Why does AI fall into the infinitely continuing loop?

   Why does AI fall into the infinitely continuing loop? 

Infinite loops, or infinitely continuing loops, mean that the system is stuck operating with the same problem without a reasonable solution. 

All our neurons operate as pairs. When the first neuron sends a message to the receiver. The receiver acknowledges the message. Those receiver neurons send the message. That the message is received. Sometimes something causes a situation. The transmitting neuron sends an acknowledgment back to the receiver. And then. Those neurons start to play a ping-pong, using those neurotransmitters. That means those neurons can fall into a situation where they just surround the same dataset in the form. Called infinity loop. 

But no problem, the outside neurons come and remove the loop. That releases those neurons to operate on a new problem. The outside system must only recognize the infinite loop, and that is quite an easy thing to do. The control system, “judge,” must just see. The system  under the “judge’s” supervision gives the same answer repeatedly. If the answer is the same multiple times, the supervisor sees that those data processor units, like neurons, can be released to operate on another problem. 

The reason why our brains would not fall into a thinking loop is this. We have so many neurons. Our neurons watch each other. And if there is a situation in which some neuron group starts to operate on the same problem repeatedly, and data starts to surround that neuron group, the outside neuron comes and releases those neurons. That means. Outside neuron destroys neurotransmitters that carry the surrounding information. And releases those neurons. 

The upper image introduces the algorithm. As you see, data travels in a circle. But sometimes the algorithm makes mistakes. The mistake can happen when the algorithm uses the wrong dataset. Or sometimes the algorithm simply returns the last mission solution to its beginning point. The system should only send a mark that it's not busy. But sometimes the router will send something else to the return point of the algorithm. When data travels in a computer. 

One wrong value causes a scandal. And in binary computers. It is not accepted. If the value is more than 1. Binary processors can operate only in states one and zero.  The stuck gate causes value 2. The system is stuck. The problem is that the system cannot null itself. And in infinite loops form when data surrounds the system. And it cannot null itself. Without stopping, those algorithms cannot take on the new mission. 

Have you ever tried to make an infinite loop in your mind? The infinite loop, or infinitely continuing loop, is the case where thoughts surround in a circle. The infinite continuum is the case where we think, “I had a dream, that I had a dream...”. This means the list of those internal spaces can continue forever. But the fact is this. Our brains cannot make an infinite loop. Or an infinite circle. Things like pi (3,14...the ratio of a circle's circumference to its diameter) are not infinite loops. They are infinite continuums. 

Outside systems can deny the infinite loops. When the system operates to solve a problem. The outsider judge system. Checks the answers. If the main system always gives the same answers, the system has fallen into an infinite loop. And the outside system orders stuck systems to dismantle the loop. And reboot the system for the next mission.

Brains can create situations that we might think of as a “virtual infinite circle”, but we are never stuck in that thing. And the reason why the AI can be stuck in those processes is this. The algorithms are like circles. But the second thing is that. The AI operates over the binary computer platforms. The AI is an algorithm group that requires the giant computer centers. There are billions of microchips in that system. But there is one weakness. When the AI or large language model, LLM, starts to solve the problem, it has a certain data handling capacity in use. 

Or, the system reserved a certain number of microprocessors for use in that problem. But if the system cannot solve the problem, the AI calls more data handling units to operate on the process. If there are no limits for that process, the system can use its entire capacity. For one problem. That is the thing. That causes the infinite loop. The computer makes its calculations. And then. It makes an error detection. Calculating the same calculations backwards. Another way is to make the error detection. Using two different lines or computers. 

If both computers have the same solutions, that means (probably) there are no errors. There is a possibility. There is a common error that causes a false answer in both computers. But the last one is faster. Than the case. Where the system detects errors. By calculating all calculations backward.  After that, the system can introduce a solution. But sometimes, Something causes situations that the system cannot detect the errors as it should. 

The infinite loop forms in the case that there is no outside actor. Or the system uses its entire processor capacity to solve some problem. In that case, the system has no resources to end the task if the processors are starting to play pin-pong with the solution. That keeps those processors busy, and they have no time to null that process.  

There is a need for an outside microprocessor. That gives an order to stop the action. If the entire system is not reserved. There is a system. That denies the main system from falling into infinite loops. 

Sometimes AI is like a child.

AI is more toddler than terminator. So, we can think that the AI is like a child who controls things like robots, vehicles, rockets, and even weapon systems. When we think about the latest advances in the world of AI and the self-advancing, and self-developing AIs that can create other AIs, we face one big problem. The problem is that if we want to be effective, we must use AI. And most of the AI companies must find their funding from private persons, or customers who are willing to pay licenses. This means that normal users or license owners are not humans. They are companies whose purpose is to maximize the profits of their owners. That means the AI offers a tool that can kick coders out of the workplace. 

And those AI companies must always make new things for their products that their customers are interested in. And buy those licenses. A significant issue in companies is that they are privately owned. They are not controlled by the governments. Private companies do what their stock-owners want them to do. They can outsource their production to countries where there are no laws that control data mining and personal data collection.  Or, they can outsource their production to countries where authorities don’t care about those laws. It is possible to buy “special permissions” to collect that data for products that benefit the defense of the state. 

People ask for some leader who controls AI development. The problem is that AI development happens under private companies. Those companies have no right to make contracts and discuss their work between companies. Those things are limited by cartel directives. That means cartels are things that can deny discussions between companies and set common goals for AI development. The big thing is that the Chinese authorities and intelligence can establish cover-up corporations in some countries and hire AI specialists to work for them. That is the new way to make intelligence work. Those actors can simply hire software specialists who are fired from their work. In China, it is impossible to establish companies without the authorities' support and cooperation. If authorities have no access to private company servers that means the company stops its work. 

And then they can use people who have EU passports and citizenships for actors who make that company for them. Those people who work in PLA (People’s Liberation Army) intelligence have methods to persuade people who left China to cooperate with them. They can say that family members who will enter military service will face problems in that army. And for cooperation, those people should establish companies in some countries. And then the entire work that those people will make will be copied to servers that are located in Hong Kong and Beijing. That is one way to do business that benefits those Eastern actors. When people use Chinese AI they also deliver information to that country. In those Eastern countries, the security laws don’t limit the authority's access to people's personal data. And that means those laws obligate only private actors. 

And then the dawn of machine learning.

 And then the dawn of machine learning.

Image: Pinterest

Machine learning or autonomously learning machines are the newest and the most effective versions of artificial intelligence. Machine learning means that the machine can autonomously increase the data mass, sort the data and make connections between databases. That ability is making machine learning someway unpredictable. And that kind of thing makes the robot multi-use systems that can do the same things as humans. 

The reflex robot is a very fast-reacting machine. The limited operational field guarantees. that there is not needed a very large number of databases. And that means the system must not search the right database very often. That makes it very fast. But if it goes out from its field it will be helpless. 

When we are thinking of robots that can make only one thing like playing tennis they can react very fast in every situation. That is connected with tennis. There is a limited number of databases. And that means the robot is acting very fast. 

When a robot or AI makes the decision it systematically searches every single database. And if there are matching details to observed action. That activates the database or command series that is stored in the database. But the thing that makes this type of computer program very complicated is that when the number of stored actions is increased the system will slow.  

If we want to make a robot that can make multiple actions. That thing requires multiple databases. And searching for the match for the situation in every database takes a certain time. So complicated actions require complicated database structures. Compiling complex databases takes time because there are limits in every computer. And in the case of a street operating robot, the system compiles data that its sensors are transmitting to its computers. 

So the conditions that this kind of system must handle might involve unexpected variables like fog or rain. And for those cases, the system needs fuzzy logic for solving problems. In that case, only the frames of the cases are stored in databases by the system creators. And that system is compiling those frames with the data sent from the sensors. 

The waiter robot can be used, as an example of machine learning.

A good example of a learning machine is the waiter robot that is learning the customer's wishes. The robot will store the face of the customer to its memory. When it asks does the customer wants coffee or tea? Then the robot will ask "anything else". And in that case, the robot can introduce the menu. 

And then the customer can make an order. There are certain parameters in the algorithm. Those are stored in the waiter-robots memory. The robot is of course storing that data in the database. The reason for that is simple. The crew requires that information that they can make the right things for the customer. But that data can use to calculate also how many items the average customer makes after a question "anything else"? 

The robot can also store the face in the database that it can calculate how often that person visits the cafeteria. Then that robot can simply store the orders below the customer's face. And it learns how often a person orders something. If some customer is ordering some certain products always. The robot can send the pre-order to the kitchen. That they can get a certain type of order. When some customers will visit often and order all the time same thing, the robot can start to say "do you want the same as usual? For that thing the system requires parameters how often in a certain time is "often"? That was an example of the learning system. 

The biggest of the wild card effects is the human factor.

   

 The biggest of the wild card effects is the human factor.

The human factor is the thing that makes predictions very difficult. 

The thing is that we can predict how the large gas masses behave in the universe. We can make algorithms that predict if there is some kind of exoplanet or civilization near some star. But we cannot predict how the human race and our society are behaving. The thing called: "the human factor" is one of the biggest of the so-called "wild card effects". 

The wild card means the surprising and non-predictable thing that has a large-scale influence on the environment. And the power of human nature is visible in the cases like COVID-19 vaccination programs. It is predicted that authorities can remove limitations caused by pandemics when the level of vaccinated people reaches 80%.  But the percentual number of the vaccinated people would rise hardly over 80%. 

And that thing is one of the examples of the human factor. All people would not take vaccines even if against the sense. These kinds of human factors are causing problems for making predictions for the future. When people are against something, they just want to follow their road even if that thing is against their benefit. 

Social media is one of the most impressive examples of wild cards. 

The cases that the human factor would terminate all predictions are examples of why computers cannot predict the future. When the AI is making the prediction it will collect all data about the cases, and then it will see are the similar signals visible in nature. But when we are thinking about technology advantages and development process that thing is out of control. 

There is one wild card that is totally out of control and that is social media. That thing is hard to control. And then another and even larger size thing is the internet. Nobody controls what people are doing at their part-time. And the information of the programming is easy to find from the net. So every person in the world can begin their artificial intelligence project. 

There is nobody who leads the R&D processes that are basing the open-source. Those kinds of things are self-controlling. And that thing means that no leader can stop those processes. When a developer makes something the response of the product is in the developer's seat. 

The same developers that can make the algorithm that can search data from the internet can also make spy programs and computer viruses. And that is the problem. But those programs can also use for searching drug dealers and slave traders. So are the surveillance programs so bad? Or is in those programs something good? 

Even if own projects by using the computers that belong to a company are prohibited everybody can buy their computer for that kind of thing. And the recommendation is that for own business is used own computers. 

The wild data causes turbulence in the environment. And the turbulence can cause. That even the strongest signals are covering. This thing makes traditional leadership very hard. The movement from the upper-controlled process will turn to a self-controlled process. And this thing makes the next challenge for society. 

https://en.wikipedia.org/wiki/Black_swan_theory

https://en.wikipedia.org/wiki/Wild_card_(foresight)

https://networkedinternet.blogspot.com/