Presentation is loading. Please wait.

Presentation is loading. Please wait.

How Solvable Is Intelligence? A brief introduction to AI Dr. Richard Fox Department of Computer Science Northern Kentucky University.

Similar presentations


Presentation on theme: "How Solvable Is Intelligence? A brief introduction to AI Dr. Richard Fox Department of Computer Science Northern Kentucky University."— Presentation transcript:

1 How Solvable Is Intelligence? A brief introduction to AI Dr. Richard Fox Department of Computer Science Northern Kentucky University

2 Solvability Theory of Computation attempts to prove whether a given problem is solvable or not –A problem is solvable if the problem can be solved by an algorithm (e.g., a computer program) –There are problems that are known to be unsolvable No one can write a computer program to solve that problem –Question: are the problems that humans solve using intelligence solvable? That is, can we write computer programs to solve problems that humans solve using intelligence?

3 Artificial Intelligence Artificial Intelligence (AI) is a field which studies and explores methods for solving problems by computer program –The problems examined seem to require the application of intelligence to solve them Researching AI yields some very difficult problems: –No one has adequately defined intelligence so we don’t know if a given AI program solves the problem –We do not understand human intelligence well enough to properly write AI programs And human intelligence is our only model available to study

4 Is AI Solvable? Can we write computer programs that solve problems that seem to require intelligence? –These problems include diagnosis, story understanding, language processing, perceptual problems, planning/design/decision making, theorizing, learning and others –For example, can we prove that diagnosis is a solvable problem? If so, how do we go about solving it? –Lets first consider an unsolvable problem

5 The Halting Problem The Halting Problem (HP) is perhaps the most well-known unsolvable problem –Given a program and its input, when executed, will the program terminate (halt)? –We cannot write a computer program that could answer this question Program to solve the Halting Problem Data Yes (program halts) or No (program does not halt) Program

6 Do people solve the HP? While we cannot write a program to solve the HP, could a human (I.e. an experienced programmer) solve the problem? –Could you, if given a program and its input, tell me whether that program would halt or not? –What would cause a program to not terminate? Infinite loop Infinite recursion Run-time error –A good programmer could probably look at a program and see if any of these occur Lets consider some examples that deal with infinite loops

7 Example 1 Will the following C++ loop terminate? cin >> x; while(x>0) do x--;

8 Example 2 A slight variation of the previous loop is given below Will it terminate? cin >> x; while(x>0) do cout << x;

9 Example 3 What about loops where we cannot step through the code because they involve input? Will this loop terminate? cin >> x; while(x>0) do cin >> x;

10 Using Intelligence To solve the previous problems, we applied our intelligence –our knowledge of computer programming –our knowledge of the Java language –our experience in dealing with infinite loops A computer program does not have intelligence, but could we write a program that solves the problem in the same (or a similar) way as to how we solved it? –if so, we will have applied intelligence to solve an otherwise thought of unsolvable problem!

11 Solvability vs. Intelligence From the previous example, we see that people can potentially solve unsolvable problems –or can they? we expect a solvable problem to have an algorithm that is always correct is a human always correct? When humans solve problems, there is some component of fuzziness or doubt –humans tend to use approximate reasoning strategies and incomplete knowledge –whereas we expect computer programs to have rigidly structured programs and a consistent and complete data base

12 So, what is intelligence anyway? As pointed out earlier, there is no single or adequate definition for intelligence How can AI succeed if we cannot even clearly define the goal? Instead of defining intelligence, we now turn to a possible test for intelligence, the Turing Test

13 The Turing Test 1950 – Alan Turing devised a test for intelligence called the Imitation Game –Ask questions of two entities, receive answers from both –If you can’t tell which of the entities is human and which is a computer program, then you are fooled and we should therefore consider the computer to be intelligent Questions Answers Which is the person? Which is the computer?

14 The Chinese Room Problem From John Searle, Philosopher, in an attempt to demonstrate that computers cannot be intelligent –The room consists of you, a book, a storage area (optional), and a mechanism for moving information to and from the room to the outside Question (Chinese) Book of Chinese Symbols Answer (Chinese) Storage You

15 Does This Pass the Turing Test? To the Chinese speaker, he/she wrote a question in Chinese and received a reasonable response and walks away satisfied – the entity in the room understands Chinese and intelligently responds –But do you understand Chinese? –Did you apply your intelligence?

16 Chinese Room: An Analogy for a Computer User Input I/O pathway (bus) Output Memory CPU Program/Data

17 Searle’s Question You were able to solve the problem of communicating with the person/user But did you understand the Chinese messages being communicated? –While we can say that the you passed the Turing Test (you exhibited intelligent behavior) we cannot say that you understood what was exhibited –Since you do not speak Chinese, you did not understand the input, output, data or the program Can we say that you actually used any intelligence?

18 By Analogy Then Since you did not understand the symbols that you interacted with, we can say that the computer does not understand the symbols that it interacts with –The computer does not understand what it does, it just does it Therefore, The computer is not intelligent, it has no “semantics,” –We often call the computer a symbol manipulating device

19 Brain vs. Computer In AI, we compare the brain (or the mind) and the computer –Our hope: the brain is a form of computer –Our goal: we can create computer intelligence through programming just as people become intelligent through learning But we see that the computer is not like the brain The computer performs tasks without understanding what its doing Does the brain understand what its doing when it solves problems?

20 Strong vs. Weak AI We differentiate here two fields of study –Strong AI – the pursuit of machine intelligence –Weak AI – the pursuit of machines solving problems in an intelligent way Weak AI makes no commitment towards whether the machine will actually understand or not Even if we believe that strong AI is not possible, we can still pursue weak AI, it seems possible

21 AI Areas of Research Knowledge-based and Expert Systems Natural Language Processing Systems (including machine translation) Perceptual Systems (including visual and speech recognition) Intelligent Agents Automated Problem Solving Systems Intelligent Databases Learning Systems Robotics, Smart Weapons, etc

22 Recent “hot” AI Topics Machine Learning –Neural networks –Data mining and statistical approaches –Learning by example (explanation based learning) –Genetic algorithms (evolving better answers) Intelligent Agents –Small problem solvers that can work on their own but also know who to contact for assistance –Examples: travel agent softbots, search engines Together, these areas are sometimes called Artificial Life (A-Life)

23 Comments? Questions?


Download ppt "How Solvable Is Intelligence? A brief introduction to AI Dr. Richard Fox Department of Computer Science Northern Kentucky University."

Similar presentations


Ads by Google