Thursday Jan 10 2013 - Lecture 2

7 Pages
Unlock Document

Computing and Information Science
CIS 3700
Yang Xiang

Thursday, January 10, 2013 CIS 3700 Lecture 2 Review • Definition of artificial intelligence • Raised the issue of what constitutes success when creating intelligent entities • Human-centered approach included acting like a human • We can judge success based on how the agent acts or thinks • Turing test has the limitation of using the human as the golden standard because it would force machines to mimic human weakness ◦ Machines have different strengths than humans, so there are skills that they can do better than a person ◦ Criticism against the 'acting like a human' approach Principled Approach • Thinking logically ◦ Build intelligent systems based on logic ◦ Not always practical to solve problems logically ◦ This approach dominated during the 1960's ◦ Argument against this approach (That intelligent systems must be built based on mathematical logic) ▪ Though logic is a powerful tool and can be used to describe many scenarios, it is limitted in practical problems that cannot be solved purely based on mathmatical logic ▪ We will not be exploring this topic in this course (saved for advanced courses inAI) ▪ Covered in the textbook in section 13.1 if interested (but it will not be covered in the tests for this course) ▪ We will use mathematical logic for much of our scenarios • Acting Rationally ◦ Arational agent acts so as to achieve the best outcome ▪ Based on the external behaviour of the agent ▪ Presumably, the agent is working as a principle of a real user ▪ Aprinciple is a client that uses the agent to accomplish a specific task (Double check) ▪ The agent has to be able to create a decision in a reasonable amount of time and within specified resource constraints ▪ If the agent can consistently achieve the best possible outcome, it is deemed successful ◦ Not necessarily imitate human, either internally or externally ▪ As long as the behaviour is rational, it is intelligent, regardless of whether or not the behaviour is similar to what a typical human being would do ▪ For some tasks, human behaviour is advantageous, and should be imitated ▪ If human behaviour is rational, imitate it, but if human behaviour in this situation would not be rational, do not imitate that behaviour ◦ Not necessarily based on logic ▪ If a problem can be solved based on logic, then use it, but if logic is not appropriate, then use another approach ▪ This approach is more general than the human-centered approach ◦ What is rationality? ▪ Doing the right thing given what is known ◦ We will adopt this approach in our course for determining whether agents are inteligent Agents and the Environment • Operationally, an agent perceives its environment through sensors and acts upon the environment through actuators ◦ Top box represents the agent and bottom box represents the environment ◦ The agent is composed of sensors, actuators and the agent's program ◦ Sensors are intended to be used to observe the environment ◦ Actuators are intended to be used to make changes in the environment • Environment in this context refers to the situational setup of a particular practical application (eg. Navigating a building, diagnosing a patient, tutoring algebra). It does not refer to the physical context of the agent. ◦ Example: Consider a robot that is navigating a building to deliver packages from one room to another ▪ Environment is the building • Building has rooms and hallways that are laid out in certain configurations • Includes the current physical location of the robot and the destination of the package ▪ Sensors • Camera • Infrared sensors • Allows the agent to see surroundings and sense distance to nearby objects • Allows identification of objects • With sensors the agent decides where it currently is ▪ Program • Accesses knowledge of the building (ie. Map) • Computes a path to the destination of the package ▪ Actuators • Motor • Wheels • Arms to push doors if required • Any mechanism that allows the robot to move and perform its function of delivering packages ▪ Large integrated loop of environment, sensors, program and actuators ◦ Example: Medical diagnostic agent ▪ Environment: • Patients • Biological signs and symptoms • Medical tests ▪ Sensors: • Thermometer (Get body temperature) • X-Ray instrument (Take images) • Other medical instruments that collect symptoms ▪ Program: • Accesses knowledge about the relationship between symptoms and diseases • Makes decisions about what disease or malady the patient is likely to have • Determine improvement of returning patients and make judgements based on attempted treatments ▪ Actuators: • Printing prescription • Limited surgical capabilities • Output mechanism for communicating the result ◦ This model is different from other programs we have written in the past ◦ Above is the regular was that programs are written and run, but this model is not what is used when programing intelligent agents • EarlyAI was focused on building something like a newborn baby that has the ability to learn to do anything ◦ What kinds of mental capacities do a newborn baby have that allows them to learn so much and accomplish so much ◦ This type of thinking is impractical ◦ This type of thinking isolates the agent from the environment ◦ It is reasonable that this type of agent would be more difficult to build ▪ There is not super-human that does everything at the expert level so it is impractical to think that we could create a machine to do • Situatedness of agent ◦ We began considering the situatedness of the agent when we realized that we could not build super agents ◦ Makes coding more practical Agent Function • Percept sequence: everything that the agent has perceived so far ◦ After the agent is booted ◦ Every time the agent perceives the environment is a percept sequence • Agent behaviour can be described by agent function that maps
More Less

Related notes for CIS 3700

Log In


Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.