The AI involves billions of algorithms.

"New research from Johns Hopkins University shows that certain biologically inspired AI architectures can mimic human brain activity even before training on data, challenging long-held assumptions about how AI must learn. Credit: Stock" (ScitechDaily, Johns Hopkins Study Challenges Billion-Dollar AI Models)

New research challenges AI and its learning process. When we try to train extremely large AI’s and their language models, we must understand that this process requires. Lots of work. When we think about things that we make every day, we must realise that. That we learned all the things. That we do. When we try to teach an algorithm, we face one big challenge. The algorithm works like this. The sensor brings data to the memory loop. 

Which compares that data with the database. And if there is a match. That memory loop will activate some action. The problem is that the action that is described to the system is the key. That activates a process. That is connected to the database. And there is always a possibility that the thing. What the computer sees. Doesn’t have 100% match. 

 With the thing. That is described in the database. The single memory loop. That is connected to a certain mission. It is quite easy to train. But when we try to train robots. That can operate. In an open environment, we are in trouble. There must be a memory loop. For every action that the system makes. The thing is that. The memory loop must stop. If the system wants to drive new data into the loop. A quantum computer that operates in multiple states. Makes that system more effective. The system can drive data to different states, and the loop must not stop. 

That makes things like fuzzy logic more effective. Fuzzy logic means a series of precise logical responses. If we compare that thing with the case. When we move our hand, the fuzzy logic means that there is an algorithm for effectively determining. The position of the hand. 

The memory loop or algorithm is the circle of commands that surrounds all the time. If we make a system that goes shopping for us. We must teach every action that the system needs in that everyday journey to the computer. When we think about a shopping trip and compare. It's with a fighter mission, the jet fighter seems more difficult. Because it's not an everyday tool. 

Image 2) Algorithm. 

The thing. What makes the human brain so effective is that it has billions of neurons. A single neuron is not very impressive. It’s a pack of memory and axons, or connections. The human brain uses billions of neurons. At the same time. The central nervous core. Drives data to billions of neurons in the same moment. If neurons don’t find an answer, they call more neurons to the mission. During that process, neurons search for matching memory units within them. And those memory units activate reactions. 

During that process. The area and number of neurons. Those involved actions grow. Until the action is handled in the entire cerebral hemisphere. If both cerebral hemispheres in the cerebrum agree. Or they get the same answer. They send the data to the other nervous system. If those cerebrums disagree, they call the cerebellum. That makes a decision on which one is right. The cerebellum's mission. It is to cut the endless memory loops. 

And when the brains send information to the nervous system. It waits for a report. That the job is done. The report travels to another neuron. That neuron is called a mirror neuron. The purpose of that is to deny. The information collimation in nervous tracks. 

The brain mimic system has two parts. The part that begins the action. And the part. That tells us that the system made the job. In human brains, there are neuron pairs that make those things. The third neuron is needed for cases. The system is stuck in an endless loop. When the first neuron sends a signal, the system does the job. And then it sends that information. Into another neuron, which tells that the job is done. In data centres, those neurons are data processing units. Those units can be multicore processors. In those systems, cores are in pairs, which makes the memory loops travel between them. 

That with things like robot fighters. The robot that goes to the shop. Makes many more actions during missions than a jet fighter. Normally, we don’t think about that thing. Normally, we learn everyday actions during life. But think about the situation. We will drop from space to Earth, and we must learn everything. That we must do. We need somebody. To tell things how to dress, what a means shop, and what vegetables are. Normally, we know those things because we learn those things “automatically”. But what if we grow in an environment? Where are there no vegetables, shops, or money? 

What if we must tell every articulation?  In our body, what must it do? When we want to raise our right foot, we just do that thing. But in the cases of computers, the system must know. What motors must it use? And where it finds those engines. The engine can have a microchip that tells the main computer. That’s the servo that operates the left knee. Then. The operational algorithm tells what the servo engine must do in certain situations. The local system shares responsibility with the main computer. And that makes the main computer’s operations lighter. 

But when we make programs, we must program every action separately. In the second image, you can see the algorithm structure. One algorithm is not very hard. But in complex AI systems. There are billions of those algorithms. There is a missing loop. That means the system should run those algorithms. All the time. The AI is a multilevel construction. The brain mimics systems. Drive data into as many memory loops as it can. In modern AI systems. The system uses other AIs to teach it. 

https://datascience.101workbook.org/05-programming/01-algorithm/01-basics-of-algorithm-structure/#gsc.tab=0

https://scitechdaily.com/johns-hopkins-study-challenges-billion-dollar-ai-models/