PHI 13 FINAL

ROBOT REPLY RESPONSE

- Computer in a _____ body could attach meaning and potentially understand natural language in a freed environment (think of child learning acquisition) The _____ concedes Searle is right about the Chinese Room scenario: - It shows that a computer trapped in a computer room cannot understand language, or know what words mean. - The _____ is responsive to the problem of knowing the meaning of the Chinese word for a hamburger—Searle's example of something the room operator would not know.

SYSTEMS REPLY RESPONSE ("MEMORIZING THE BOOK")

- Concedes that the man in the room does not understand Chinese. But, the _____ continues, the man is but a part, a central processing unit (CPU), in a larger system. The larger system includes the huge database, the memory (scratchpads) containing intermediate states, and the instructions—the complete system that is required for answering the Chinese questions. So the _____ is that while the man running the program does not understand Chinese, the system as a whole does. Searle's Response: - In principle, the man can internalize the entire system, memorizing all the instructions and the database, and doing all the calculations in his head. He could then leave the room and wander outdoors, perhaps even conversing in Chinese. But he still would have no way to attach "any meaning to the formal symbols". The man would now be the entire system, yet he still would not understand Chinese. For example, he would not know the meaning of the Chinese word for hamburger. He still cannot get semantics from syntax.

SEARLE AND THE CHINESE ROOM

- Counter-argument to Turing Test - The thought-experiment: + _____ imagines himself alone in a room following a computer program for responding to _____ characters slipped under the door. _____ understands nothing of _____ and yet, by following the program for manipulating symbols and numerals just as a computer does, he produces appropriate strings of _____ characters that fool those outside into thinking there is a _____ speaker in the room. + The narrow conclusion of the argument is that programming a digital computer may make it appear to understand language but does not produce real understanding (this is the distinction between syntax and semantics). + The broader conclusion of the argument is that the theory that human minds are computer-like computational or information processing systems is refuted. Instead minds must result from biological processes; computers can at best /simulate/ these biological processes. The claim is that even if reasonable natural language responses are being generated that are indistinguishable from ones a native _____ speaker would generate, there is no "understanding" since only meaningless symbols are being manipulated

INFORMALITY COUNTERARGUMENT

- Definitive rules of conduct where man has to regulate themselves, they would be no better than a machine, however if there are no such rules, man cannot be machines 3 laws: - Laws of conduct + It seems to run something like this: "If each man had a definite set of rules of conduct by which he regulated his life, he would be no better than a machine. But there are no such rules, so men cannot be machines." The undistributed middle is glaring. I do not think the argument is ever put quite like this, but I believe this is the argument used nevertheless. There may however be a certain confusion between "rules of conduct" and ''laws of behavior" to cloud the issue. - Law of behavior 1. By "laws of behavior" I mean laws of nature/biological as applied to a man's body such as "if you pinch him he will squeak". If we substitute "laws of behavior which regulate his life" for "laws of conduct by which he regulates his life" in the argument quoted the undistributed middle is no longer insuperable. For we believe that it is not only true that being regulated by laws of behavior implies being some sort of machine (though not necessarily a discrete state machine), but that conversely being such a machine implies being regulated by such laws. 2. Human beings are informal + We have no explicit procedures for the unexpected, only for what is expected (eg. stopping and waiting for green at traffic lights ) + /It is not possible/ to produce a set of rules purporting to describe what a man should do in /every conceivable set of circumstances./ Argument and Turing's Response: - Argument: It is not possible to produce a set of rules for every moment a man accomplishes a task, if he did have rules, he would be like a computer, but he isn't, so they're not alike - Turing's Response: Just because we don't follow rules of behavior, doesn't mean we don't follow rules of conduct just like a computer!

CONTINUITY COUNTERARGUMENT

- Digital like pixels: distance and well defined, discrete (Discrete-state system, VN architecture, Turing machines, UTM, etc.) analogue: not discrete/all smeared states like a record (human brain works this way) Argument: - Humans are made of a _____ system, not a discrete system where everything is done serially, human brains compute in parallel. Therefore, computers can't be intelligent because parallel machines can compute serially but not vice versa (searle article). Turing's Response: - The imitation game would in no way be a _____ system, but if played properly, it wouldn't even matter because nobody *should* tell the difference between the computer and the human (that's the point of the /imitation/ game)

PAUL CHURCHLAND AND NEURAL NETS

- Highly Simplified model networks have been useful in suggesting how real neural networks might work and in revealing the computational properties of parallel architectures. - This network is a device for transforming any one of great many possible input vectors(activation patterns) into a uniquely corresponding output vector It is a device for computing a specific function - Exactly which function it computes is fixed by the global configuration of its synaptic weights - The model network over simplifies the structure of the brain, but it does illustrate several important ideas + Parallel architecture provides a framatic speed advantage over traditional computer 1) Advantages increases as the network grows in size 2) Processing dependent of both number units + Massive parallel system means the system is fault-tolerant and functionally persistent = graceful decay + Stores large amounts of information that can be accessed in milliseconds - Not all parallel processing is ideal for all computational functions therefore classical symbol manipulation machines are typically superior - Parallel systems are described as not manipulating symbols according to structure sensitive rules + It's one of the many cognitive skills that a network may or may not learn to display + /Rule governed symbol manipulation is not its basic mode of operation / - /Searle's/ argument is directed against SM machines:connection weights (vector transformers) are not threatened by the Chinese room argument + Believes parallel processing to be devoid of real semantic content + Outlines the failure of parallel processors with the chinese gym proceeding his argument of the chinese room + Neglects to mention that his simulation requires 10^14 when the human brain has 10^11 neurons and averages over 10^3 connections + His response will require 10000 earths therefore one gym is not a fair simulation + Agrees though he qualifies his claim by saying " any other system capable of causing minds would have casual powers (st least equivalent to those of brains" 1) Agrees that artifical mind does not need all casual powers of the brain, just those relevant to conscious intelligence Conclusion of article: - Searle bases positon on common sense intuitions about the presence or absence of semantic content - Churchlands base theirs on the specific behavioral failures of the classical SM machines and on the specific virtues of machines with a more brainlike architecture - The brain is making systematic use of computational advantages, but its not the only physcial system capable of doing so + The machine can be of any medium with parallel processing + The contrasts between their opposing views show computational strategies have vast and decisive advantages over others

COMBINATION RESPONSE

- Searle is the whole system - _____ of both the system and robot responses + Robot = processing by the ability to sense and behave + System = simulating the original but in different narratives - Assessing the examples of the responses relying on intuition - If Searle already shot down the other 2 arguments then why would he accept a combination of the 2? - Under the rubric "_____", Searle also considers a system with the features of all three of the preceding: a robot with a digital brain simulating computer in its cranium, such that the system as a whole behaves indistinguishably from a human. + Since the normal input to the brain is from sense organs, it is natural to suppose that most advocates of the Brain Simulator Reply have in mind such a combination of brain simulation, Robot, and Systems Reply and therefore some argue it is reasonable to attribute intentionality to such a system as a whole. ++ Searle agrees that it would be reasonable to attribute understanding to such an android system—but only as long as you don't know how it works. As soon as you know the truth—it is a computer, uncomprehendingly manipulating symbols on the basis of syntax, not meaning—you would cease to attribute intentionality to it.

NEWELL, SIMON, AND PSS

A physical symbol system (also called a /formal system/) takes physical patterns (symbols), combining them into structures (expressions) and manipulating them (using processes) to produce new expressions. PSS Hypothesis: - It is a position in the philosophy of artificial intelligence formulated by Allen Newell and Herbert A. Simon. - A physical system has necessary and sufficient means for general intelligence. 1) Implies both that human thinking is a kind of symbol manipulation (because a symbol system is necessary for intelligence) and that machines can be intelligent (because a symbol system is /sufficient/ for intelligence) - A physical system: 1) Obeys laws of physics, realized by engineered systems made of engineered components 2) Not restricted to human symbol systems - /Necessary/: any system that exhibits general intelligence will prove upon analysis to be a PSS 1) Therefore, humans must be PSS! 2) Part of a sufficiency test for intelligent action - /Sufficiency/: given that it is necessary for intelligence that the system is a PSS, any PSS can be developed into an intelligent system provided it is of sufficient size. - Physical: clearly obeying laws of physics, realizable by engineered systems made of engineered components - Symbol: An entity, a physical pattern that can occur as a component of another type of entity, not restricted to human symbol systems - Symbol System: a set of entities - /Designation/: expression designates an object, if the system can affect the object or behave in ways depending on the object - /Interpretation/: given an expression (set of symbols), the system can perform the process PSS's use heuristic search to solve problems ("games" from lecture) - Heuristic search: the solutions to problems are represented as symbol structures. A physical symbol system exercises its intelligence in problem solving by search, i.e., generating and modifying symbol structures until it produces a symbol structure 1) Means-ends problems - A means-ends problem first determines the goal then steps to achieve that goal. For instance, getting from point a to b using smaller steps to reach that goal. Each step brings you closer to the goal. 2) Generate solutions via framing: + Frames are information structure for stereotyped situations - How the Turing Machine follows PSS: syntactic processing, stored program concept in computers (interpretability of symbols, & heuristic search)

DESIGN STANCE

A second method, which involves taking the "_____", is to attribute functions to the system or its parts and to predict that the system will function properly. We can predict that a jogger's pulse will increase as she heads up the hill because of what we know about exercise and the proper function of the circulatory system. Dennett describes an alarm clock. Its function or _____ is to wake a person up at a designated time. It has a button for snooze. It is _____ to execute a function. (In other words, ignore physical components and just on basis of _____, predict that system will behave as it is _____ to behave)

INTENTIONAL STANCE

A third method, which involves taking the "_____", is to attribute beliefs and desires to the person, and then to predict that he will behave rationally, given those beliefs and desires. In Dennett's view, a system with beliefs is a system whose behaviour, while complex and difficult to predict when viewed from the physical or the design stance, falls into patterns that may be captured with relative simplicity and substantial if not perfect accuracy by means of the _____. (Figure out what the object's purpose is, what the beliefs and desires it ought to have are with respect to its function in the world (e.g., stop sign on the street versus it being washed up on the beach, it was designed for traffic), and predict that it will act to do what it was made to do)

DISABILITIES COUNTERARGUMENT

Argument: - Computers lack feature x, and x is important and necessary*** for intelligence, so computers are therefore not intelligent - Examples of x: + "Be the subject of its own thoughts" + "Make mistakes" + "Do something really new" Turing's Response: - These _____ are irrelevant because it is an unfair comparison. If you compare a race between man and the plane, the plane will obviously win with respect to speed being an example of feature x

LADY LOVELACE COUNTERARGUMENT

Argument: - Computers lack originality (ex/ thoughts), and are therefore not intelligent Turing's Response: - How do you know if the work of a human is truly original, let alone a computer's?

CONCIOUSNESS COUNTERARGUMENT

Argument: - To be sure that machines think is to be the machine Turing's Response: - This is the solipsist point of view: only way to know that a human is thinking is to be the human, and we can't be other people so this is not enough + He's saying that we might each have just as much reason to suppose that machines think as we have reason to suppose that other people think

TURING: CAN MACHINES THINK

Basic argument: - Brains are like computers (computationalist view): + In the brain we have a CPU that is distinct/separate from where the memory is stored AKA Von Neumann Architecture(Strong AI according to searle) - Brains are really just a bunch of algorithms that analyze and make decisions (outputs) Establishes Turing Test - Test background : + Designed to distinguish AI [Artificial Intelligence] from normal computers - Test design: + Imagine you are having a conversation with two people. One is a computer, the other is a human. It is debated as to whether or not you should know one is a computer. Regardless, if the computer carries on the conversation normally like a human, then it passes the Turing Test. Premises (Dorsey says this structure *should* be applied to every argument): - P1. Humans are intelligent have feature X - P2. Computers lack feature X - P3. Feature X is necessary condition for intelligence - C. Computers are not intelligent

FODOR, PYLYSHYN, AND COMPUTATIONALISM

Computationalism: - Strong: humans are /Good Old Fashioned PSS/. Look at the brain, identify the processor. Memory generates symbol structures in memory then searches it to solve problems. The brain is a PSS and have Von Neumann architecture 1) GOF PSS: Von Neumann architecture with a distinct memory store and processor. Solves problems by means-ends and search trees. - Weak: GOF are useful tools for studying humans or human cognition, it is a tool, and does not claim that the brain is a GOF PSS. Von Neumann architecture : - Is a theoretical design for a stored program computer that serves as the basis for almost all modern computers. 1) A von Neumann machine consists of a central processor with an arithmetic/logic unit and a control unit, a memory, mass storage, and input and output. - Processing of symbols and expressions are processed by a distinct processor. 1) Neural nets do not have an independent processor, they process in parallel - Manipulates symbols and expressions - UTM also manipulate symbols with symbols! LOTH : - Thought as language: language can be combined in multiple ways - Rules are syntactic (not semantic): do not have to tell you what components mean - Constituents = basic units of thought - Productivity: we can create many thoughts - Compositional - Grammar (of thought) 1) Purely formal 2) How to put PS together to form actual thoughts and not gibberish - Executor 1) The thinker 2) Knows grammar 3) Purely formal rules Computationalism vs. Connectionism - Neural nets can't be PSS because they have parallel processing (i.e., they don't have Von Neumann architecture) - Homunculus problem: if brains are neural nets 1) Thinker within the thinker's mind repetitive pattern, a reason why computationalist example of "what if there is a thinker in the brain doing the thinking" doesn't work VN Architecture - no parallel processing - avoid the homunculus problems Language of Thought (LOT): - basic units of grammar/ syntax - symbols do not need meaning Traditional Computers>Neural Nets

DREYFUS, MICROWORLDS, AND FRAMES

Dreyfus argued that human intelligence and expertise depend primarily on unconscious instincts rather than conscious symbolic manipulation, and that these unconscious skills could never be captured in formal rules. SHRDLU: Designed as a piece of AI - It's a 3D system that can d o things/perform tasks - simulates robot arm which can move a set of shapes - Humans engage with the computer ex/ ask questions, issues commands - SHRDLU uses syntax, semantics( bc/ people must give it commands in order for it to work, so it can be seen as "understanding"), and facts about the blocks to perform tasks - Restricted domain: hence name of micro-world (doesn't give insight to real world) - LIMITATIONS: fixed representation, no semantic depth, just carries out process (Searle says this in general about computers) - "It knows what it owns, but it doesn't know what it is like to own something" (pg 150) ...also Lady Lovelace's argument from the Turing section - Application of micro-worlds: computer vision

NEURAL NETS

Feed forward: activation is flowing only forward in the _____ - Not common of neurons in our brains Back propagation: trial and error format used to correct weights to desired amount - Nodes later in the net can affect nodes earlier in the net - Human brain has back propagation-but it's rare. Most NN depend on back propagation but this isn't really how neurons operate Basic features: - Dynamic: can change over time - Can change through feedback - Graceful decay: + Although some nodes may be lost overtime, it does not inhibit the overall productivity of the system

TURING MACHINE

How they work: - A _____ is an idealized computing device consisting of a read/write head (or 'scanner') with a paper tape passing through it. - The tape is divided into squares, each square bearing a single symbol--'0' or '1', for example. This tape is the machine's general purpose storage medium, serving both as the vehicle for input and output and as a working memory for storing the results of intermediate steps of the computation. - The input that is inscribed on the tape before the computation starts must consist of a finite number of symbols. - However, the tape is of unbounded length. The read/write head is programmable. - The _____ can read the symbol currently under the head, write a symbol currently under the head, or delete and rewrite if there is already one. Move the tape left or right one square, change state, or halt. - A program or 'instruction table' for a _____ is a finite collection of instructions, each calling for certain atomic operations to be performed if certain conditions are met. Big idea: can machines think? - Better representation of our brain processing - Read/write machine; can move left and right

FINITE STATE AUTOMATA

How they work: - _____ is a simple idealized machine used to recognize patterns within input taken from some character set. - The job of an _____ is to accept or reject an input depending on whether the pattern defined by the _____ occurs in the input in an automated fashion. - It has a finite set of states, a special start state, a set of final (or accepting) states, an input finite in length, and a set of transitions from one state to another labeled with characters. Example: - We begin in the start state. If the next input character matches the label on the transition from the current state to a new state, go to that new state. - Continue making transitions on each input character. If no move is possible, then stop (go to trash state), or if in accepting state, then accept. Big idea: - Very long, convoluted and taxing - Not a very good model of how our brain processing works - Read only machine (can't write or erase like a Turing Machine ) - Cannot handle infinite states such as a^n b^n because that would require counting: do not have memory !!!

DENNETT AND THE INTENTIONAL STANCE

Objectivism vs. interpretationism - Objectivism: + Objective facts. In class he discussed it like a virus. Either you have the virus or you don't. - Interpretationism: all in the eyes of the beholder. + Choose whether or not to have that belief. In class he discussed what makes a good spouse. It's very subjective because what one characterizes as a good spouse might totally be different from someone else.

PHYSICAL STANCE (H2O MOLECULE)

One method, which involves taking what Dennett calls the "_____", is to apply our knowledge of _____ law. We can predict that a diver will trace a roughly parabolic trajectory to the water because we know how objects of approximately that mass and size behave in fall near the surface of the Earth. A _____ is a _____ thing. It is made up of two _____ and one _____.

FRAME PROBLEM

_____ - _____ definition: an information structure for stereotyped situations 1) Ex: when you go to a restaurant you know you have to see if you need to wait to be seated, wait for a menu, decide on what to drink, etc. 2) Need to identify base things that are a part of stereotyped situations - Real AI cannot happen in microworlds- if SHRDLU was put in the real world (or a simulation of the real world), it would have no idea what to do 1) _____ are supposed to help move from microworld to real world when presented with an unfamiliar subject, a computer does not know what information is important and what is unnecessary - While a chatbot or search engine can reasonably function using pure word-to-word textual references, they falter when topics become more specialized; chatbots become incoherent, search engines tend to present only the most popular or obvious information.

SIMULATION (BRAIN SIMULATION) RESPONSE

_____ the computer to that of a native Chinese speaker's brain, and it should, in theory understand Chinese if each synapse, each firing, is the exact same (not duplicated, but _____) Searle's response: - Even getting this close to the operation of the brain is still not sufficient to produce understanding, because it's still _____ and not a duplication of the system - Water pipe _____ + The brain _____ is _____ the wrong things about the brain- as long as it _____ only the formal structure of the sequence of neuron firings at the synapses, it won't have _____ what matters about the brain: its ability to produce internal states

SEARLE, THE CHURCHLANDS, AND THE CHINESE GYM

_____' Argument ; - The luminous room argument + Imagine a man waving a magnet in a darkroom. Could waving the magnet around produce light? + Conclusion: Electricity and magnetism are neither constituted of nor sufficient for light 1) Axiom 1: electricity and magnetism are forces 2) Axiom 2: the essential property of light is luminance 3) Axiom 3: forces by themselves are neither constituted of nor sufficient for light - They say that _____ is not warranted to make the claim that you can't get semantics from syntax (_____'s 3rd axiom) + We don't even understand semantics- _____ can't say that what is happening in the _____ is not semantical - The only way to make the claim is to reverse engineer the brain, through parallel systems (neural nets) + Need to be able to build the _____ itself, and actually reverse engineer the brain to see if the program did understand _____ _____'s Argument: - _____ Thought Experiment (Chinese Room rework): + The thought-experiment-structure: 1) _____ now imagines he is inside a gym with several other men, none of whom speak Chinese. But they all follow instructions properly, and their responses fool whomever is outside the room into thinking they speak Chinese. 2) This system still does not understand Chinese + The meaning of the rework: 1) This is more a response to parallel processing and neural nets, no?