Computationalism Flashcards
What is computation?
- The manipulation of symbols.
- Computation provides the rules for manipulating those symbols on the basis of their shapes.
- Computation is logic, and logic is purely formal.
- What does purely formal mean? The symbols that are manipulated are representational units, i.e. symbols “stand in” for something else, however no necessary relation between the symbol and what is represents.
- To perform computation, you need input / output mechanisms.
- Limitations of computation: Limited in terms of memory, processing-speed and time. Can be controlled or automatic.
- Three requirements:
- Three requirement: implementation-independent, systematically interpretable and formal symbol manipulation.
What are the three requirements for computation?
- Implementation-independent: Computation can be implemented in multiple ways.
- Formal symbol manipulation: You can manipulate the shape of the symbols, but not the meaning. We attribute meaning to the symbols. The symbols are arbitrary in relation to what they represent. E.g. different number systems.
- Systematically interpretable: Computation uses syntax that is constant, stable and consistent. The same output despite different representations. Rules/algorithms specify the computation. The system means something within the system.
What is computationalism?
- Cognition is computational process.
- Cognition is
1) implementation-independent
2) systematically interpretable
3) symbol manipulation - Ergo, we can talk about the mind with no reference to the brain. In principle, anything could have a mind and not necessarily a brain.
- Goal: Make theories of human computation that have the same functions and break down the same way as human cognition.
What can we use computationalism for?
- The mind is a complex information-processing system. We can use computationalism to analyse such complex phenomena, make them more accessible and understandable.
- Point: Investigating how specific areas of the brain work (e.g. the neural activity, synapses etc.) does not tell us how the mind works. In the same way, only investigating the behaviour of humans or the physical properties of the brain does not tell us how the mind works.
Who is David Marr?
- 1945-1980
- Neuroscientist
- Main research areas: Vision and the cerebellum
- Combined mathematics with deep insights on brain function.
- Died of Leukaemia, aged 35.
What does David Marr say about levels of analysis?
- “If one hopes to achieve a full understanding of a system as complicated as a nervous system (…), then one must be prepared to contemplate different kinds of explanation at different levels of description that are linked, at least in principle, into a cohesive whole, even if linking the levels in complete detail is impractical.” (From the book Vision 1982)
What are the three levels in David Marr’s levels of analysis?
- The computational level
- The representational and algorithmic level
- The implementational level
What is the computational level?
- The most abstract level.
- A precise formulation of the problem.
- The goal of the system / the computation.
- The output of the computation / the end-stage / the object that will be computed.
- What and why questions.
- Important step! More important to understand the nature of the problem than examining the mechanism/algorithms.
- Important point! Must be treated as an information-processing problem.
What is the representational and algorithmic level?
- What steps are taken to achieve the goal. How computations are carried out in steps.
- Representation of the system (input/output). Trade-off - certain representations are well-suited for specific algorithms and vice vers
- Algorithms, i.e. series of steps. The operations performed in the system.
- Example: Addition, i.e. there are multiple ways in which you can add 47 to 21.
What is the implementational level?
The physical system of the computation. How can the representation and algorithm be realized physically?
Describes how the steps are executed physically.
Physical processes such as neuroanatomy, activated brain regions, synaptic mechanisms, action potentials, inhibitory interaction, neurophysiology.
Language as an example of a computational system.
Computational level: Goal of language is producing sentences. The definition of a sentence is a series of words in a specific order. Defining rules with syntax and grammar.
The representational and algorithmic level: The process, the thing that the mind does step by step. Combining words into sentences. What kinds of representation do we use? What language?
The implementational level: Neuroscience - which areas of the brain is activated when producing and comprehending language, e.g. Broca’s and Wernicke’s area.
Human recognition as an example of a computational system.
Computational leve: Patterns of human recognition. The output is the face.
Representational and algorithmic level: The processes we use to recognise a face. Do we look at each feature of the face and combine them together? Or do we use a more holistic approach?
The implementational level: Physical activity in the brain when recognising faces. Occipital lobe for simple stimuli, temporal lobe for faces, fusiform face area.
What is artificial intelligence?
A field that aims at constructing a “thinking machinery”, i.e. computing machines that can execute core mental tasks such as reasoning, decision-making, problem solving.
Examples: Chess player, driverless car
What is strong AI?
Thinking is manipulation of formal symbols.
By designing the right program with the right inputs and outputs, you are literally creating thinking minds.
Minds are implementation-independent.
The Turing Test can be used as a scientific test to determine success or failure.
Goal: To design programs that will simulate human cognition in a way so that it can pass the Turing test.
“The mind is to the brain as the program is to hardware”.
What is weak AI?
Computer models are seen as useful tools for studying the mind in the same way that they are useful in studying the weather, economics etc.