<< content Chapter 7
Robotís pathways and encapsulated work (part3)
How exactly does the laser system of the plane know what atoms to hit and when to hit them? How does the plane train the laser system? These are just some of the questions we will be exploring in this section. The idea is to create complex encapsulated work and assign these encapsulated work to fixed software functions using the universal computer program. The laser system has to be aware of all visibility levels in the clarity tree and to train itself to recognize commands from a hierarchically structured team of virtual characters.
Encapsulated work is done by entire station pathways. Each station pathway has one or a team of virtual characters working together and there are relational links between virtual character interactions. In the last section we explored how encapsulated work can be trained in fragmented sections. The training starts from the bottom up, whereby work has to be encapsulated and assigned to fixed software functions using the universal computer program.
ItĎs kind of hard to explain this process because the steps are so complex. I will be giving examples instead to illustrate this process.
Making videogames to train the plane (atom manipulator)
A videogame is created to help the virtual characters to do their tasks and to communicate with higher level commanders. A videogame is set up, whereby the controls of the plane are linked to certain goals that are given to virtual characters. The videogame also has tools and software to help the virtual characters to accomplish their goals.
There are two points I want to make: 1. the videogame is created by virtual characters and can be modified. 2. the pathways of virtual character store the usage of the videogame. These two points are very important to understanding how work is encapsulated. FIG. 31 is a diagram depicting a captain and 5 lower level workers (all are virtual characters). The captain is the main virtual character and the workers are other virtual characters that follow the command and supervision of the captain.
Each worker is assigned to certain areas of the environment. Usually, they are assigned to spaced out areas in the environment, each worker has to do tasks in their own boundary. Software in the videogame can manage interactions and conflicting problems. FIG. 32 depicts the current environment divided into 5 equally spaced out areas and each area is assigned to one worker. For simplicity purposes a simple example will be given. Imagine that there are 100 randomly scattered atoms in each area and these atoms donít move. The videogame is for the workers (players) to use a laser system to hit atoms so that a desired result will occur. The tasks are given to the workers by the captain via the videogame. Letís just say that the captain wants the workers to work together to move the atoms in the target area. The captain wants certain atoms in the targeted area to move at a certain speed and direction. The job of the workers is to play the game and to follow the rules and objectives of the captain.
The laser can shoot x amount of laser beams and each laser beam can be in any intensity. The workers have to set the coordinates of where to shoot the laser beams, how strong does the laser beams have to be, how many laser beams to shoot, and when to shoot the laser beams. Part of the videogame is to try something and if that strategy doesnít work then try another strategy. This trial and error process will loop itself until a desired result occurs.
This is where human intelligence is needed in order to play the videogame. Each worker is intelligent at a human level and they are able to receive commands from someone and to achieve these commands by using intelligence. In other words, the workersí pathways store how it thinks and senses while they play the videogame. The station pathways are the instructions to control the laser system in the plane to accomplish tasks.
This example is basically like the game of pool, where a player has to determine how hard to hit a ball and where to hit the ball so that the ball will bounce other balls around. The goals and rules of pool can be changed and the human player can still adapt to the game. The videogame for the laser system is no different.
Each worker can share laser systems or each can have their own laser system. In fact, the plane can have one laser system and all workers have to share resources. Software will determine what terminals of the laser system are given to what workers.
Building the videogame interface functions between the captain and the workers
The captainís pathway and the workers pathways donít have to be happening at the same time. The videogame can be set up to define tasks for workers and to let them submit the desire output. For example, the captain can be running at 1 millisecond per frame and the workers can be running at 1 nanosecond per frame. The captain will use the universal computer program and trick his pathways on clicking buttonA, then it will define what it wants the workers to do and what the desired output should be. As soon as the workers receive the instructions they will be hard at work trying to achieve the goals set by the captain. They can use the process of trial and error, whereby they try strategies until a desired result occurs. When the workers are satisfied with their work they will submit a desired output to the captain. Since the captain is running at a slower speed than the workers, the captain will receive his desired outputs quickly.
This method is slightly different from the previous universal computer program examples, but it comes from the same ideas. Referring to FIG. 33, the station pathway is done in the time machine. The captain is the main virtual character and the workers are the other virtual characters that must follow commands given by the captain. The captain will create a dummy software, in which it presses a buttonA. Then it will send commands to the workers, which are running at a faster speed than the captain. After the workers receive the commands they will be hard at work trying to accomplish the commands. They will work as a team, using trial and error, and to produce a desired output. When this desired output is done it will ďonlyĒ submit the desired output. The videogame will ask the workers what it wants to output and it will output the strategy that works the best.
Letís say that the command was to use the laser system to shoot atoms and to let them bounce around until they hit 50 atoms in the targeted area. The 50 atoms have to move to the right and it has to travel at a certain speed. The workers will work together using the videogame to create that desired result. Sometimes they might make a mistake and they use software to correct that problem. Their work is over when the laser system does hit atoms in the environment and they bounce around, hitting 50 atoms in the targeted area. The 50 atoms in the targeted area are moving to the right and they are moving at the speed specified by the captain. Once this desired output is reached, the workers will capture these instructions into the videogame and execute the codes to control the laser to physically carry out the instructions. When the laser does its job, the environment will be changed and the 50 atoms in the targeted area are moved according to the captainís commands. The workersí pathways to control the videogame to fire the laser system are pegged to buttonA.
Because the workers use trial and error to carry out the commands of the captain, there are some instructions in the pathway that might have to be bypassed. Self-organization and pain/pleasure by the workers will determine which of the instructions in the workers pathways are important or not. Usually, the workers are skilled in what they do and they can play the videogame and get it right the first time. If not, at least, they get better and better as they play the videogame.
The idea is to capture the work done by the workers (the virtual characters) and to assign this encapsulated work to a fixed software function (buttonA). The captain controls the ďdummyĒ buttonA and the captain uses the videogame to send commands to the workers. In the future, the captain can simply press buttonA to get the desired results without any workers. The pathway with the captain pressing buttonA is relationally linked to the workersí pathway. If many examples are trained with the captain and the workers (a station pathway) for this problem, then a universal type of pathway is created. Users can press the buttonA and the encapsulated work will occur.
How the plane moves
The plane moves by using the laser system to bounce atoms around the environment and to push the planeís exterior surface. FIG. 34 is a diagram depicting how the plane moves in different directions. When moving forward, the target area is behind the plane. The atoms have to move forward and push the plane forward. When moving backward, the target area is in front of the plane. The atoms have to move forward and push the plane backwards.
Moving forward, backward, right, left, at an angle and so forth require manipulating the joystick of the plane. When the captain wants to move the plane, he has to gently push the joystick slowly at first, then position that joystick to the speed it wants to travel. The joystick isnít a fixed function like a button so itís kind of hard to put encapsulated work into a joystick.
The captain has to use software to train the joystick in increments. FIG. 35 is a diagram illustrating three increments of the movement joystick. In the first increment, the captain pushes the joystick forward slightly, then he has to have the workers use the laser system to manipulate the environment. Next in the second increment, the captain pushes the joystick forward harder, then he has to have the workers use the laser system to manipulate the environment. Finally in the third increment, the captain pushes the joystick forward harder, then he has to have the workers use the laser system to manipulate the environment.
The first increment might include the command of moving 100 atoms in the targeted area to push the plane forward. The second increment might include the command of moving 300 atoms in the targeted area to push the plane forward. The last increment might include the command of moving 9,567 atoms in the targeted area to push the plane forward. Each atom might be given a force. For example, the first increment might include light force, while the last increment might include medium force.
The captain has to do this for all speeds and directions of the plane. Self-organization will do the rest to average out how the joystick is handled and what the desired output are in every increment.
The controls of the joystick will only work if the plane doesnít change its shape. If the plane does change its shape the joystick has to be modified. When the plane has to manipulate objects in its environment a different joystick is needed and this joystick will have to be trained with many different objects in the environment. For example, if the joystick can lift objects in the environment it has to be trained with lifting many different types of objects. Lifting a book is different from lifting a truck. The joystick has to be trained with lifting a book and also lifting a truck. When the opportunity presents itself, and there is a table in the environment, the AI of the plane will know what encapsulated work is needed to lift the table. The AI will find the pathways in memory that has an object that matches to the size, shape and weight of the table.
The joystick increments of training donít have to be perfect. The software from the videogame will manage the increments. However, letís say the increments are self defined by the captain. FIG. 36 is a diagram depicting increments trained at non-spaced out manner. The encapsulated work in each increment may not be correct all the time. But because of self-organization, the joystick increments average itself out and a smooth joystick movement results.
All controls of the plane including radio buttons, software interface functions, joysticks, monitor, switches and so forth has to be trained in this fashion. Work has to be encapsulated repeatedly. The more complex the task is the more encapsulated work is present.
The one thing I want to note is that in regular virtual character pathways, the software instructions and functions are not stored along with the pathways. Only the virtual characterís experience with the software is stored. This separates virtual character pathways and software programs into separate data.
When the plane wants to use a virtual character pathway (or station pathway) to do work, it needs a physical copy of each software used in the pathways. For example, if the pathway records the virtual character using internet explorer to search for information from the web, it will get a physical copy of internet explorer and it will use the pathways to control the certain functions in the software.
This method works because if you have a function in a software and this function is represented by a button. The virtual character pathways record the pressing of the button. The result is the function executing after the button is pressed. All of the steps in the function and the computer codes to execute the function are not stored in the virtual characters pathways. The pathways get the function from the physical copy of the software. Another benefit is that the virtual character pathways can be used to work on similar software. For example, instead of using internet explorer, the AI can use netscape.
Videogame training (details)
The videogame has tools that let the workers see their area in a clearer manner. The software can display the 3-d shape of one atom, a molecule, or a group of molecules. The workers need this tool to determine how two atoms will interact with each other. FIG. 37 is a diagram showing how two atoms are positioned in different areas. The job of the worker is to us the laser and determine how the beam of light will hit the first atom so that it can bounce the second atom in a certain direction and speed. This process will be called E1.
E1 can be viewed in any angle or dimension Ė the monitor can show a sky view of the atoms or it can show a 3-d angled view of the atoms. The videogame has image software to show the worker details of E1.
E1 is just one task of the worker. In diagram E5, the job of the robot is to zap the first atom and let it bounce around until it reaches the atom in the targeted area. In some sense this problem is just like the game of pool. The worker has to work in sections. First it has to know how the laser can hit the first atom to bounce the second atom towards atom3. When that is successful the robot will use the videogame software and record the instructions. Next, it has to find out a way to use atom3 to bounce atom4 toward atom5. This will go on and on until the instructions to bounce atom1 to atom7 is perfected.
The process of trial and error has to be done. During each try, the robot can use the videogame software to save certain behaviors and use this behavior in the future. The worker also has the ability to analyze the atoms microscopically to see where the atom should be hit in order to generate a desired output. If the worker made a mistake, he can retry the last play and see where the atom was hit and to use software to determine where the atom should be hit in order to bounce the atom in a certain direction and speed.
Referring to FIG. 38, the bouncing of atoms has to be done in sections (E1, E2 and E3). The worker will start with E1, then when it is successful it will start on E2, next when it is successful it will start on E3. Along the way, it will use the videogame tools and functions to help accomplish its goals.
Training small distances then longer distances
Similar examples will self-organize in memory. Of course, the more simple the example is the easier it is to find a pattern. The more complex an example is the harder it is to find a pattern. One simple example is E1 and a complex example is the diagram in FIG. 38. Basically, the longer the first atom is to the target area the more complex the example is.
Referring to FIG. 39, the videogame will first present short distance examples from the laser system to the target area. As the worker gets better and better at playing the game, the videogame will present longer distance examples. As the worker plays the game, patterns are found and math equations are set up for bounce behaviors. The idea is that the target area can be anywhere and the environment can have any number of atoms and they can be positioned anywhere, the pathways will still be able to shoot the laser to move atoms in the target area.
The self-organization is very important because it generates hidden objects. These hidden objects will be in the form of math equations that can cater to infinite possibilities. For example, in E1, the second atom can be anywhere, but the hidden object (a fixed math equation), will help bounce atom2 to atom3 with the same force and direction.
Self-organization will create floaters in memory. The most important floaters will be outlined while the least important floaters will not be outlined. Since the controls of the laser system is from intelligent workers (or virtual characters) then the strongest floaters in memory are intelligent pathways. This is important because some neural networks use random training at the beginning to set the foundation for the AI. In the atom manipulator nothing is random and everything is based on intelligence. It has to be guided intelligence because if you try to train the videogame to randomly hit balls, the desired outcome will not be met regardless of how many times you train the videogame. The videogame has to be trained by an entity with human-level intelligence.
The pattern to E1 and E5 is the laser shoots one beam starting from the closest atom. Then it has a target area and the first atom has to bounce around until it hits an atom in the target area. The patterns found between similar examples will set up math equations for the laser to hit atoms. If the laser system is trained adequately the result is: you can set the target area anywhere (near or far) and the environment can have any number of atoms positioned in various areas, the laser system will still have the instructions to move atoms in the target area. That is the ideal outcome of this videogame.
By training it using short distances at first and then longer distances, behavior of bounces can be grouped together. Referring to FIG. 40, notice that in all three gameplays there are repeated behavior. E1, E2, E3, E5 are all repeated behavior. Instead of trying to find patterns in E1, E2, E3 and E5, there are copies already stored in memory and these copies contain hidden objects. For example, in the third gameplay, E2 and E3 already exist in memory and the AI doesnít have to worry about finding hidden objects for these two sections. The AI will try to find patterns in J1 and J2. They will compare this example to similar examples already stored in memory to find the hidden patterns. Even entire gameplay like E5 can be encapsulated. This makes it easier for the pattern recognition to find patterns and to find hidden objects.
More complex examples
The illustrations given above are very simple. The atoms are stationary and there is only one target area. In a more complex situation, the atoms are constantly moving and the laser system has to predict where these non-intelligent atoms will be in the future, so that it knows how to shoot the laser to bounce atoms to the target area. In real life, wind moves quickly outdoors, while wind in a room moves slowly. The laser system has to train itself to work in a dynamic environment.
Also, in the real world, the distance from the laser to the target area might be billions and billions of atoms/molecules. The laser system in the plane doesnít have to be perfect at an atomic level. As long as air is manipulated in the target area, the laser successfully did its job. For example, there can be infinite ways that atoms can bounce around to get to the target area. If the laser can execute one successful way to bounce atoms to the target area, that would be considered a success.
By tracking every atom, electron and em radiation, the atoms can bounce in a way that will minimize interacting with other atoms. By minimizing atom interactions energy from the bounce is conserved. Letís say that you wanted the laser to shoot an atom against a gust of wind. The objective is to avoid any atom that will hit the atom in the direction of the wind. By using the signalless technology, the laser can know where all the atoms of the wind are and to bounce an atom around to avoid any interacts with them. Itís kind of like navigating a ship through an asteroid belt. Because the signalless technology tracks all atoms, electrons and em radiations, the chances of success are very high.
The signalless technology gives the atoms a sense of intelligent guidance. If you try to randomly fire an atom against a wind gust, most likely the wind gust will prevent the atom from getting through. Itís kind of like randomly navigating a ship into an asteroid belt. The atom manipulator does things in a hierarchical manner. It might not be able to track every single atom or em radiation, but it can track larger objects like molecules or tiny particles. Instead of using the laser to fire an atom at a gust of wind, it can fire a molecule.
Thus, the atom manipulator can accomplish tasks in an approximate manner. This is why the videogame trains the laser system in a hierarchical manner. This is why the plane has a clarity tree that sees things in different levels of clarity. And this is why multiple virtual characters have to train the laser system at all visibility levels, either simultaneously or independently.
By the way, the signalless technology is only concerned with tracking atoms at the moment and what will happen in the short future. That is the difference between the atom manipulator and the time machine. The time machine is a more difficult technology to create because, a perfect timeline of all atoms, actions and events have to be mapped out not only for the short past/future, but distant past/future. The atom manipulator can be a much easier technology to build. The atom manipulator only needs to track non-intelligent objects and it can guess where the intelligent objects might be. For the most part organic species are much larger than a molecule. Even viruses are made up of thousands of molecules.
The atom manipulator can track as much non-intelligent atoms/molecule as possible and use physics to determine where they will be located in the short future. Tracking solid matter is easy because they need force in order to move, but tracking gas and liquid is harder. The atom manipulator will do very well in gas/air because atoms can move freely.
Team work to accomplish tasks
A station pathway is a team of virtual characters working together to accomplish goals. This team of virtual characters can be structured in any manner. The diagram in FIG. 41 shows that a station pathway is structured in a hierarchical manner. A captain is in charge of 5 workers. His task is to monitor the visibility level D2 and D3, while he instructs his workers to do tasks in visibility level D4. A videogame software will be used between the captain and his workers. This videogame provide tools for communication and aid in accomplishing tasks.
The videogame is specifically designed to control a laser system to hit atoms and let them bounce around until atoms in the targeted area of the environment are manipulated. The videogame comprises multiple workers that are assigned to different areas in the environment. The software in the videogame will allocate the laser system each worker will use and which areas they have to focus on.
The captain will input into the videogame a target area and specific instructions to move atoms in the target area. The videogame will send programmed instructions to specific workers to do things based on the input by the captain. Next, the workers will work together and by themselves to accomplish the goal the captain wants to accomplish.
Gameplay is a term used for each bounce pathway plotted by a worker. For example, in the last chapter E1 and E5 are gameplays. Each worker has to know that he/she is not working alone and that the environment changes based on all 5 workers. The gameplays each work does will be inputted into the videogame and the video monitor of the environment will change as a result of the gameplay. Each player has to confirm that they want to use a gameplay before it can be used to update the environment. Often times a worker will devise 20 gameplays and select the most optimal one to be inputted into the videogame.
For simplicity purposes all atoms in the environment are stationary and they donít move unless they are acted upon. As each worker inputs their optimal gameplays the environment changes. The videogame software has to keep track of which atoms are used by which workers. If one atom is used by worker1 and worker2 wants to use the same atom, the videogame will forbid worker2 from using that atom. Multiple usage of atoms will lead to conflicts between workersí gameplays.
In order to solve this problem there are three methods that can be used in combinations: 1. common knowledge of atom priority. 2. the videogame software outline atom priority. 3. the captain defines areas in the environment that has priority or not.
(1) Referring to FIG. 42, common knowledge of conflicting gameplay can be learned in books and manuals. Strategy books can be read to better play this game and how to interact with other workers. One strategy might be to stay away from atoms closest to other workersí boundary area. For example, in the first area in the diagram, all atoms that are located in area 4 will have top priority, while atoms located outside area 4 will have low priority.
What this means is that the worker can use the atoms within area 4 and be confident that these atoms will not conflict with other workers gameplays. Common knowledge in books will also give strategies for the workers to identify sections of atoms that might have top priority. There might even be steps that a worker has to go through to find the priority of atoms. One of these strategies is to communicate with other workers and to come to a compromise when they design their gameplays. Workers have to communicate especially when they have to use atoms from another workers boundary area.
(2) In the second method, the videogame has to outline, for all the workers, the priority of atoms. To minimize gameplay conflicts the software will give the workers a prioritized area based on whatever gameplay has already been inputted or in the working state. When a gameplay is inputted, then it is confirmed that atoms used in the gameplay are reserved. If a worker is in the working state of a gameplay, the atoms used should also be communicated with other workers because other workers might be using the same atoms. The videogame will look through the inputted gameplay and the working state gameplays and prioritize all atoms in the environment for all workers.
(3) The captain will be observing all gameplays inputted by the workers and he might be disappointed with some gameplays because gameplays might have high levels of conflict in a particular area. The captain might delete unwanted gameplays and tell certain workers to redo their gameplays in a different area. Thus, the captain can help to prioritize the atoms in certain areas.
Working together to play the videogame
Referring to FIG. 42, the target area falls in the boundary of worker2 and worker3. That means these two workers must do most of the work. Worker2 and worker3 controls the closest lasers to hit the atoms in the target area. Other workers have to bounce atoms around for longer distances. This means that worker2 and worker3 have to work closely to plot gameplays that will reach the target area. They might argue back and forth using sentences like: ďno, that is my atom. Go get the nearest atomĒ or ďbut that will take a long time to get to the target areaĒ or ďif you use that atom and I use that atom, both of our gameplays will lead to the target area in the shortest timeĒ or ďyou concentrate on this area and I will concentrate on that areaĒ. Sentences like these will be exchanged back and forth between workers to make their optimal gameplays.
The videogame software can also reassign worker1, worker4 and worker5 to help worker2 and worker3 to do their jobs -- to help them plot out gameplays. The videogame software is essentially giving each worker new boundary areas and new goals to achieve. This will reallocate resources of the videogame, wither that be workers or laser terminals.
The videogame software has tools that the workers can use in addition to help from the captain. Calculations of atom interactions can be done quickly by AI software. If you watch an episode of CSI, you will notice that these detectives use software as tools to find information. The workers are using the videogame software in the same manner.
When everything is said and done, all workers have accomplished their goals set by the captain. They have resolved their differences and come to a compromise. The unified gameplays will be the desired output that will be sent to the captain. The gameplays will be stored as encapsulated work by all the virtual characters in one station pathway (in this case, the station pathway is one captain and 5 workers). This station pathway will be universalized and the captain will assign this station pathway to a fixed software function. These station pathways, if trained adequately will represent the AI of the plane; and will be used to control the laser system for one function in the future.
Predicting the future and each gameplay
By the way, the gameplays are plotted in a future timeline because we are dealing with predicting the environment in the future. The future prediction function needs to predict how the inputted gameplays will affect the environment in the future. This example is easier because the atoms in the environment are non-moving and they donít move unless acted upon.
Everything has to be trained in the virtual world. The signalless technology will capture a short sequence of the environment and track all atoms, electrons and em radiations. Then, this short sequence will be presented to the workers who will control the laser system to manipulate that environment. The short sequence only records the environment without any tampering.
The future prediction function is used to predict what the environment will be in the short sequence if the laser system was used. This means that every gameplay inputted into the videogame will update the future predictions to include the gameplay. The videogame is responsible for modifying the short sequence and providing an accurate depiction of what that short sequence will be if the inputted gameplays are used to manipulate the environment.
The complexity isnít as difficult because the laser system only manipulates a small fraction of the atoms from the environment. Only the bounces and the atom interactions as a result of the laser beams fired by the plane need to be changed in the short sequence.
The question about how does the modifications of the short sequence happens should be asked? The answer is by using the simulation brain on how atoms interact with each other. The bounces can be calculated by matching pathways concerning atom bounces. The laser interacting with the original atom can be calculated by finding a pathway that matches to that object interact. The simulation brain stores the behavior, properties, object interactions for a given object or groups of objects.
For example, the simulation brain has to be trained with many examples of how atoms interact with other atoms or how electrons act with other objects. A laser is one object and a molecule is another object. A universal pathway has to be trained regarding the interactions between the laser beam and the atom. The future prediction will use these learned pathways to fabricate what might happen to the environment if the laser system was introduced in a given short sequence.
Training the atom manipulator in the virtual world and testing it in the real world
Why does the training of the atom manipulator have to be done in a virtual world, why not in the real-time? The reason why is because itís very hard to build a laser system with fixed functions and fixed computer codes. It has to be trained through a videogame. We have to give the atom manipulator a training session (inputted gameplay) in the virtual world. This would mean all the work that is needed to control the laser system for one training session has to be done in the virtual world. All of the debates between workers, all the captainís orders and all the encapsulated work has to be done within a fraction of a millisecond.
In the computer, time is void and depends on the processing speed. This can be used as an advantage because all the work needed to control the laser for one training session can be done in the virtual world during runtime. After the work is done, the laser system can test the training session during runtime to see if the predicted results of the laser system are correct or wrong.
For example, we can use one training session and the plane can fire beams of light at atoms in the environment based on the training session. The robot piloting the plane will observe if the predicted future of the training session (inputted gameplay) is accurate or not. If the future prediction is accurate, then the laser successfully fulfilled its mission for that one training session. If it failed, then it can train it with a more desired training session in the future. The atom manipulator (the plane) will learn as more training is presented. Each training session has to be perfect or near perfect so that the AI can average all the controls and what these controls do to the laser system. There might be mistakes made, but self-organization of station pathways will average everything out.
FIG. 43 is an illustration depicting how the atom manipulator will train itself using the virtual world to design one training session and using the real world to do the physical training.
First, the current environment is inputted into the plane. The plane will use the signalless technology to track all atoms, electrons and em radiation from a targeted environment. This short sequence will be handed over to virtual characters in the time machine. These virtual characters can work as a team or by themselves. They will design the training session using a videogame software and they will also create a future prediction for the training session. Next, when the captain is satisfied with the training session (inputted gameplays by the workers) he will transmit this information to the robot in the real world and the training session will be executed to be tested in the real world. The robot will see if the desired output has occurred. If not, then the robot will tell the workers in the time machine to do a better job in the future or to input some advice.
Training the atom manipulator in a dynamic environment
The examples above only describe atoms that are stationary in the environment. In real world situations atoms/molecules move in a dynamic way. They move based on physics and chemistry laws. Wind outdoors moves fast, while wind indoors moves slowly. The plane has to be trained under many situations.
The example above must be adapted to include a dynamic environment. Instead of atoms staying in a fixed position, the atoms are constantly moving. The workers have to come up with gameplays that will predict future positions of atoms. Where will a certain atom be in the future and how can the workers bounce that atom to hit other moving atoms? The future prediction function has to also do a good job in predicting the short future environment, and also, to modify this short future environment as the workers input new gameplays.
Training an adaptable laser system
Letís say that the first training session was a failure and the future prediction is also a failure. The plane has to adapt and to teach itself another training session. This time, the future prediction will modify itself based on the updated current environment. By training the plane with sequences of adaptable updates, the AI can learn what is desired and what is not desired. It will form patterns to keep the desired training sessions and to delete the bad training sessions. The plane can also know that the bad training sessions are not wanted and that this isnít what the robot pilot is looking for.
However, it is better to train the plane perfect or near perfect in every training session. The more desired training the plane goes through the more likely it will behave in a desirable way in the future.
Training to correct previous mistakes
If one training session is badly executed and the results are wrong, the plane can modify the previous training session to make it correct. For example, if the first training session is wrong and the laser miscalculated, the plane can come up with a second training session that will correct the first training session. The second training session might include introducing more laser beams into the environment to bounce misguided atom bounces back to their original course.
This can be done repeatedly until all previous mistakes are corrected or a desired outcome results.
This method is important because the plane (the atom manipulator) doesnít predict the exact future, but an approximate future using its laser system. The future prediction function isnít concerned with predicting every atom, electron and em radiation on planet Earth every fraction of a millisecond. It is only concerned with predicting every atom, electron and em radiation within a focused area. Intelligent objects like human beings and animals are extremely hard to predict and they are usually ignored. However, when dealing with air and the open sky, most of the objects are non-intelligent and they are easily predicted. The simulation brain stores non-intelligent objects and their interactions. Since non-intelligent objects are based on physics and chemistry, they are systematic and they have repeated patterns.
Also, people and animals sense and act every second. Atoms and molecules act every fraction of a nanosecond. To the atom manipulator, intelligent objects behave very slowly and they can be considered stationary objects. Thus, the atom manipulator manipulates the environment so quickly that the infinite possibilities of an intelligent object donít really matter. Large intelligent objects like human beings and animals think slowly so they wonít affect the atom manipulator. Small intelligent objects like viruses and spores are too small so they wonít affect the atom manipulator.
However, it is prudent that the plane have the ability to predict an accurate future of how the laser system changes the environment. It must also predict what pre-existing objects in the environment will do in the future as a result of the laser beams. The plane should use adaptive methods to change the environment in the moment that something unexpected happens.
Copyright 2007 (All rights reserved)