FAQSearchEmail

humanlevelartificialintelligence.com   

  
 chapter10

Home | Videos | Contact Us   

 
Home
HLAI
UAI
Videos
Books
Patents
Notes
Donation

     
 

                      

                         << content                                      Chapter 10

4.  Giving speeches and having conversations

Giving a speech inherits most of the intelligent pathways for writing a book.  Intelligent pathways in memory are structured hierarchically.  Writing sentences on paper is similar to speaking sentences to a public.  They both share commonalities. 

FIG. 53 is a diagram depicting the hierarchical tree that shows the relationships between writing a book and giving a speech.  The common intelligent pathway used is translating thought to sentences (FIG. 54).  The intelligent pathway to write a book is used to write a book and the intelligent pathway to give a speech is used to give a speech.  Both intelligent pathways inherit the root node, which is translating thoughts into sentences.  For each task, there are slight differences for T4 and the robot’s conscious have different rules that it must follow.     

FIG. 54  

 

FIG. 55

 

In intelligent pathway T4, the last instruction will point to a different pathway to do things according to the task.  If the robot was giving a speech, the pathway will be to open its mouth to speak the remembered sentence.  If the robot was writing a book, the pathway will be to move his hand and write the remembered sentence on a paper. 

Tasks and rules are very important to the robot when giving a speech.  The robot has to know the people in the audience.  A speech for an audience comprising children will be different from a speech for an audience comprising politicians.  With the children audience, the robot will have rules that it must follow, such as:  “use elementary words”, “don’t use bad language”, “speak slowly”, “make the audience laugh” and so forth.  The job of the intelligent pathways changes as the robot understands the current situation.  The intelligent pathways modify the tasks and rules and knowledge of the computer program inside the robot’s conscious.

This is why the robot has to learn speeches in a variety of environments.  He should learn to give a speech in front of politicians, school teachers, the public, in a classroom, in a family gathering, in a concert and so forth.  At least, learn the knowledge of how to give a speech under various audiences.  College classes on speeches are a good place to learn how to give a speech.  Thus, the robot will be able to give speeches under any situation.  He can give a speech in a news conference, or in a family party or in a debate class.          

In the hierarchical tree, under give speech, the robot will be able to select the appropriate intelligent pathways for that given situation.  All speeches will comprise very similar general instructions.  The specific intelligent pathways will have instructions to do specific types of speeches.    

 

5.  Writing computer software programs

The writing of computer software is very similar to writing a book.  Although, the rules of grammar and the rules of software engineering are different, they both have rules that you have to follow according to specific tasks.

The robot has to have knowledge about object-oriented programming, which includes discrete math functions like if-then statements, for-loops, and statements, or statements and so forth.  Learning classes, methods and recursions are also a must to write meaningful software.

Once the knowledge and the rules of software engineering are stored in the robot’s brain, he has to know the steps to write software.  These steps can be read from basic programming books or learned through trial and error.  When the robot takes computer science classes, the professor will give assignments to the robot, ranging from basic to complex.  A simple assignment like writing a java applet that will output the text:  “hello word”, might be the first assignment.  As the robot learns more about computer science and actively write software, he will get better and better.  By the time he graduates from college, the robot will have the knowledge to write complex software.   

By the time, the robot gets into college, his brain already has the capacity to learn anything.  Learning how to write software or learning how to build bridges or learning how to draw the blueprints to a city is possible.

It might of taken the robot 20 years to develop the skills to write a book.  However, when it gets to college, it will take the robot only several months to learn how to write a decent software program.  The reason why is because the intelligent pathways are so well developed that it can teach itself how to learn.  The way the robot store and retrieve information has been optimized.

 

FIG. 55 is a diagram showing what knowledge the robot should know about computer science.  These various knowledge will be needed to translate an outline of a problem into a software program.    

The knowledge of object oriented programming will be configured in memory according to how knowledge in science books are configured (in previous chapters).  Teacher lectures of chalkboard demonstrations will help to organize the data in memory.  Also, visual diagrams in the computer text books will help to organize the data in memory.  When the teacher draws diagrams on the chalkboard and finger points to numbers and shapes, he is providing a very strong method for the robot’s brain to configure data in an organized way.

FIG. 56 

Referring to FIG. 56, classes and functions can be represented by a square diagram with inner shapes representing certain elemental parts.  For example, the square at the top is the declaration, where the objects and functions are defined.  The circle at the bottom is the main body and that is where objects are used and functions are called.

In the second diagram is a visual diagram of recursion.  The rootnode is the parent, which is made up of a class and the child node is an inner copy of the parent.  Other data and facts about binary trees can be included in this diagram.  During the learning phase, the teacher is drawing this diagram and he is pointing here and there and writing down things on the diagram.  The robot will store static as well as linear data from the lecture and configure data from the lecture in an organized manner for memory storage.

As far as accessing data from the knowledge learned about object oriented programming, the data is stored in memory as 5 sense data (either as static data or linear data or both).  The intelligent pathways to analyze is the key to finding out how things work.  For example, before writing a software program, the robot will activate the class diagram and analyze the diagram.  Information extracted from the intelligent pathway, via search patterns, will determine the fruit of the search.  The robot might think:  “oh, that is where the declaration goes or that is where the body of the problem should be or that is where the functions are called”.  Another example is the human hand.  If the teacher asks this question:  “what does a human hand look like”, the robot will extract a visual image of the human hand and using intelligent pathways in memory to output what the visual hand looks like.  There is no data in memory that explains the detailed aspects of a human hand.  The intelligent pathways analyze the image to extract the information of what the human hand look like, during runtime.   

FIG. 57 is a diagram depicting the steps the robot will take to write software.  The robot will be given a problem by the professor.  He will take the problem and write an outline for a software program that will solve the problem.  Next, the robot will translate the outline into software functions.  Finally, he will test out the software program so that it is working correctly; and solves the problem given by the professor. 

 FIG. 57

The outline of the software program is very important.  At this stage, the robot needs to know how the software program will work and use computer science knowledge to translate the outline into computer functions.  For example, if the robot is writing a computer program on a binary tree, he has to use recursive classes.  The outline of the binary tree will probably be a diagram of a hierarchical tree.  If the computer software is a chat network, the outline might be multiple interconnected nodes with a hub in the middle. 

For me personally, I would draw out the outline of the software program I am planning to write.  I do this by drawing circles, squares and connected lines.  I also write down the functions of each node and what data is contained in each node.   

Referring to FIG. 58, when the robot decides to write a software program, knowledge will start to pour into its conscious.  At first, a general computer program will be used to manage tasks and rules.  As the robot follows intelligent pathways in memory to write software, the computer program inside the robot’s conscious will get more specific.  It can write any software program regardless of how complex it might be. 

The intelligent pathways contain the step-by-step instructions to manage multiple layered tasks and to be aware of possible rules that the robot needs in order to accomplish tasks in the task container.  Task interruptions will be solved by the robot’s conscious as well.  There are limits to how certain computer statements work.  The robot has to be aware of these limits.  There might be number limits to a for-loop because the program might run out of disk space.  Or certain variables can only use certain data types.  The robot has to know these rules when writing the software.  

 FIG. 58

Learning how to write software and understanding the knowledge of writing software is one thing, actually doing the assignments is another thing.  The robot has to write software starting from simple software programs to complex software programs.  The robot’s brain learns knowledge through trial and error and the pathways have to go through a bootstrapping process, whereby intelligent pathways build on itself. 

 

Here is a list of simple to complex software programs:

 

1.  write a program to output “hello world”.

2.  write a program to convert Fahrenheit to Celsius.

3.  write a database program to store customers using a binary tree.

4.  work in a group to write software to an operating system. 

 

When the robot learns to write a simple program like program1, he can use that knowledge to write program2.  In program2 you have to know how to output string text to the monitor so that users can see the number conversions.  Program2’s knowledge can be used in program3 because program2 has the ability to manage classes and calling functions from the main body.  Finally, knowledge from program3 can be used to write codes in program4.     

Thus, by learning and writing software programs from simple to complex, the robot can form smarter intelligent pathways to cater to complex tasks. 

Hierarchical intelligent pathways are also considered when learning how to write software programs.  Intelligent pathway (S1), which is a universal problem-solving pathway, can be used to write software.  If the robot fails in completing the assignment, he will loop itself to try again.  He will modify his goals, change his strategies and take another action.  This loop will repeat itself, until the robot completes the assignment, which is to write a specific software program.

To make the task of writing software more complex, imagine that the professor wanted the robot to use Unix to write the codes.  The robot’s brain has two tasks to juggle, one task is to operate the Unix system and the second task is to write the software program.  Let’s say the robot knows how to write software programs, but doesn’t know how to use the Unix system.  The professor will give a primer on the Unix system and the robot has to read the instructions and understand how to operate his account. 

FIG. 59 is an intelligent pathway, called W1, to open and use the Unix system.  W2 is another intelligent pathway to write a specific software program.  The robot’s brain manages the two tasks (W1 and W2) and merges them together to physically allow the robot to write the software program.  The robot will first login to the Unix system using his user name and password.  Then he will open a writing program called pine.  Next he will write the software program according to the assignment.  He can test the software program by executing commands in the Unix system.  The robot will apply the trial and error method to refine the software program.  Finally, after the tests are over the task of writing a software program is done.  However, in order to accomplish writing the software program (W2), the robot had to use the Unix system (W1).  The robot’s conscious managed performing two of the tasks in order to finish writing the software program.    

 

 

Learning to make decisions

Making decisions is based on a lifetime of learning from school.  Teachers teach the robot how to make decisions.  Assignments are given by teachers to instruct the robot to do tasks on decision making.  These assignments include:  worksheets, homework assignments, and real life examples.  The robot will also learn to make decisions based on trial and error.  The knowledge learned in school forms the foundation for intelligent pathways to make decisions.  The robot will modify these decision making pathways through the process of trial and error.   

When the robot is at its early stages of life, it will have to build its pathways from simple data then as it gets older and there are more data in memory it will organize the pathways into complex intelligence.  Just like how we humans have to learn to walk, to talk, to move, to eat, these machines have to go through life the same way.  Let’s illustrate the gradual forming of simple data into intelligent data by outlining a series of stages.

 

1.      innate reflexes

2.      trained to do things

3.      sequential events

4.      sentence commands

5.      give robot option commands

6.      practice makes perfect

7.      copy other peoples behavior

 

Decision making comes from many factors and the 7 stages of human intelligence form the intelligent pathways to make decisions. 

Teachers will teach lessons on decision making, ranging from easy to complex.  An easy decision making is selecting an item from a group of items.  A complex decision making is answering a question for a physics exam.

Decision making is just one form of a task.  The robot has the option of aborting the task or continuing a task or postponing a task.  Once a decision has been made that decision can be modified.  A human being can decide to go to Mcdonalds for lunch.  However, when he gets to Mcdonalds, he might change his mind and go to Pizza hut instead. 

 

Examples

Imagine a teacher wanting the robot to select one movie from 100 different movies.  The teacher will help guide the robot by giving him criteria steps (they don’t have to be in linear order) that would help narrow down the decision making process.  The teacher might say things like: 

 

1.  “what kind of movie category do you like to watch?” --  “action and comedy movies”

2.  “here are the choices.  What actor/actress would you like to see?” – “Arnold”.

3.  “based on his movies, which one would you like to select?” – “terminator 2”.

 

This is just one example of selecting a movie from a group of different movies.  There might be a situation where the robot has to select from 5 movies or 2 movies.  The robot can use the example above to select from an arbitrary number of movies.  This example will self-organize with other similar examples and a universal intelligent pathway will be created in memory to make decisions related to movie selections. 

What about buying items in a supermarket?  The decision making process will include similar steps to the example above.  In other words, it inherits some instructions from the universal intelligent pathway to select movies.  However, buying items is a different situation and requires different instructions.  Teachers will teach the robot how to buy items in the supermarket in a rational way.  The teacher will make up assignment worksheets that will ask the robot to select items in a supermarket.  In the worksheet, there might be two items.  One item is very expensive and the other item is very cheap and both items look and taste the same.  The worksheet will ask which item should the robot select.  The correct answer is the cheaper item.  If both items are the same but their prices are different, then it’s obvious that the robot should buy the cheaper item.  Let’s call this intelligent pathway (D2). 

The last example is simple.  To make the decision making process more complex, new facts should be included such as:  the cheaper item is a generic brand and is known to have product complaints, but the expensive item is a name brand and has a good reputation. 

Another factor is how much money does the robot currently have?  If the robot decides to select the expensive item, but doesn’t have the money, then by default, he has to select the cheaper item.        

Complex decision making builds on itself through a bootstrapping process.  In a complex decision making process, the robot has to use many encapsulated linear decision making pathways.  FIG. 60 is an illustration.  Intelligent pathway D2 is to select 1 out of 2 items.  D2 is encapsulated in D1.  Next, D2 might trigger a new decision pathway such as generating 10 categories based on D2’s decision and selecting one out of the 10 (D3).  Next, to narrow down the choices, a D4 is used to come up with new facts about D3’s decision.  Finally, with all the selections made from D2-D4, the robot will use multiple if-then statements to come up with one decision (D5).  Intelligent pathways D2-D5 are all previously learned in the past.  Through steps (via sentences and visual examples) by teachers, the robot is able to create intelligent pathway D1 to make a very complex decision. 

FIG. 60 

 

 

Decision making can also be learned and copied

The lessons from teachers should have relations with observation of decision making.  This way, the robot can form intelligent pathways to make decisions based on observation and not by teachers teaching step-by-step examples.

FIG. 61 is a diagram depicting D2.  The pathway shows the steps of the robot that will lead to a decision.  To the right of pathway D2, is the robot observing someone making a decision similar to D2.  Instead of seeing the hidden steps of making the decision, the robot only sees the problem and the decision.  In D2, there is:  the problem, the steps, and the decision.  In the observation pathway the robot only see the problem and the decision. 

Through relational links between D2 and the observation, the robot will assume that the person is using the missing steps to make the decision. 

FIG. 61

 

With this said, the decision in D2 can also change if the decision making in the observation is different.  Based on many examples, the intelligent pathway D2 will be modified based on the observation of decision making.  The robot will try to make decisions based on what the average of society will do.  The teachings establish the foundation for decision making.  However, the observation of decision making and trial and error will decide how intelligent pathway D2 will evolve.   

Many factors are needed for decision making.  The intelligent pathways store the probability of selection.  This probability could be a learned thing or it could be a hidden object that is not known to the robot.  A learned probability is if the robot learned to use a calculator to determine the probability of something.  The robot might be taught to calculate the probability of something in its mind.  What are the chances an event will happen?  When making a decision based on flipping a coin, the robot knows that there is a 50 percent chance, it could be heads or tails.  This form of probability is learned.  The math results of the probability will determine what the robot will select. 

In some cases, the robot can do calculations in the mind and to use that info. to select items.  In other cases, the robot has to work out the math problem on paper and to use the answer for the decision making process.  Yet, in other cases, the robot is not aware of the probability and the probability is based on hidden data generated by the patterns stored in intelligent pathways. 

 

Complex if-then statements

The intelligent pathways to make decisions are usually based on complex if-then statements (or common discrete math functions).  Many similar pathways are learned in memory and these similar pathways self-organize to form patterns.  These patterns establish computer programs in intelligent pathways to activate the linear steps to generate a decision.  Intelligent pathway D1 is one example.  D1 can have many variations and if the robot follows D1 in multiple same situations, the linear steps will be slightly different. 

The reason why is because the intelligent pathways store a computer program that will select different steps, at given times, to accomplish a goal.  The goal in this case is to make a decision or to select items.  Following intelligent pathway D1 will not always generate the same steps to make a decision.  However, the computer program in D1 will most likely come out with the correct decision at the end. 

Generating complex if-then statements are done through patterns.  The robot will use intelligent pathways to come up with meaningful logic to accomplish a task.  The intelligent pathways are universal and it will generate complex if-then statements (or any combination of discrete math functions) to generate a decision.  Universal means that the intelligent pathway was trained with many similar examples of solving a problem and it can solve the problem under various circumstances. 

 

Types of decision making

There are many different types of decision making.  Random selection is one type or logical decision making is another type.  A situation where the robot has to select from a variety of food is one example of logical decision making.  The robot has to decide on 3 food:  hamburger, spaghetti or hot dog.  He understands he can only choose one food to eat.  The robot will look into each items’ powerpoints in memory.  The powerpoints is the desirability of that item.  A function in selected pathways in memory will determine which items’ powerpoints are highest and output that information to the robot’s conscious.  The robot will select the item with the most powerpoints.  For example, if the hamburger has 10 powerpoints, the spaghetti has 18 powerpoints and the hot dog has 14 powerpoints, then the robot will select the spaghetti.   

Things can be more complex.  The robot loves to eat spaghetti and the powerpoints are high, but what if the serving of the item is small?  The factor about size is confronted next.  Although the robot selected the spaghetti, the size of the hot dog might over power the decision to pick the spaghetti.        

What if the hot dog had onions?  The robot is allergic to onions and can’t eat any food that has onions.  The robot will automatically turn down the hot dog and activate a new problem:  select between the hamburger and the spaghetti.  The hamburger is larger in size than the spaghetti, but the robot loves spaghetti more than the hamburger.  Either the robot will randomly pick or it will come to a compromise (via friendly conversation).  At the end, the robot picked the spaghetti. 

You can see from the linear decision making problem that it takes the robot time to resolve the many conflicts in the decision making process.  However, the more the robot learns how to make decisions, the better it will be at the task.  Also, the process of trial and error will evolve the intelligent pathways learned by teachers.  The steps to making a decision should be based on using logic to narrow down the possible choices -- breaking up the complexity into chunks and tackling each chunk separately. 

Other decision making, like random selection, is based on hidden data in pathways that randomly pick an item.  The robot learns to pick randomly by observing decision making situations.  For example, if the robot watches a game show and he sees players select the 2nd door more than the other doors, then when the robot is confronted with the decision of selecting random doors, the robot will choose the 2nd door. 

I was asked by a friend once this question:  “pick a number between 1 and 10”.  My answer was 7.  My friend told me that 80 percent of all people answered 7.  The reason why is because he told me to pick a number between 1 and 10.  My obvious thought was 5.  I remember he said it could be any number, so I didn’t pick 5, instead I picked a number between 5 and 10 which is 7. 

This type of example shows how predictable human beings are.  Even though they learned knowledge differently and live different life styles, society made them into the average joe. 

Other types of decision making include selecting pictures and movie clips from webpages.  When people go online they make decisions in terms of what they are searching for online.  The search engines aren’t really good at finding exact things online.  They give the user an approximate search, but all the hard work is done by the user.  When I was writing my patent applications it took me hours just to find 30 images that I needed for my drawings.  All 30 images online needed some form of modification, such as resizing or adjusting the contrast, in order to be used in my patent application.  The Patent office has rules that the drawings must follow and I had to make sure the images follow these laws. 

If the search engine was intelligent at a human level then all I needed was one search and my 30 drawings would be found.  In fact, the search engine would help me modify the 30 drawings and download them into my computer. 

That’s not the case with modern day search engines.  More complex type of decision making is to compare images or compare movie clips and select from a group of items.  There can be a mixing of different items.  The robot might have to select from 3 websites, 2 movies, and 3 pictures.  Only one item can be selected for purchase. 

Comparing a movie to a website is difficult because each item is completely different in many aspects.  It’s because of years and years of learning that the robot is able to compare different media types.  Teachers have taught the robot many different examples and he has personally engaged in decision making situations in his life.  This knowledge and experience of comparing different media types is self-generated by the robot actively making decisions.  In order to actively make a decision, teachers have to teach the robot how to compare and choose.        

One such method taught by teachers to compare different media types is to identify aspects of each item and compare these aspects.  The comparison might reveal facts from each item that might benefit the robot.  If the robot was comparing a website to a movie, maybe content of the two items can be compared instead of presentation of data.  Maybe the financial worth of the website can be compared to the financial worth of a movie.  Let’s say that a rich person had a choice to make between buying a website or owning the rights to a movie, he will have to compare the two items.  He must compare different aspects of each item and ultimately reveal how each item can benefit the rich person.  After debating with itself, the rich person will weigh the benefits and select one of the two items.   

 

<< content               next chapter >>

 

 

Home | HLAI | UAI | Books | Patents | Notes | Donation

Copyright 2006 (All rights reserved)