FAQSearchEmail

humanlevelartificialintelligence.com   

  
 chapter9

Home | Videos | Contact Us   

 
Home
HLAI
UAI
Videos
Books
Patents
Notes
Donation

     
 

                     

                         << content                                      Chapter 9

 Accomplishing various tasks in the real world

This chapter will emphasize on tasks that the robot can do in the real word, such as:  writing essays, writing software programs, playing videogames, giving speeches, having a conversation, and so forth.  But before I get into each task, I must explain the gradual forming of intelligent pathways in memory.  Intelligent pathways in memory go through a bootstrapping process, whereby new data is built on top of old data.  The robot uses trial and error and school lessons to keep pathways that lead to pleasure and forget pathways that lead to pain. 

The search function to activate related element objects from a target object can actually be learned.  If the target objects were:  cat and dog.  An element object activated might be:  “cats don’t like dogs”.  This fact is a stereotype of the two target objects:  cat and dog.  Below are diagrams showing target objects and their strongest element objects. Diagram E1 is a paper that has writings on it, showing what the strongest stereotypes are to the target objects.  Diagram E2 is a spoken sentence that tells the robot what the strongest stereotypes are to the target object.  FIG. 46 just shows the different medias that can be used to teach the robot what the strongest element objects are to a target object. 

FIG. 46

 

In the future, when this pattern is formed and an intelligent pathway is created, the robot might be able to modify connection weights between the target object and its element objects.  Intelligent pathways might be able to change data in memory according to the diagrams in FIG. 46.  For example, in E1, the target object is artificial intelligence and the stereotypes are element objects that have strong association with the target object.  What if the robot had only the first 3 element objects stored in memory?  Based on the pattern, the robot’s brain will add in the last two element objects and store that in memory.  Maybe the order of how strong element objects are can be changed because of observing E1.  The robot might have neural network as the least strong element object.  After looking at E1, the robot’s brain made element object, neural network, stronger.

Diagrams E1 and E2 can also tell the pathways that when the target object is identified by the robot, the stereotypes will activate in this linear order.  In the future, when the robot recognizes Chinese people, stereotypes will activate such as:  kung fu, fried rice, wok, and ni hau. 

The above method is important for my next lesson. 

 

1.  Writing a sentence (A1)

Teachers in the past has taught the robot simple things like the ABCs, nouns, verbs, phrases, subjects, predicates, pronouns, numbers, and so forth.  The identification of sentences are also learned, such as designating sentences into these categories: declarative, question, exclamation, comment, etc. 

The robot should know what objects are nouns, verbs, and adjectives.  Learning sentence structure is also vital.  From this point, the robot has intelligent pathways to construct sentences based on its thoughts.  The robot’s conscious will activate images, sounds or 5 sense data and the robot has to translate these activated thoughts into sentences. 

The translating of thought to sentences is actually learned previously.  A teacher will show a picture and it will describe the picture using sentences.  For example, a teacher will show a picture of a person jumping over a table; and the teacher will ask the students what does the picture mean?  Either students or the teacher (or both) will give the correct answer.  The teacher will say:  this picture is depicting “a person jumping over a table”.     

How does the last lesson relate to writing a sentence?  If the robot activated a picture of a person jumping over a desk, he has to translate this image into a sentence.  In order to do that, the previous lesson on identifying objects and actions in an image will be used.  The robot will identify the scene as a person jumping over a desk. 

FIG. 47 is a diagram depicting the process of translating conscious thoughts into sentences.  First the robot activates a thought.  These thoughts could be any 5 sense data:  sight, sound, taste, touch or smell.  It could be a still picture or a movie sequence.  An intelligent pathway in memory will be used to identify objects, events or actions in the still picture.  A simple id would be to output words like:  person, jump, above desk.  Next, another intelligent pathway will be used to convert the words into meaningful grammar sentences.  The robot knows that the subject comes first and then the predicate.  The subject will start with a noun.  Then, the predicate will include a verb that the noun belongs to and the other noun that is being affected.  At the end, the sentence:  “the person jumped over the desk” is constructed.  The robot will use another intelligent pathway to remember the sentence and, using his hand, to write down the sentence on a paper.       

FIG. 47

 

Images and movie sequences convey a thousand words.  This is why the conscious activate images more than any of the other 4 senses.  If a question is asked such as:  “what color is a heart”, the robot will activate an image of a heart and use an intelligent pathway to analyze the color.  The output is the color of the heart.  If the robot was asked another question such as:  “what shape does a heart look like in a human body?”, the robot will activate a heart from a human being, analyze the image and output its shape.  Thus, the image of a heart contains many data.  Using images to store data is better than using voice sequences.  A voice sequence is a sentence spoken by someone.  “the color of a heart is red” is one voice sequence.  If we store all facts related to a heart using voice sequences, it would overwhelm the robot’s memory.  The image of a heart contains many facts related to a heart.

 

2.  Writing paragraphs (A2) 

Writing paragraphs will encapsulate writing sentences.  A paragraph comprises a series of linear sentences.  The paragraph gives facts about a subject matter (an object, event or action). 

Teachers will teach the robot at the beginning of the paragraph to brain storm on a subject matter.  Maybe the goal of the robot is to write facts about the relationship between brain injuries and football.  During the brainstorming part, the robot is searching for specific data on football and brain injuries.  The target objects are football and brain injuries and the robot has to search in memory for element objects related to the two subject matters.  The outputs are simple facts pouring into the robot’s conscious and an intelligent pathway will select which fact to use for the paragraph (FIG. 48).     

FIG. 48

 

Sometimes, writing a sentence is done so quickly that, when an idea is selected, the sentence is automatically constructed.  It’s like speaking.  When ideas pop up, the words just come out of a person’s mouth.  Minor things like grammar checking and checking sentence structures are bypassed.

Once the first sentence is done it will continue to brainstorm more ideas and construct the second sentence.  Other facts will activate such as:  “the second sentence should continue from the first sentence” or “don’t use the same sentence structure from the previous sentence”.  While the robot is writing the paragraph, the robot’s conscious gives him rules and goals to do in order to complete the task. 

In some cases, the robot might have 3 good ideas and he will use the first idea and remember to use the 2 other ideas in later sentences.  The computer program inside the conscious will organize the ideas so that the robot knows which ideas to use first, second and last.  For example, the robot selected 3 ideas in memory and constructs the first sentence.  Next, the robot will remember the second idea and proceed to construct the second sentence (at the beginning of the second and third sentences, the robot doesn’t have to brainstorm any ideas).  Finally, the robot will remember the third idea and proceed to construct the third sentence.  These ideas can be from fragmented movies that activate in the robot’s conscious.  The robot’s brain simply selects data from these movie sequences. 

The three ideas can be from one still picture or one movie sequence. 

The intelligent pathway in FIG. 48 is only one pathway used to write a paragraph.  Referring to FIG. 49, when the task of writing a paragraph is on the robot’s mind, he will automatically generate a computer program to accomplish the task.  General tasks and rules will pour into the robot’s conscious.  As the robot selects specific intelligent pathways to write a paragraph, the tasks and rules located in their containers become more detailed.  The intelligent pathways provide more knowledge of what the robot should do in the future to accomplish writing a paragraph.  

FIG. 49 

These intelligent pathways actually define an optimal computer program to write a paragraph.  When should the robot do this and do that?  What rules does the robot follow for a given task?  What are the limited choices in a given situation?  The computer program inside the robot’s conscious will manage the complexity of the task. 

The robot does things sequentially in order to write a paragraph.  Specific intelligent pathways are selected at certain times to do things.  If you observe the complexity of a software that can write a book, you will notice that the software has thousands and thousands of individual functions.  There is a function to id words, there is a function to check grammar, there is a function to string words together to form sentences, there is a function to search for ideas and so forth. 

In a human brain, the complexity is managed by using specific intelligent pathways for given situations.  For example, when the robot has to brainstorm ideas, only functions that search for ideas are being executed in the robot’s conscious. 

Writing a paragraph is very complex because there are multiple layers of tasks being done at that same time.  The computer program inside the robot’s conscious will manage all layered tasks.  In one task, the robot wants to write a paragraph.  Encapsulated in this task are other subtasks such as brainstorming ideas, translating ideas into sentences (A1), and checking for grammar errors.  If grammar errors are found, the computer program will find a way to correct the problem.  After finishing the paragraph, the robot might have to look at the overall sentences in the paragraph and to make sure that everything is done correctly, according to writing rules.

Doing tasks, following rules and solving interruptions of tasks are all managed by the robot’s conscious.  The robot’s brain will select optimal intelligent pathways from memory at each iteration of the task:  writing a paragraph.  These intelligent pathways will form an efficient computer program inside the robot’s conscious to accomplish the task in a linear manner. 

 

3.  Writing a book (A3)

Writing a book is a much more complex task than the previous two tasks (A1 and A2).  Intelligent pathways in memory build on itself and become more complex.  In terms of writing a book, the teacher must teach the robot to do an outline at the beginning of the task.  The outline will be a summary of what the robot has to do in order to write a book.  The outline can be a hierarchical order, whereby the data goes from general to specific.  Some areas of the outline can be blank so that the robot can fill in the missing data during the writing of the book.

The rules and goals of writing a book are much more complex than writing a single paragraph.  The computer program inside the robot’s conscious has to provide the correct rules of writing a book.  These rules include:  “how long chapters should be, what the overall content of the book will convey, how much revisions should be made before completing the book, when to check the outline, how to modify the outline, how to solve conflicts in the outline and so forth”.  The rules should be about the overall book and not just about a specific chapter or section. 

FIG. 50 is a diagram depicting an intelligent pathway to write a book.  These are general steps that are required to write a book.  There can be hundreds of other intelligent pathways learned to write a book.  However, the diagram illustrates the most likely steps to write a book.  First, the robot has to brainstorm ideas for the book.  Next, an outline is created that summarizes the chapters in the book.  The construction of the outline is taught by English teachers.  Next, the robot has to do “work” – he has to follow the outline and write the chapters.  After completing the task of writing the chapters, the robot has to check for grammar errors in all pages of the book.  He must visualize how the reader will respond to the book and what the robot should do to make the book better.  The checking of the outline, checking grammar errors and checking the overall book can be done in any sequence order.  The final step is to do 3 revisions of the book before it can be published. 

FIG. 50

 

FIG. 51

 

FIG. 51 is a diagram depicting the outline of the book.  The robot will generate the outline and he will write the chapters according to its summary.   

In English books, the steps to writing a book can be given and the robot has to learn the steps and generate its own linear intelligent pathway in memory.  For example, the robot is reading an English book and the chapter is explaining the linear steps to writing a book.  The robot will create intelligent pathway32 in memory based on what he is reading.  This pathway will be further strengthened when the robot is given assignments to write a book.  Over the years, after writing many books, the intelligent pathway32 will change and new steps are added or old steps are deleted or modified.  Trial and error by the robot will determine how intelligent pathway 32 will evolve. 

 

Gradual learning      

Writing a book is extremely complex and in order to write a book (A3) the robot must first learn how to write a paragraph (A2) and learn how to write a sentence (A1).  In the writing of a book, A2 and A1 will be used many times in addition to many other intelligent pathways.  It took a human being a total of 20 years to fully learn the English language and to write a full book.   

Many years of assignments and tests and lectures are needed to form the intelligent pathways to write a book.      

A more simple type of task to A3 is to write short essays or to write letters.  The task of writing a book comes from simpler tasks like writing a short essay.  In fact, all tasks of writing anything are structured in a hierarchical tree.

Writing a book à writing a short story  à writing an essay  à  writing a letter à writing several paragraphs  à writing one paragraph  à  writing a sentence. 

Writing a short story is encapsulated in writing a book and writing an essay is encapsulated in writing a short story.  The intelligent pathways are structured in a hierarchical manner and there might be a general computer program to write any length book (short or long).  This general intelligent pathway is located at the top of the hierarchical tree.  At the bottom of the tree are specific intelligent pathways to write specific types of books (FIG. 52).

 FIG. 52

  

     

<< content               next chapter >>

 

 

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

Copyright 2006 (All rights reserved)