<< content Chapter 9
Universal artificial intelligence for machines
The first invention I designed back in 1999 was the universal artificial intelligence program. This is a software that can basically play any videogame. One day in 1999 I was playing Mortal Kombat and I got very bored playing with the computer. I decided to write a software that can play Mortal Kombat. As I played other games, I was wondering if it was possible to write a universal AI software that can play any videogame.
I think I proposed many different universal artificial intelligent programs from 2006 to 2008. In this chapter we will review on some of these UAI programs.
A universal artificial intelligence program can control a car, a plane, a forklift, a boat, a train, a motorcycle, an air control tower, and so forth. The artificial intelligence can be used on any machine to do any human task. In other words, the artificial intelligence is universal.
One proposed idea was converting a machinesí sensed data from the environment into a videogame and having a robot play the game to control the machine. I proposed a virtual world where the robot (a virtual character) is controlling a videogame to control a physical machine. The virtual world changes when using different physical machines.
Another idea was to build a dummy physical robot that has ďlimitedĒ pathways to drive a car or fly an airplane or control any machine. The robot can download pathways for a specific type of machine. This idea is very useful because the robot can control any physical machine, even cars and planes that were built in the 1920ís. Instead of buying an AI car or AI plane or AI truck, the physical robot can simply get into a car, operate it, get out of the car, get into a plane, operate it and get out of the plane. These dummy robots work very well in sewing factories. They can mass produce clothing by using many different sewing machines. They can work in a team like organization to accomplish sewing tasks. For example, a group of dummy robots can cut out the fabrics and give the fabrics to another group of dummy robots, which are responsible for sewing the parts together. Finally, another group of dummy robots will add the finishing touches to the clothing such as nailing buttons, cutting excess strings, ironing the clothing and packaging the clothing to be shipped to department stores.
Controlling a car
The idea for an autonomous car is for a car to drive on its own based on minimal user input. The user might give voice commands like: drive home, drive to the nearest library, drive to the beach, and so forth. The AI car must obey the userís commands and to safely accomplish these commands.
The user can also input commands to the AI car through an onboard computer. He might have to fill in a form and press a submit button. The AI car will process the command after the user submits the command form.
This AI car is supposed to act as an intelligent entity that can not only follow simple commands, but to give opinions, alert the user to danger, diagnose the hardware and software of the car, and so forth.
The data structure of the AI car
Teams of virtual characters will be controlling the AI car. Each virtual character is intelligent at a human level and they can think and act like a human being. Training has to be done first before the AI car can drive autonomously.
The AI time machine is used to encapsulate all the work done by teams of virtual characters. In training mode, the AI car will record both the work done by the robot pathways (the user) in the virtual world and the virtual character/s pathways in the time machine world. These pathways will be stored in the universal brain. In standard mode, the AI car can be fully automated by extracting pathways from the universal brain and tricking these pathways in a virtual world to make the virtual characters do work (FIG. 36).
In this case, the AI time machine serves as a central brain for a physical machine or an army of machines. The universal brain comprises two other brains: 1. virtual world brain, which stores robot pathways (the user). 2. time machine world brain, which stores virtual character pathways (the workers).
The first step is to input the current pathway into the AI car. Sensed data from the AI car like vision and sound will be part of the current pathway. Another part of the current pathway is the activities of the user in the AI car. Things like voice commands from the user or software commands are stored in the current pathway. Another type of data stored in the current pathway is the various electronic devices in the car such as internet access, videogame, TV, air conditioner, phone calls and so forth. For example, a user might call a friend to tell them that he will be late for a party, the AI car will record this information in the current pathway.
The current pathway is a snapshot of what is happening in and around the AI car. A pathway must be extracted in the universal brain that best matches to the current pathway, which is called the optimal pathway (note: the optimal pathway also factors in future predictions).
The current pathway will be matched to a robot pathway in the virtual world. The best match, called the optimal pathway, will be extracted. The optimal pathway will extract its dependable virtual character pathways in the time machine world. These virtual character pathways contain the instructions to operate the AI car. The AI car will trick the virtual character pathways, called a station pathway, in a virtual world to do work. The work is used to control the AI car in an intelligent manner.
In terms of driving a car, the current pathway contains information about what is in and around the AI car. Things like cameras and sensors can be set up inside and outside the car, and the sensed data is part of the current pathway (eg. a camera is mounted in front of the car to see the road). Another data stored in the current pathway are camera information inside the car that focuses on inputs from the user (commands from the passenger).
On the other hand, virtual characters are analyzing data from the current pathway via a software program. Because there are lots of information to look through at any given moment, the virtual character uses a software program to organize data and to help present the most important data to the virtual character. Under normal circumstances, the main virtual character (called the captain) will analyze videos of a user and to listen and watch for commands.
The captain's job is to interpret the user's commands and to carry out orders to lower level workers. If the user gives an order, such as: "drive me home", the captain will use human intelligence to identify the user, know where he lives, and give the order to drive to that destination. If the user gives a command to check his email, the captain will check for any new email. If the user gives a command to turn on the radio, the captain will turn the radio on based on what he thinks the user would like to hear.
In another case, the user might see a UFO in the sky and ask the AI car what that thing in the sky is. The captain will control a software program to view the cameras around the AI car and to focus on the UFO in the sky. It will analyze the video and interpret what he thinks is in the sky. This can only be accomplished if the captain analyzes the user's body movements, such as the user's pointing finger. He must determine which area in the sky the user is pointing to; and to view this area of the sky using the cameras mounted on the exterior of the AI car.
The point I'm trying to make is that the virtual character (the captain) is analyzing important data from the current pathway via a software program. The captain's main job is to analyze the user, manage commands, and to follow every command given by the user.
FIG. 37 is a diagram depicting a pathway from the universal brain. This pathway stores a team of virtual characters working together to control the AI car. The inputs are the information from the AI carís senses and the userís input (robot pathways). The outputs are accomplished work done by the team of virtual characters (virtual character pathways).
This pathway depicts the AI carís activities over a long period of time. The user inputs commands and the virtual characters give outputs (notice that the current pathway is just a small sequence in the linear inputs and outputs). The current pathway will move incrementally in the pathway as time passes. The pathway shows the linear tasks that the team of virtual characters has to accomplish based on the linear inputs from the user. For example, below is a list of linear inputs the user gave to the AI car.
1. drive home
2. I want to see my email
3. search for the cheapest TV in this area
4. I change my mind, drive to Pizza hut instead
5. its getting hot in here, turn on the AC
6. how much longer before we get to Pizza hut
7. thatís too long, drive me to the closest fast food restaurant
These commands are the inputs of the AI car. The outputs are the work done by the virtual characters. Not only does the AI car have to drive around, but it has to do tasks given by the user such as checking email, turning on the AC, answering questions from the user, solving interruptions, doing research over the internet and so forth.
Referring to FIG. 38, a team of virtual characters is called a station pathway. There is a captain, who is in charge of decision making and there is a driver, who is in charge of driving. An AI car is a simple example and it doesnít really need two virtual characters to operate. In the next example, we will discuss why itís important that a team of virtual characters work together to control a complex machine.
Each virtual character has human level intelligence and is using technology to do their work efficiently. The virtual characters will most likely be using the AI time machine to do work. Other software programs can be used such as the windows operating system, a web browser, search engines, or a software calculator or any apps on an iphone.
Referring to FIG. 39, the car software is a software program specifically handcrafted to help the captain do his job. Itís functions are actually adaptable (I will explain this later). As far as working in a team, each member will have a handcrafted software program designed for their roles. The captain will have a software program to manage multiple tasks and make decisions, the driver will have another software program to drive the car, the intelligence officer will have another software program to gather useful information. The team software will have interface functions so that one member can communicate with another member.
Each virtual charactersí roles, rules, status, powers, limitations and objectives are based on common knowledge found in books, reports, college courses and so forth. Every member of a business knows what their roles are because of knowledge learned in business school. These virtual characters have gone to college, studying in specialized fields. The captain was trained to solve problems, follow commands, solve interruptions, manage multiple tasks, make decisions and so forth.
FIG. 40 is a diagram depicting one station pathway to do multiple tasks from the user. The input from the user is to drive home. The captain will identify the user and do research, such as find out where the user lives. Next, the captain will use GPS software and plot out a route from the current location of the AI car to the userís home. Finally, the captain will send this information to the driver and he will navigate the car either manually or using automated software.
The next input from the user is to check for emails. The captain will open up an internet browser, login to the userís email account, and send the new emails to the user. The next couple of inputs from the user are to do research over the internet.
Tasks that are done over and over again can be assigned to fixed software functions in the AI time machine, either manually by a virtual character or automatically. The AI car might detect that an input from the user like ďdrive homeĒ can be assigned to virtual character pathways like E1 and ďcheck for emailsĒ can be assigned to virtual character pathways like E2. These two newly created fixed interface functions will be assigned to the car software for the captain to use in the future. Work that has already been done numerous times can be saved in the AI time machine as accessible interface functions for virtual characters. For example, Letís say that the AI time machine has a function to read a website and to summarize the content to the virtual character. If the user in the AI car inputs the command: ďI want you to read this website and summarize its contentĒ, the captain can use the AI time machine to accomplish the task, instead of doing the task manually. The captain will take the output from the AI time machine and give this information to the user.
In an autonomous car, the user simply has to tell the AI car where to go and what to do and the AI car simply extracts pathways from the AI universal brain to do work. No real virtual characters are needed, during runtime, to do the driving.
Controlling an armored car
A more complex AI machine is an armored car designed to do battle. FIG. 40 is a diagram depicting 4 virtual characters working together to do battle. The goals of the AI armored car are based on the constant goals of the captain. One of its goals is to constantly monitor the current surroundings to look for dangers. If danger is identified, the team of virtual characters will work together to get the occupants in the AI armored car to safety. Another goal of the captain might be to follow orders given by the user.
These orders might include driving supplies to a destination where there is a war zone going on, driving safely from one destination to the next, doing battle in a blockade, attacking enemy fortresses, and so forth. The more tasks that are trained by teams of virtual characters the more capabilities the AI armored car will have.
Orders that are given by the user which includes battle commands must be cleared by superior officers. This is standard procedure when it comes to battle commands. The captain will analyze the user and determine what the userís rank is and see if he has the authority to do battle based on rules set by the army. For example, if the user is a soldier and he wants to attack a small fortress, the captain will know (based on common knowledge) that he doesnít have the authority. However, if the user is a lieutenant and he wants to attack a small fortress, the captain will follow his command. The captain knows what is allowed and what isnít allowed by the user. He also knows the hierarchical rankings of users and their limitations. The captain is using common knowledge learned in military school to determine hierarchical rank and limitations.
Referring to FIG. 40, the captain is the person responsible for decision making for the AI armored car. The driver is responsible for driving the car based on the destination given by the captain. The captain is responsible for changing destinations. The shooter is responsible for shooting enemies and to protect the occupants in the AI armored car. Finally, the intelligence officer is responsible for gathering information from the internet and sensing devices to help the team do its job.
Letís use an example of how the AI armored car works. The user gives a command to the AI armored car to go to a certain city. The team of virtual characters will do as commanded. The captain will plot out the course using a GPS device. He will send the destination information to the driver, which drives the car. Upon reaching the city, the AI armored car is ambushed and there is a blockade in front of their path. The captain will tell the user and the occupants that the AI armored car is now taking control of the situation. This basically means the team of virtual characters will not process any commands given by the user.
The teamís goal is to work together to get safely out of the area. The intelligence officer will monitor streaming data from satellites to find out about its surroundings. The intelligence officer might spot two enemies behind the car and tell the shooter this information. The shooter will either take out the enemies or wait for identification before shooting. Under the rules, the shooter doesnít need permission from a captain to fire if he thinks the armored car is in danger.
The communication between team members is by voice. Some information can be conveyed electronically, but the majority between the team members is by voices.
The captain will ask the intelligence officer to find a safe route to go to with minimal resistance. The intelligence officer will use sensing data, satellite data and any electronic data to find the safest route. The intelligent officer will give this information to the driver and he will drive the AI armored car there.
The captain will get constant updates from all team members and he has to make decisions that will benefit the team.
The autonomous armored car will work by extracting station pathways from the universal brain. Station pathways are teams of interconnected virtual character pathways. Each virtual character pathway in the station pathway are tricked in a virtual world to make each pathway think that they are doing work. Thus, no real virtual characters are needed, during runtime, to operate the AI armored car.
Future prediction for the AI armored car
If the AI armored car is ambushed, the AI will predict the future in what will probably happen. These future predictions include the actions of the AI armored car, as well as, the activities of the enemy. Thousands of alternative cases will be predicted and the AI armored car will select the best future pathway that benefits itself.
Thus, the station pathways or groups of virtual character pathways are not acting based on the best current environment, but based on the current environment, as well as, the best future prediction.
In order to do this, everything I talked about in this patent application must be used. The signalless technology maps out the current environment atom-by-atom. It identifies all enemies and objects in its surroundings. Another team of virtual characters, besides the team of virtual characters controlling the AI armored car, has to predict the future actions of each enemy and what they will do in the future.
Each future prediction will use a different virtual world. The team will extract virtual character pathways from the AI time machine and trick them in these alternative virtual worlds. The virtual character pathways in a virtual world (a future prediction) with the best results will be the virtual character pathways selected to control the AI armored car to act in the future.
Time dilation between levels of virtual characters
FIG. 41 is a diagram depicting time dilation between levels of virtual characters. 1 nanosecond of the captain is 10 seconds for the intelligence officer. 1 nanosecond is equivalent to 4 weeks for the 6th virtual character. The times for each virtual character are different because some jobs have to be done quickly. The intelligence officer has to do his job really fast so that the captain will get results quickly. Maybe the intelligence officer has several lower level virtual characters working for him. These virtual characters can also have same or different time speeds.
A more efficient system is to have an adaptable time dilation between virtual characters. Maybe the speed of the captain can speed up if his input is needed for the intelligence officer to do his work.
A more complex machine is an entire starship. In Star Wars, they have these large imperial starships that have thousands of workers. These workers include: captains, shooters, intelligence officers, engineers, pilots, lieutenants, maintenance workers and so forth. In order to build an autonomous starship, teams of virtual characters are structured hierarchically to give commands. A series of captains might be responsible for the actions of the starship. Each worker is responsible for following orders from their hierarchical chain of command. For example, the shooter follows orders from a first officer, the first officer follows orders from a lieutenant and the lieutenant follows orders from the captains.
In order for the starship to be autonomous, a user is commanding the starship. He will input commands into the AI starship and the captains have to manage and accomplish each command based on military rules. For example, there are things that the user can and canít do. If the user gave the command to attack an innocent planet, the AI of the starship, will do research before this command is executed. The captains might identify the user and determine his rank. Next, they will follow military rules of attacking an innocent planet and what constitute as right and wrong.
In the transformers cartoon, there exist a robot that can transform into 6 machines: a robot, a tank, a lion, a fortress, a gun, and a plane.
What if there is a machine that can change its hardware. For example, a car can change into a plane or a boat or a truck or a forklift. The AI time machine can train any type of machine. If the machine is a car, the AI time machine stores pathways from that car. If the machine is a plane, the AI time machine stores pathways from that plane. If there was a universal machine that changes its hardware, the AI time machine will extract pathways for that present machine. For example, if the machine is a car it will extract pathways in the AI time machine on cars. If the machine is a plane it will extract pathways in the AI time machine on planes.
For each type of machine, itís software program will be different. If the machine is a car, there will be a specific software program used by the team of virtual characters. If the machine is a plane, there will be a specific software program used by the team of virtual characters.
Training of the transforming machine will be done separately. It isnít recommended that one captain be trained on three different types of machine. Maybe one captain can train in two different machines, but not 3 or more. Each captain should be skilled in limited fields.
Maybe the captain can be the same, but the other virtual characters are replaced as the machine transforms. For example, a universal machine can transform into 6 different machines. The captain remains the same regardless of what machine it changes into. However, the other virtual characters are replaced (this method should be used to train the universal machine).
This method is used because the captain should be the same person regardless of the machine type. The change in other virtual characters is because each VC has to be skilled in their field.
Note: each virtual character is using many technologies to do their work. They can use software or electronic devices. A virtual character can use a search engine to access knowledge over the internet or they can use photoshop to make an image sharper. For example, if the user asks the AI car to open an image file and to make the image sharper, the virtual character has to go into the userís computer, use the windows operating system to access the image file, open photoshop, and do work to sharpen the image. Finally, after the work is done, the virtual character has to send the file to the user in a viewable manner.
Hierarchically structured machines working together
A more advance version of the AI machine is to have hierarchically structured machines that work together to accomplish tasks. FIG. 42 is a diagram depicting hierarchically structured military machines. Each machine is fully automated and doesnít require any user input. The president is human and he is the only person that gives commands to the machines. The president is the user.
The lieutenant and the colonel are virtual characters and they are used to coordinate all the groups and give each group tasks to do. The president will give an order, the lieutenant will devise a plan and the colonel will divide tasks and tell each group what they have to do.
Each machine will have communication software that will send and receive inputs/outputs from its superior officer. For example, machine1 will get inputs from group1 and group1 will get inputs from colonelA and colonelA will get inputs from the lieutenant and the lieutenant will get inputs from the president.
The job of the lieutenant is to talk to the president and his cabinets about what must be done. The lieutenant will devise a plan to achieve the goals of the president. These goals are sent to the colonels so they can devise a strategic plan. The plan will be broken up into parts and given to two groups. Each group will break up their tasks into smaller pieces and given to individual AI machines.
Conflicts in the hierarchical chain of command are solved by common knowledge and if each ranking officer wants to question a command, they can go through standard procedures to be heard.
This hierarchical structure can be applied to any business or industry. Hierarchically structured machines can be created for planes or cars. AI towers can be created, each controlled by team of virtual characters, to coordinate landing permission from planes in their air space. In the case of cars, hierarchical structured machines can be created for vehicles to travel on the streets autonomously. Hierarchical traffic towers can be stationed in various areas to coordinate the autonomous vehicles. Individual cars can receive permission and tasks from hubs and these hubs can receive permission and tasks from traffic towers.
In all of the autonomous machines (structured hierarchically), there are human beings who are controlling these machines by giving commands (usually, its verbal commands). For example, in a autonomous traffic system, individual AI cars might have a human passenger who is giving commands. Each individual AI car has to follow instructions from hierarchical traffic towers, in terms of traffic rules and safety rules. If a human passenger wanted to go to destinationA, but the route to destinationA is blocked by a mudslide, the AI car will not follow the commands from the human passenger to go to destinationA.
These hierarchical traffic towers will feed danger and forbidden routes, and specific traffic rules to each AI car. If a politician is in town and uses certain routes, AI cars near the forbidden route will be informed and they have to plan new routes to their destination.
Copyright 2007 (All rights reserved)