10 - Language of thought (1)

Minds, Brains & Machines 🧠🤖

Classical Architectures: Physical Symbol Systems (PSS) 🤔

The Physical Symbol System Hypothesis says that a physical symbol system is needed and enough for intelligent action. This idea tries to explain intelligence as computation by using symbols.

  • Necessity claim: Intelligent systems must be PSSs.

  • Sufficiency claim: PSSs are enough for intelligence.

  • Key aspects of a Physical Symbol System:

    • Symbols: Physical patterns.

    • Complex symbol structures: Symbols can be combined.

    • Symbol manipulation processes: The system contains processes for manipulating symbols and complex structures.

    • Self-representation: The processes for generating and transforming complex symbol structures can be represented by symbols within the system.

The Computational Theory of Mind 🧮

This theory asserts that the mind is a computer. Understanding the mind requires addressing three levels:

  • Computational level: What function is being computed?

  • Algorithmic level: How are these problems solved? What procedures and representational codes are used?

  • Implementation level: What physical mechanisms implement these processes? This relates to the brain's structures.

Examples of PSS đź’ˇ

  • The +1 Machine: A simple example illustrating a PSS. It takes an input number and adds 1, producing an output.

  • Syntax as computation: Parsing sentences can be modeled as a PSS, with operations like copying, deleting, and pasting symbols (words or phrases) to produce a representation. Ohta, Fukui, and Saka (2013) provide a model demonstrating this.

  • Marr's primal sketch: This visual processing model can also be viewed as a PSS. It involves taking luminance values and applying rules and assumptions to derive representations, such as edges.

Functional Decomposition & Primitive Processors

  • Functional decomposition: Breaking down a cognitive process into subcomponents, each analyzed computationally and algorithmically. This continues until reaching primitive processors—components that cannot be further decomposed computationally. At this point, understanding shifts to the natural sciences (neuroscience, etc.) to explain how intelligence is realized physically.

    • Example: Multiplication can be broken down into simpler operations like addition. Ultimately the process would break down to something like AND-gates.

    • Example: Understanding Language can be decomposed into processes like word recognition, syntax recognition, parsing, and semantic interpretation.

Key Points to Remember đź“Ś

  • An algorithm computes a function by deriving a representation of the value from a representation of its arguments.

  • Not all functions are computable.

  • Any computable function can be computed by a Turing Machine, but not all computers are Turing Machines.

  • A Universal Turing Machine can compute any computable function given the correct program.

  • The PSS Hypothesis states that the mind is a PSS, suggesting symbol manipulation is sufficient for intelligence.

  • Mental functions can be decomposed to primitive processors understood as physical systems (like AND gates).

The Tower Bridge of Computation 🌉

Cummins (1996) proposed a model represented visually as a tower bridge, showing the different levels of analysis: Input, Representation, Algorithm, and Output. This model visually summarizes the relationship between the different levels of computation, from input to output, emphasizing the role of representations and algorithms.

10 - Language of thought (1)

Cognitive Architectures: The Language of Thought

Course: COGS/PHIL 2160Key Topics:

  • Cognitive Architectures: Different frameworks to understand the mind.

  • Classical vs. Connectionist Architectures: Traditional symbol manipulation versus neural network approaches.

  • Language of Thought: Can machines truly think?

  • Symbol Manipulation: Physical symbols represent mental states (beliefs and desires) influencing behavior.

  • Physical Symbol Systems (PSS):

    • Necessity: Intelligent systems must be PSSs.

    • Sufficiency: PSSs alone are enough for intelligence.

    • Key figures: Allen Newell & Herbert Simon.

The Role of Symbols:

  • Symbols are the building blocks of mental representations.

  • Syntax vs. Semantics: Understanding the difference between the structure of symbols (syntax) and their meanings (semantics).

  • Symbols facilitate infinite thought processes through systematic combinations.

Causation and Meaning:

Exploring the relationship between beliefs, desires, and actions through logical mapping. The Language of Thought Hypothesis emphasizes how structured physical states represent meaning and dictate behavior.

Summary Points:

  • Mental state explanations depend on understanding symbolic meanings.

  • Productivity and Systematicity: Language and thought can create endless representations from limited symbols.