Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
Do you really want to delete this prezi?
Neither you, nor the coeditors you shared it with will be able to recover it again.
Make your likes visible on Facebook?
You can change this under Settings & Account at any time.
Transcript of Artificial Intelligence
What is artificial intelligence?
Artificial intelligence (AI) refers to the use of computer science to develop systems that are able to sense their environment and base its actions on that input. In short, they are able to function on their own.
There are many different divisions of artificial intelligence that are all equally important.
These divisions include: how AI shows knowledge, learning, planning, how it processes human language, manipulation, perception, social intelligence, creativity, and general intelligence.
Knowledge representation is how the AI system distinguishes bits of information within the knowledge domain.
Logic plays a large role in what symbols are used to represent information within the system. Also, a knowledge domain depends on how expressive it is.
What "expressivity" means is the more expressive a knowledge system is, the easier it is to identify which elements are which. This also makes the system more complicated in terms of logarithms.
If you have a system of representation that is very simple in terms of expressiveness, that doesn't mean that the system is bad; it just means that it's more likely to be complete.
Intelligent behavior is a result of good knowledge representation; after all, what good is the information unless you can make sense of it?
This intelligent behavior should include reasoning, drawing conclusions, and correct deductions. It should also include information that is declarative (facts known to be true) and procedural (how the knowledge representation knows to perform tasks).
A good way to tell the two apart is this:
Declarative knowledge lets you know what you need to do, and all the information you need to do it.
Procedural knowledge lets you know how you go about doing a task.
McCarthy, John. "What is Artificial Intelligence?." Stanford.edu. N.p., 17 Nov. 2007. Web. 16 Sept. 2013. <http://www-formal.stanford.edu/jmc/whatisai/>.
"Artificial Intelligence." Wikipedia. N.p., n.d. Web. 16 Sept. 2013. <http://en.wikipedia.org/wiki/Artificial_intelligence>.
"Knowledge representation and reasoning." Wikipedia. N.p., n.d. Web. 16 Sept. 2013. <http://en.wikipedia.org/wiki/Knowledge_representation_and_reasoning>.
Uluoglu, Belkis. "Declarative / Procedural Knowledge." Designophy. N.p., n.d. Web. 16 Sept. 2013. <http://www.designophy.com/designpedia/design-term-1000000001-declarative-.-procedural-knowledge.htm >.
"What Is Procedural Knowledge?." Wisegeek. N.p., n.d. Web. 16 Sept. 2013. <http://www.wisegeek.com/what-is-procedural-knowledge.htm#didyouknowout>.
A good knowledge representation system will:
Have a large, complete amount of information
Have the ability to be read by humans (natural language)
Be easy to update and edit
Have intelligent activity
Planning in artificial intelligence refers to how automated systems execute action sequences, such as how autonomous robots operate.
"Automated planning and scheduling." Wikipedia. N.p., n.d. Web. 16 Sept. 2013. <http://en.wikipedia.org/wiki/Automated_planning_and_scheduling>.
Planning is not a simple thing; consideration for multiple factors must be taken. Planning often involves a lot of trial and error.
The Intelligence agent of the artificial intelligence system needs to be able to plan ahead for the goals it wants to accomplish and how it will accomplish it.
Learning, or machine learning, refers to how artificial intelligence can learn to differentiate data.
The AI can learn to represent and generalize data; generalization is just how the AI is able to operate after it has recognized new data.
There are a few different algorithm types associated with machine learning. Simply put, algorithms are planned out diagrams of how decisions should be made.
The types of algorithms involved with machine learning are:
Supervised learning algorithms operate where both of the input and the output is known; you would use this when you want to map out possibly unknown values.
Unsupervised learning deals with where you know the input but not the output. You would use this when you want to gather data about outputs.
Semi-supervised learning is a combination of supervised and unsupervised learning.
Transduction is used to predict outcomes for tests based on what it has learned.
Reinforcement learning is how the AI interacts with the environment; it learns how to get the most out of its actions in the environment.
Natural Language Processing
Natural language processing is how the AI is able to communicate with humans. Specifically, it's the interaction of machine language and human language.
The AI has to be able to understand the complexities of human languages. Some words are used as verbs, nouns, and adjectives; its meaning changes with the context.
The AI software needs to be programmed to be able to make sense of the input.
Natural Language Processing, or NLP for short, can perform many tasks such as these:
Machine translation is used to translate from machine language to a selected human language. This is one of the harder tasks to accomplish due to the many different things that need to be programmed.
Auto (or automatic) summarization is used to look over a document of text and present the most important information.
Natural language understanding is used to convert human language into commands for the AI to follow.
Part of speech tagging is used to "tag" words according to what part of speech they are.
Question answering is just that; a question is asked and an answer will be provided.
Word segmentation is mainly used in languages such as Chinese and Japanese; this process separates characters into their respective words because of the lack of spacing.
Parsing involves separating a sentence using proper grammar rules. When using AI for this, a sentence could possibly have multiple meanings because of the AI's extensive ability to diagram the sentence.
Optical Character Recognition is when the AI can recognize text from an image shown to it.
Motion and Manipulation (Robotics)
The motion and manipulation part of AI is directly related to robotics. Robotics uses artificial intelligence for many things.
Robotics uses AI for navigation, mapping, motion planning, and localization.
Mapping is figuring out what is in the environment and being able to know where each object is. Essentially, it's building a map.
Motion planning is working out what path you need to take in order to navigate the environment successfully.
"Machine Learning." Wikipedia. N.p., n.d. Web. 18 Sept. 2013. <http://en.wikipedia.org/wiki/Machine_learning >.
"Algorithm." Wikipedia. N.p., n.d. Web. 18 Sept. 2013. <http://en.wikipedia.org/wiki/Algorithm >.
"Natural Language Processing." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/Natural_language_processing >.
"Reinforcement Learning." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/Reinforcement_learning >.
Perception includes both machine and computer perception. They both have their own distinctions.
Machine perception is when the AI receives from sensors, such as motion sensors, touch sensors, microphones, etc. Computer perception is when the AI receives visual input and turns it into machine legible code used for making decisions.
Computer perception can be used for facial, object, and voice recognition.
Social intelligence refers to how artificial intelligence is used to make robots that are able to imitate humans.
The robot should be able to sense emotional changes in humans and correctly respond to the emotions shown by humans.
Social intelligence is a mix of psychology, computer science, and cognitive science (the science of how the mind works and why it works that way).
Also, a robot equipped with social intelligence should be able to predict what emotions will be elicited when another emotion is displayed.
In other words, it has to plan out what emotions will have a certain reaction.
Another part of social intelligence is being able to display emotions in addition to perceiving them.
Creativity, when dealing with AI, means constructing an Artificial Intelligence system that can possess human creativity, understand how creativity works, and work on making things more creative.
This creativity is typically identified by the criteria of: the idea being different from others, rejecting previous ideas, defining unclear ideas, and the idea is formed through hard work.
General intelligence, also called Strong AI, refers to how artificial intelligence might be joined one day to form an artificial intelligence system that will be able to surpass human intelligence.
Another associated concept is AI-complete. AI-complete means that a task or algorithm is not able to be solved simply and that it matches human difficulty.
This requires the AI to be able to apply reason and multiple other aspects, such as machine translation, a knowledge representation, and creativity.
Comparing and Contrasting Artificial Intelligence in Robotics
To the left is Kismet, a robot constructed for interacting with humans. It can take input through audio and visual sensors and can also understand body language. It expresses emotions through the eyes, eyebrows, lips, head, jaw, and ears.
Different types of artificial intelligence have their own role in robotics.
The range of abilities for robots using artificial intelligence is very wide. Complicated artificial intelligence equipped robots can go to work for the military, while vacuum robots can learn the layout of a room to avoid bumping and nicking furniture.
Another example is voice-recognition being used. Apple's Siri, for example, can recognize questions and commands fairly well.
Military robots can be programmed to sense explosives. The robot learns from this input and knows how to respond accordingly, such as setting off an indicator or proceeding with a set of actions.
Both voice recognition and explosive sensors are similar in that they both take input from their environment and perform actions from that input.
They differ in the fact that voice-recognition obviously cannot be used to sense physical data and explosive sensors cannot detect audible data.
Robots that use artificial intelligence to navigate can also differ from each other.
A toy for children to use will be less complicated than a robot that is large in scale and needs to move through rough terrain.
On a basic level, both objects are able to do the same thing, just to very different degrees.
For example, WowWee toys sells toy robots that use artificial intelligence in different ways.
The Roboraptor is designed to respond to input from its controller, but it also an infrared (IR) sensor. This means that it can detect when objects are nearing it if it's walking around autonomously.
It can also react with a "mood" when certain factors in the environment provoke it. The Roboraptor will also interact with other WowWee robot toys.
Decision logic in robotics is simply how the robot makes decisions on what to do based on the logic programmed into it.
An example of decision logic would be how Boolean logic operates. Boolean logic sees if an input (such as a sensor) reads as true or false. The robot then performs the action for what each indicator says.
For a robot that has to avoid walking into objects, a true reading might be that an ultrasonic sensor detects an object within a specified distance.
A false reading would say that the sensor does not detect any objects within a certain distance.
You could program the robot to continue walking if the output is false, and to stop if the output is true.
You can get very complicated with decision logic, especially if you have a robot equipped with multiple sensors. This example is very basic.
Boolean logic uses "gates" to process data. They are: NOT Gate, AND Gate, OR Gate, NOR Gate, NAND Gate, XOR Gate, and the XNOR Gate.
The NOT Gate is also called the Inverter Gate because it takes in the input and outputs the opposite. If a zero is the input, the output is one, and vice versa.
The input is referred to as "A" and the output is "Q."
The AND Gate basically works so that if the inputs A and B (the input) are 1, for example, then Q (the output) should be 1.
The output will only be 1 if both A and B are 1, as illustrated below.
The OR Gate is used where A and B are inputs and Q is the output; for example, if A OR B equal 1, then Q is 1.
The NOR and NAND Gates are very similar to each other. They are made by combining either an AND or OR Gate and a NOT Gate.
You can see below how these Gates function.
Another two gates that are used are the XOR and XNOR Gates. The X stands for "exclusive," as in the XOR Gate, if A OR B is 1, but not both of them, then Q is 1.
Boolean logic, when applied to robotics, usually works in a true or false format.
Comparisons, such as greater than, less than, greater than or equal to, less than or equal to, and equal to are also used in Boolean logic.
A simple example of using Boolean logic would be this:
Let's say that you have a robot and you want it to move backwards only when the light sensor sees black and when it is moving.
You need to factor in both the light sensor and the moving wheels. The robot will only move backwards when both values are true.
You would use the AND case for this since you need both values to be true, not just one.
light sensor < 10 & wheel motors = 1
8 < 10 & 1 = 1
true & true
Note that for the motors, 1 equals true and 0 equals false.
Truth tables are tables that use math to work out how logic finds all possible outcomes.
We have used truth tables earlier in the presentation to explain the different cases of Boolean logic.
For example, if you have X and Y as the input, and you want to find an instance in which X and Y are both true, you would use an AND table. This can be written as or x AND y.
x AND y
x y x AND y
The logical expression for that truth table would be:
When X and Y both equal 1, that is the only time when the case of x AND y can be true. All other input combinations result in a false value.
The OR case is represented as x OR y or as .
A truth table for this expression would be this:
x OR y
x OR y
The logical expression for this truth table would be:
X or y is true when x or y is equal to one, or when both x and y are one. The only false instance is when neither x nor y equal one.
The NOT case is just a negation of the input value, so its truth table is very simple.
It is represented by either "NOT x," "!x," or .
The logical expression for this would be:
When x is equal to zero, the output is one, and when x is equal to one, the output is zero.
Venn Diagrams for Boolean Algebra Problems
Venn diagrams illustrate well how the AND, OR, and NOT cases work.
A true result is only achieved when both x and y share the desired output.
The OR Case is true when either x or y equal the desired output, or when x and y have the same value for the desired output.
The NOT Case is true when the output is the opposite of the input.
"NLP." Webopedia. N.p., n.d. Web. 19 Sept. 2013. <http://www.webopedia.com/TERM/N/NLP.html >.
"Parsing." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/Parsing >.
"Computer vision." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/Computer_vision>.
"Machine perception." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/Machine_perception >.
"Cognitive science." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/Cognitive_science>.
"Kismet (robot)." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/Kismet_(robot) >.
"File:Kismet robot 20051016.jpg." Wikipedia. N.p., n.d. Web. 19 Sept. 2013. <http://en.wikipedia.org/wiki/File:Kismet_robot_20051016.jpg>.
"Computational creativity." Wikipedia. N.p., n.d. Web. 20 Sept. 2013. <http://en.wikipedia.org/wiki/Computational_creativity >.
"Strong AI." Wikipedia. N.p., n.d. Web. 21 Sept. 2013. <http://en.wikipedia.org/wiki/Strong_AI >.
"AI-complete." Wikipedia. N.p., n.d. Web. 21 Sept. 2013. <http://en.wikipedia.org/wiki/AI-complete>.
Harris, Tom. "AI-complete." Howstuffworks. N.p., n.d. Web. 21 Sept. 2013. <http://science.howstuffworks.com/robot6.htm>.
"What Are the Different Types of Artificial Intelligence Technology?." Wisegeek. N.p., n.d. Web. 22 Sept. 2013. <http://www.wisegeek.com/what-are-the-different-types-of-artificial-intelligence-technology.htm>.
"What Are the Applications of Artificial Intelligence?." Wisegeek. N.p., n.d. Web. 22 Sept. 2013. <http://www.wisegeek.com/what-are-the-applications-of-artificial-intelligence.htm>.
"iRobot Roomba® 790." iRobot. N.p., n.d. Web. 22 Sept. 2013. <http://store.irobot.com/product/index.jsp?productId=12991591&cp=2501652&jsessionid=nTtpS1dc6hldlfB3p58BN9T0GCWgQfCj5X5gmG6nhh3KCKBXYd1k%21-1669999542&s=A-ProductAge&parentPage=family>.
"What Is the Connection Between Artificial Intelligence and Robotics?." Wisegeek. N.p., n.d. Web. 22 Sept. 2013. <http://www.wisegeek.com/what-is-the-connection-between-artificial-intelligence-and-robotics.htm>.
"WowWee Robots." WowWee Toys. N.p., n.d. Web. 22 Sept. 2013. <http://store.wowwee.com/robots/>.
"Roboraptor." WowWee Toys. N.p., n.d. Web. 22 Sept. 2013. <http://store.wowwee.com/robots/roboraptor.html>.
Brain, Marshall. "How Boolean Logic Works." Howstuffworks. N.p., n.d. Web. 23 Sept. 2013. <http://www.howstuffworks.com/boolean.htm>.
Brain, Marshall. "How Boolean Logic Works - Simple Gates." Howstuffworks. N.p., n.d. Web. 23 Sept. 2013. <http://computer.howstuffworks.com/boolean1.htm>.
"Boolean Logic." Robotc.net. N.p., n.d. Web. 23 Sept. 2013. <http://www.robotc.net/NaturalLanguage/pdf/hp_boolean_logic.pdf>.
"Truth table." Wikipedia. N.p., n.d. Web. 23 Sept. 2013. <http://en.wikipedia.org/wiki/Truth_table>.
"Boolean algebra (logic)." Wikipedia. N.p., n.d. Web. 23 Sept. 2013. <http://en.wikipedia.org/wiki/Boolean_algebra_(logic)>.
Truth tables such as this can be represented with T and F or 1 and 0.