3. Common Sense
A program has common sense if it
automatically deduces for itself a
sufficiently wide class of immediate
consequences of anything it is told
and what it already knows
4. John
McCarthy
• 1955: Develop the phrase
“Artificial Intelligence”
• 1959: Programs with
Common Sense
• 1960: First LISP
implementation
• 1971: Turing Award Recipient
5. Intelligent System
Core Features:
1. Representation of all the behaviors.
2. Interesting changes in behavior must be
expressible in a simple way.
3. All the aspects of behavior must be improvable.
4. Need of partial success concepts.
5. Need to have subroutines which can be
included in procedures as units.
6. Intelligent System
Core Features:
1. Representation of all the behaviors.
2. Interesting changes in behavior must be
expressible in a simple way.
3. Improving mechanism should be improvable.
4. Need of partial success concepts.
5. The system must be able to create subroutines which can
be included in procedures as units.
7. Advice Taker
• Is a proposed program for solving problems by
manipulating sentences in formal languages.
• The basic program will draw immediate
conclusions from a list of premises.
• These conclusions will be either Declarative or
Imperative sentences.
9. What is NOT Advice
Taker?
• Heuristics are NOT embodied in the
program
• Procedures and heuristics are described
in the language itself
10. Advice taker
• An example:
• Assume that I am seated at my desk at home
and I wish to go to the airport.
• My car is at my home also.
• The solution of the problem is to walk to the car
and drive the car to the airport.
11. Some Premises we can
make..
A predicate “at”
“at(x,y)” is a formalization of “x is at y”
14. Ambiguities
• Semantics of the language
Meaning of “at” is controversial
• Can have multiple syntactically correct but
semantically incorrect inferences towards the same
goal.
16. Time complexity
• Even an inference algorithm is meant to be
sound and complete this may take an
exponential time even in well defined finite
state environments.
17. Can we match a logic
agent to a human
• We can’t make a generalized
inference in a constant time in which
most of the human beings are
capable of
18. Thank You!
• Dileepa Fernando
• Rashmika Nawaratne
20. The Present…
• The symposium on common sense reasoning
held at The Courant Institute – 2001
1. Better sense of how to develop a workable
methodology for formalizing common sense
2. How to divide the larger problem up into more
manageable parts
21. Why this subject is
difficult?
Simple tasks use implicit assumptions which cannot be
determined without proper (deep) study about
human mind.
• Real world Example:
• Research paper:
Editor's Notes
Expensive: Hard to ImplementGood afternoon all,The main idea of today’s presentation is to describe the expensiveness of a logical agentSo before going to the advance details we’ll simply have a look at what a logical agent is
Logical agents is the kind of an entity thatapply inference to a knowledge base to derive new information and make decisions our context of logical agents we mainly concern about:Knowledge about the environment which percepts are comingAbility of inference the actions happening in the environmentAbility to arrive at specific conclusions through the inference mechanism
And another thing which we need to consider is the common sense.Beinga human, what do you think the common sense means?Basically it’s the judgments we take when we encounter percepts or facts in the nature.When it comes to programs or machines, we can define common sense as given in the slide.Basically what meant by the definition is a program which has common sense, can automatically deduce all the consequences in its domain, by the percepts which it acquire from the environment and by the things which it already knows about the domain.
Today the paper we are presenting you is Programs with Common SenseWhich was a research of John McCarthyand was published in 1959Introduced the phrase Artificial Intelligence in 1956Famous for researches which are based on Intelligent SystemsIn the slide you can see some of his contributions to the field of Artificial IntelligenceProceedings of the Teddington Conference on the Mechanization of Thought ProcessesPaperdiscuss:Programs to manipulate in a suitable formal language common instrumental statements Program will draw immediate conclusions from a list of premises
According to John McCarthy’s research, it defines 5 main features that a ideal intelligent system should have;All behaviors (Of the environment) must be representable in the system. Interesting changes in behavior (Environment) must be expressible in a simple way. All aspects of behavior must be improvable. In furthermore explaining, the improving mechanism should also be improvable.4. The IS must have concepts of partial success because on difficult problems decisive successes or failures come too infrequently. Becausethere are natural problems which do not have a optimal solutions under the given conditions.5. The mechanism that selects subroutines should have concepts of interesting or powerful subroutine whose application may be good under suitable conditions. (In a simple sense what this means is it should have knowledge to choose the correct algorithm according to the conditions)
In order for a program to be capable of learning something it must first be capable of being told it. Therefore definitely there should be a simpler way to capture behaviors within the intelligent system.The simpler way which is as described in the paper is using sentences to represent behaviors in the system.The solution which is presented in the paper is the Advice Taker.
Is the proposed program by John McCarthy for solving problems by manipulating sentences in formal languages. The expectation of advice taker is improve its behavior merely by making statements to itPremises: A sentence in the knowledge base which can be used to infer.Ultimate objective of Advice Taker is to make an IS that learn from itsexperience as effectively as humans do.
Imperative: How to do the itDeclarative: What to doWhen an imperative sentence is deduced the program takes a corresponding action. In the latter part of the presentation we’ll explain how to process these sentences in the Advice TakerAdvantages of Imperative Sentences: A procedure of imperatives is carried out faster Know previous knowledge of the machine needs to be analyzedAdvantages of Declarative Sentences: Use of previous knowledge is acceptable Order is not as important as in imperative sentences, can have afterthoughts Previous state is less important so allows for less need of instructor to know previous state
Now let us see what is the difference between the advice taker and previous programs made earlier.The main difference between it and other programs or proposed programs for manipulating formal is that in the previous programs the formal system was the subject matter but the heuristics were all embodied in the program. In this program the procedures will be described as much as possible in the language itself and, in particular, the heuristics are also described.Now we’ll consider the example of the behavior of the advice taker…
Paper explains the advice taker’s behavior with a general example. Note that this is very high level. Hence most of the implementation detail are not taken care of.
First of all we have to build a knowledge base with the information we have. The following are the basic premises which describe the scenario. Though these definitions are conflictual at certain levels We will continue with this notation for the time being.
For human, inference rules are obvious. Transitivity and the implications stand at the common sense levelBut, for this machine we have to embed these rules as well.
This is the stage in which we infer certain conclusions. As human it is a trivial task for us to substitute “home” for X.For an agent we have to specify a way of doing the correct substitution.This shows how we arrive at conclusion by deduction. When some premise is known to be true, all the statements implied by that will also be true.
Semantics are very important. For instance, at(I,desk) means I stay near by the desk. at(I,home) implies I am inside home. When using these two definitions for “at” we can’t always make sure that transitivity would work.at(I, desk A) and at(desk A, desk B) does not imply at(I, desk B)
The inference technique stated before may sound interesting. But the problem is that agent cannot decide what premises to use first for deduction, in constant time (most of the cases) This search may be done recursively. At the worst case time complexity will have an exponential order.Implementing this solution efficiently is beyond the scope of the paper.
Human can make most inferences in constant time which is magical. With some magic, a human being is able to pop up some past decisions from memory and optimize the search. Fetching the data from the memory is done within a constant time which is impossible with an agent in general.
Anything like a formalization of common sense is so far from being accomplished that—if it is achievable at all—it is not even possible to estimate when we could expect the task to be completed.2001: This effort is yielding a better sense of how to develop a workable methodology for formalizing common sense, and of how to divide the larger problem up into more manageable parts.seeks to account for many areas of knowledge attempts to see how this formalized knowledge can be brought to bear on moderately complex common-sense reasoning problems.
The motivation for using logic is that even if the eventual implementations do not directly and simply use logical reasoning techniques like theorem proving a logical formalization helps us to understand the reasoning problem itself. The claim is that without an understanding of what the reasoning problems are, it will not be possible to implement their solutions.