Exam 2: Physical symbol system hypothesis
2/11/2026
one way to view information processing perspective:
useful for understanding both artificial and biological minds
help ask questions
remain agnostic to whether bio minds are strictly information processing systems
stronger interpretation:
minds aren’t as studied from info processing angle
both bio and artificial are info processing systems
info processing is all you need to explain things like intelligence
60s/70s: Newell and Simon with first overarching theory for cog sci
follows info processing ideas, but uses them to articulate hypothesis about bio and artificial minds (Physical Symbol System Hypothesis)
Physical Symbol System Hypothesis (PSSH):
forms basis of almost all AI (good old-fashioned AI - GOFAI) until late 20th century
since 90s a diff form of AI has become more prominent (learn after exam 2)
meant for this to apply to biological intelligence
claim: being a physical symbol system is necessary and sufficient for intelligent action
if it’s not a PSS, it can’t act intelligently - if it is a PSS, it acts intelligently
implication: if a human acts intelligently, it must be a PSS, and if something is a PSS, we don’t need anything else to explain its intelligent behavior
PSS 4-point definition:
a PSS has symbols (elements that stand for something); these are physical patterns (these physically exist), ex. words on a page, bits in computer
these symbols can be combined to form symbol structures (more complex expressions that jointly stand for something), ex. sentences made of words, group of 0s and 1s that represent something
PSS has processes/rules for manipulating its symbols (creating them, changing them, combining them), these processes have to be purely mechanical: no intuition or insight allowed
these processes or rules of point #3 can themselves be represented as symbols within the system, ex. operations programmed into calculator
Newell and Simon were inspired by Turing’s work; Turing machines are textbook example of Physical Symbol System
Questions:
Q1 (point 1 and 2). Which parts of a Turing machine would correspond to the symbols and symbol structures of points 1 and 2?
symbols: actual symbols printed on the tape,
symbol structures: whole line of tape after computation, groups of symbols that jointly mean something
Q2 (point 3 and 4). Which part of a Turing machine would correspond to the stored rules of points 3 and 4?
the Turing machines table to follow the program through it’s process, for point 4, the Turing machines table uses the programming of the 0s and 1s that the machine is reading from the input, the system knows what to do based on whether is it a 0 or 1
rules contained in machine table, machine has programming that responds to those rules
machine head allows machine to see what is on the tape
Q3 (point 3 and 4). We have said that “representations” and “algorithms” play a central role in cognitive science. Which of Newell and Simon’s concepts corresponds closely to representations, and which one to algorithms?
representations: first 2 points, all the numbers are representations (can stand for other numbers, pixels, etc)
algorithms: point 4, purely mechanical, no insight, rules are concrete
the PSSH consists of 4-points definition of a physical symbol system, but the understanding that being a PSS is necessary and sufficient for intelligent action = the claim is that the human mind is a PSS
Newell and Simon summarized view:
a physical symbol system is a machine that produces through time an evolving collection of symbol structures
thinking is simply transforming symbol structures from one into the other according to rules
Newell and Simon 2 key components to intelligence:
essence of intelligence is the ability to solve problems (ability to work out, when confronted with a range of options, which option best matches certain constraints)
all AI machines they built are meant to solve problems
intelligence requires generality: an ability to respond effectively to a wide range of conceivable situations, not just being effective in some restricted domain (others have also thought generality is key component of intelligence)
put together: intelligent system has a general ability to solve many different problems it may be confronted with
Newell and Simon’s General Problem Solver:
example of good old-fashioned AI
purpose: solving problems
fairly abstract description of what a problem is, so using their approach many different problems can be phrased in abstract terms that the GPS can deal with
terms in which problems need to be phrased
states:
initial state: ex. mouse wants to get to cheese, but there is a maze separating them
solution state: ex. mouse gets the cheese
all other states in between: ex. the mouse in the maze
permissible transformations: an action that connects one state to another different state, that changes the situation from one state to a different state
ex. mouse moving up in the maze, moving right, etc.
in the language the GPS speaks. a problem is a list of states, one of them initial, one of them solution, and a list of permissible transformations
in the language of GPS, a solution is a sequence of permissible transformations that transforms initial state into solution state
Point of PSS according to N and S:
take the symbol structure that represents input
use the system’s processes/rules for manipulating the symbols in that symbol structure, until you have the symbol structure that represents the output
GPS does this by:
takes symbol structure that represents problem: set of states and permissible transformations
uses some processes/rules for manipulating the symbols in that symbol structure, until it has the symbol structure that represents the solution: a series of permissible transformation that link from initial state to solution state