Minds, Brains, and Computers

• Instructor: Zhexi Zhang
• Email: [email protected]
• Office Hours: Tuesdays 12pm-2pm in Room 234, Department of Philosophy or Zoom
• Course focuses on minds, brains, and computers

Course Content

• The mind is that which thinks (believes, desires, etc.)
• Course addresses thoughts as the content of mental states
• Course will examine the transitions between thoughts, such as reasoning, inferentialtransitions, and associative thinking
• One hypothesis is that the mind is a computational system
• Computational theory of mind indicates the mind processes information
• Studying the mind with this approach requires cross-disciplinary methodology
• It is hypothesized that the human mind is physically implemented by the human brain
• Focus on Artificial Intelligence (AI) to determine how to build a thinking computer
• One approach involves Turing-style AI programs like The Logic Theorist
• Another option involves connectionist networks or artificial neural networks
• Week 1-3 covers the classical computational theory of mind, involving rule-governed formal systems manipulating physical symbols based on syntactic properties
• Week 4-7 discusses how computational theory relates to intentional explanations, focusing on the semantic content of psychological states
• Week 8-9 examines the relationship between computational theory and neuroscientific theories, especially the connectionist theory as a rival to the classical theory of mind
• Week 10 focuses on how computational theory has largely focused on digital computers, while analogue computers involve covariational relationships between representational vehicles and content
• Focus on if formal/digital, connectionist/artificial neural and analog systems can produce intelligent action and intentionality

Newell & Simon Key Concepts

• A. Newell and H. A. Simon's work "Computer Science as Empirical Inquiry: Symbols and Search" (1976) is studied
• Topics covered are necessary and sufficient conditions, Physical Symbol Systems (PSS), Physical Symbol System Hypothesis (PSSH), General intelligent action and Heuristic search and problem solving

Necessary and Sufficient Conditions

• Necessary conditions must be met for something of interest to be the case
• Example: A is necessary for B when B cannot occur without A
• Falsity of A guarantees the falsity of B if A is necessary for B
• Sufficient conditions are enough for something of interest to be the case by itself
• Example: A is sufficient for B when A's occurrence ensures B's occurrence
• Truth of A guarantees the truth of B if A is sufficient for B

Examples of Necessary and Sufficient Conditions

• Being a mammal is a necessary but not sufficient condition for being a dog
• Being a German shepherd is a sufficient but not necessary condition for being a dog

Question Regarding the Conditions for Having a Mind

• Potential controversial necessary conditions are exhibiting intelligent action, having intentionality (thoughts), emotions, reasoning capacities, consciousness, and self-reflection

Definition

• Precise definitions combine necessary and sufficient conditions
• Necessary conditions provide criteria for belonging to a category/concept
• Sufficient conditions guarantee something belongs to a category/concept
• A good definition requires more than "being a mammal" to define "being a dog."

The Mind - Key Ideas

• Is hard to have a precise definition of the mind
• Newell & Simon suggest general intelligent action is a necessary and sufficient condition
• Physical Symbol System hypothesis also discussed

Scientific Methodology

• Newell & Simon studied symbolic systems, focusing on software
• Examines structural requirements for general intelligence
• Studies qualitative characterizations/laws enabling general intelligent action by look at other disciplines

Newell & Simon Hypothesis

• A physical symbol system has necessary and sufficient means for general intelligent action
• Empirical data is collected from research in artificial intelligence and cognitive psychology
• Their hypothesis is an empirical generalization supported or challenged by evidence

Physical Symbol System Hypothesis

• PSSH states a physical symbol system has necessary and sufficient means for general intelligent action
• Any system that exhibits general intelligent action is a physical symbol system
• Physical symbol systems of sufficient size can be organized to exhibit general intelligent action

Physical Symbol Systems

• Physical patterns implement symbols
• Symbols carry information about the world
• Physical symbol systems combine symbols to form or change structures by following algorithms

Turing Machine

• Turing machine is an important example of physical symbol system
• Machines manipulate 0s and 1s using instructions to transform symbols and transition to the next state
• The current state, input symbols, and rules determine the actions
• Primary actions are writing symbols, moving left/right, and changing states
• The Turing machine is an abstract device that requires infinite memory and time for computation
• Turing proved the computability/uncomputability of some mathematical problems

Digital Computers as PSS

• Programs are converted into machine code for the CPU
• Transistors act as switches, representing binary symbols (0/1)
• Binary symbols combine into complex strings following syntactic rules
• Algorithms direct computers to convert binary strings

PSS Computations

• PSS computations are governed by algorithmic rules
• Rules are specified in a finite number of steps
• Rules are unambiguously followed by a computer
• Yields an output for any valid input
• Physical symbols do so without understanding of the symbols' meaning and operates purely syntactically

Physical Symbol Systems - Transitions in Thoughts

• The semantic content of thoughts explains thought processes
• Algorithmic rules are followed without semantic understanding
• Thought processes requiring semantic understanding cannot be algorithmic

Formal Logic and PSS

• PSS operates on the syntactic/logical structure of thoughts regardless of semantic meaning
• Logicians formalized thoughts using meaningless tokens according to syntactic rules
• Formal logic utilizes strings of symbols that have semantic meaning to represent information
• These same symbol strings can be transformed according to algorithmic/purely syntactic rules without needing processes for semantic meanings

Physical Symbol System Hypothesis - Key Points

• Hypothesis states physical symbol system has the necessary and sufficient means for general intelligent action
• General Intelligent Action provides the same scope of intelligence with limits to speed and complexity as human action

General Intelligent Action

• Early chess programs were successful against human players, but were not based on neural networks
• Systems must generate potential solutions, test structural symbols, solve the problem with representation and have limited processing capabilities to arrive at a solution quickly by using heuristic search

General Problem Solver (GPS)

• GPS was built in 1956
• GPS represents the problems, strategies to generate potentia solutions and the goal
• The means-ends analysis plays a central role in test
• The GPS detects differences between current/goal situations and selects what reduces said differences

Example: The Traveling Salesperson Problem

• 24 identified routes with constraints on initial state and the goal - this is the problem!

PSS and "general intelligent action"

• Analogy - Problems previously faced could be solved using the means end analysis which can assist with identifying potentials, evaluate difference, and select what reduces it
• Early AI can solve problems just like humans with comparable thought processes

PSS - Concerns

• Means end is useful but problem can easily be HUGE - Heuristic may be useful in rapidly accessing info
• Issue - There is no clear compatibility between Semantic processing with syntactic
• Searle's Chinese Room (Week 5) argues against this possibility.