humanAI.com

 
taskengine
timemachine
patents
notes
donation

 

 

 

                                  << content                              Chapter 4

Autonomous prediction internet

There are two states to the prediction internet:  1.  manual work by teams of virtual characters.  2.  autonomous work by teams of virtual characters.  In the first state, each virtual character has a complete brain and they can think and act with human level AI.  In the second state, virtual character pathways are extracted from the universal brain and they are tricked into doing work.  In this case, work means predicting the future. 

FIG. 15 is an example of manual work done by teams of virtual characters.  Each virtual character has a full brain and they can think and act like a human being.  They will manually work on predictions in the prediction internet (FIG. 15 is in training mode). 

FIG. 15

FIG. 16 is an example of automated work done by teams of virtual characters.  An AI system will extract virtual character pathways from the universal brain and trick each pathway into thinking work is done.  In this case, work means predicting the future (FIG. 16 is in standard mode). 

FIG. 16

The AI time machine is the key to understanding how this method works.  In training mode, the virtual character has to do things manually to train the AI time machine to do prediction tasks (FIG. 15).  A lot of training is needed in order for the standard mode to work properly. 

In standard mode, no manual work is needed.  A user simply accesses the interface function (the input) from the AI time machine and the desired output will automatically be displayed to the user.  The prediction work is based on an AI system that extracts virtual character pathways from the universal brain and tricks these pathways in a virtual world so that work is done (FIG. 16).  The universal brain stores pathways from all virtual characters. 

 

Summary of the autonomous prediction internet

In an autonomous prediction internet, the AI system has to mimic the behaviors of teams working in the prediction internet.  In previous chapters, I talked about how teams of virtual characters work together to predict one football game.  This is just a simple example.  A more complex example includes predicting every football game in the NFL league as each game starts.  FIG. 17 is a diagram depicting how a pathway in the AI time machine can be trained to predict all NFL football games played on Earth.  Each prediction will start as soon as each football game starts. 

The reason that the prediction for each NFL football game is predicted as soon as it starts is because the teams of virtual characters will have an easier time doing their predictions.  They would filter out rare events like the quarterback is sick or the receiver was unable to attend the game.  By doing the predictions at the beginning of the game, all players, referees, coaches, and fans are accounted for. 

FIG. 17

In order to predict all NFL football games, the teams of virtual characters have to use software to be informed on when games begin.  For each game, information from electronic devices and cameras are sent to the prediction internet for processing.  The prediction internet, in this case, isn’t predicting one football game, it is predicting multiple football games that are happening at the same time.  Thus, each football game will be given a prediction tree and teams of virtual characters will be working hard to predict each games’ future events. 

If this prediction internet is trained often (using training mode for the AI time machine),  an “autonomous prediction internet” will be created (FIG. 17).  The behavior of the autonomous prediction internet can be assigned to fixed software functions in the AI time machine.  Finally, a user can predict the future for all NFL football games without real virtual characters doing work during runtime.  In other words, a user can type into the AI time machine that he wants to know the outcome of all NFL football games and the AI time machine will instantly output future events of each NFL football game that is currently being played.  The output will most likely be a short video summary of each game, highlighting the dramatic moments in the game and presenting the final score.         

 

Using the autonomous prediction internet to predict the past, present and future

In this chapter, we will discuss a complex example, whereby the prediction internet has to predict not only the future, but the present and the past.  We will make the prediction even more complex by stating that we want to predict all events, objects and actions on planet Earth.  All events, objects and actions that happened in the past, are presently happening now, and will happen in the future will be predicted accurately using the AI time machine. 

FIG. 18 is a diagram depicting a pathway in the AI time machine that will accomplish the task.  The AI time machine is in standard mode and a user can accomplish tasks through the AI time machine.  In this case, the user wants to predict all events, objects and actions for planet Earth for the past, present and future. 

FIG. 18

Of course, the AI time machine has to be trained with many examples (using training mode).  When there is an adequate amount of training this pathway can be used in standard mode.  The autonomous prediction internet will extract virtual character pathways from the universal brain and trick the pathways in a virtual world to predict past, present and future events on Earth.

In FIG. 18, the job of the virtual characters is to create a central prediction outline and to coordinate all the teams that will be doing the predictions.  This central prediction outline specifies what the goals and rules are for anyone participating in the prediction.  One goal is to devote 70 percent of team resources to predict the present, 25 percent will be devoted to predicting the past and 5 percent will be devoted to predicting the future. 

All events, objects and actions in the past, present and future are stored in an interconnected web.  A simultaneous way of predicting events in the past, present and future will yield the best results.  There is no point in predicting the future if we haven’t predicted the present yet.  Also, there is no point in predicting the past if we haven’t predicted the present yet.  For example, there is no point in predicting the future actions of a quarterback, if we haven’t predicted the quarterback’s current brain state.  By predicting the current thoughts of the quarterback, the virtual characters can understand the quarterback’s future goals.  By understanding the quarterback’s future goals, we can understand how his body will move in the future.    

70 percent of team resources are used to predict the present because past and future events depend on present events.  Only 5 percent of team resources are devoted to future prediction because predicting the future is so darn difficult. 

Another goal in the central prediction outline is to continuously predict the past, present and future.  In one minute of the prediction internet, the teams of virtual characters might predict 70 years into the past with pinpoint accuracy.  In the second minute, the teams of virtual characters might predict 2 million years into the past with pinpoint accuracy.  In the third minute, the teams of virtual characters might predict 40 trillion years into the past with pinpoint accuracy. 

As time passes, the timeline of Earth in the prediction internet gets more detailed.  These teams of virtual characters aren’t interested in predicting events they already know, they are interested in predicting events that they don’t know.  The central prediction outline should contain this goal and all virtual characters who do predictions have a clear understanding of all goals and rules contained in the central prediction outline.  

The autonomous prediction internet (API) will mimic the behaviors of teams working in the prediction internet.  Specifically they will mimic the goals and rules specified in the central prediction outline.  The AI of the autonomous prediction internet will extract virtual character pathways to do work in the prediction internet that mirrors how teams of virtual characters are doing work in the prediction internet.

For example, the API will devote 70 percent of resources to predict the present, 5 percent of resources to predict the future, and 25 percent of resources to predict the past.  All virtual characters will predict only events that are not stored in Earth’s timeline.

As the autonomous prediction internet is running, the timeline of Earth becomes more detailed.  Events in history are more accurate and detailed; and future events are more accurate and detailed.  The longer the API is running the farther into the past and future it can predict. 

There are some slight differences between teams of real virtual characters that do predictions and the autonomous prediction internet.  One big difference is that the real virtual characters can do complex predictions.  Each virtual character has a full brain and they think and act like real human beings.  On the other hand, the API extracts virtual character pathways from the universal brain and tricks these pathways in a virtual world to do work.  Because of this, the API can only do simple or limited amount of work.  The API also has to be trained adequately in order to output optimal predictions. 

There can exist a dual system, whereby the real virtual characters are working in the prediction internet, as well, as the API.  You may recall that work from the API can be assigned into the AI time machine as pathways.  The real virtual characters can encapsulate work done by the API into the AI time machine.  This means that the real virtual characters can use the AI time machine to accomplish tasks that can be done with the API. 

FIG. 19 is a diagram depicting two types of teams that are working on the prediction internet simultaneously.  Each real virtual character has a full brain and they are using technology to predict events in the prediction internet.  On the other hand, the API extracts virtual character pathways from the universal brain and tricks these pathways in a virtual world to predict events in the prediction internet. 

FIG. 19

A good idea is to use the autonomous prediction internet to do predictions on simple events, while the real virtual characters do predictions on complex events. 

A software program can be created to monitor the API to make sure that it is predicting events accurately.  If the software finds out that the API is constantly outputting wrong prediction data, then the software will tell the API to stop predicting in certain areas and tell the real virtual characters to do these predictions manually instead.  If the API is doing a very good job and the prediction output is equal or better than the real virtual characters, then the software will tell the API to devote more resources to certain predictions. 

If the API is trained adequately it should be able to do any prediction that a real virtual character can do.  The API works much faster than a real virtual character and the computer processing needed to accomplish a prediction task is a fraction of what a real virtual character needs in order to accomplish the same prediction task.

 

<< content               next chapter >>

 

 

  

Home | Task Engine | Time Machine | Patents | Notes | Donation 

Copyright 2007 (All rights reserved)