistic view of mathematics” is “inadequate” because there are many different ways to interpret mathematics (i.e., different phraseologies). 1936 Turing gives a fundamentally persuasive analysis of “computation” by inventing the idea of (what was soon called) a Turing Machine. The human interface, the human context of a shareable command, is demonstrated to be fundamental to the nature of computation. How did Turing do this? Cambridge 1934-6 (Turing’s Graduate Fellowship Years) Wittgenstein’s Blue and Brown Books (1933-35): • Construed logic anthropologically, in terms of “language-games” in which humans operate “mechanically” with signs, step-by-step, following command-tables. The question of general training is raised. • “Can a machine think?” is a grammatical question requiring careful investigation, it is not yet clear: we can e.g. say that there is “thinking in the hand’, but this does not refute the idea of human consciousness as irreducible. Human beings act both “mechanically” (e.g., as workers), and “creatively”. • Investigation of the concept of “thinking” must attend to how concepts are used in everyday life. Grammatical “experiments” with specific, sometimes invented “phraseologies” – i.e., comparison of different language-games – is the method. Turing 1936: “On computable numbers, with an application to the Entscheidungsproblem” Turing’s greatest paper turns on Wittgenstein’s method of language-games (Floyd 2012). Wittgenstein’s Brown Book §41: Hilbert’s Entscheidungsproblem (1928) Turing solved the following problem: Is there a definite method (algorithm) for deciding, Yes or No, whether or not a sentence of a theory couched in a formal system of logic follows from the axioms? (Hardy (1929): Do mathematicians make discoveries by turning the handles of a miraculous machine?) Turing’s 1936 answer to Hilbert: No! 1. There exists no general algorithm (no “definite method”) to decide whether or not one sentence follows from another in a logic. “Thinking” is not in general reducible to algorithms. 2. More generally: there is no general algorithm to determine in advance what an algorithm will do on a particular input: the idea of working out the consequences of a concept or set of equations cannot be in general foreseen (Wolfram, A New Kind of Science: computational irreducibility). 3. Turing (1954): This shows that there will always be a need for humans to use common sense. Turing’s Analysis of a “Definite Method”: A Basic Logical Point To obtain a positive result that there is a machine or algorithm or “definite method” that can accomplish a task, you must write down the method or algorithm, make the machine, or at least prove that it exists. But to obtain a negative result, that something is impossible, you must give a persuasive general characterization of what doing that something would have looked like.