The AI requires powerful microchips.

"By directly leveraging light signals received from distributed acoustic sensing systems, the proposed photonic neural network architecture provides massive gains in accuracy and efficiency over conventional electronic computations. Credit: N. Zou (Nanjing University), edited

Imagine fiber optic cables acting as vast sensor networks, detecting vibrations for everything from earthquake warnings to railway monitoring. The challenge? Processing the enormous data flow in real-time."(ScitechDaily, This AI Uses Light Instead of Electricity and It’s Mind-Blowingly Fast)

"Traditional electronic computing struggles, but researchers have merged machine learning with photonic neural networks, using light instead of electricity to process distributed acoustic sensing data at incredible speeds." (ScitechDaily, This AI Uses Light Instead of Electricity and It’s Mind-Blowingly Fast)

AI is an algorithm and physical platform combination just like all other computer solutions. Running computer programs is impossible without microchips. The problem with complicated algorithms is that they require so much computer power that the system requires truck-sized systems. And those systems require lots of energy. The system can use morphing neural networks that make the algorithms lighter for individual computers. But the paradox is this. 

The more powerful computers make those morphing neural networks more powerful. That is important for data security. 

More powerful computers can break codes that weaker computers make. And that causes the weapon race in microchip research. Fast microchips make those systems more effective. Another paradox is that effective systems make it possible to run more complex algorithms. So advances in AI and algorithms follow the same route as other computer programs. More effective microchips allow developers to develop more complicated programs that require more calculation power. 

That kind of system uses lots of power. In electricity-based microchips, resistance causes very big problems. Resistance causes vibration and data loss. The answer can be superconducting computers but those systems require very big coolers. Or material that can be superconducting at room temperature. It's possible to make room-temperature superconductors. Using very high pressure. 

But if the pressure system's shell is broken. Pressure causes terrible danger. In pressure superconducting systems. Pressure anchors those particles in their places. 

Maybe, nanotechnical, small-size tubes where the nano-diameter wire travels can allow researchers to create a safe pressure system. That system removes vibrations from the wire that goes inside it. 

Another way to make the system is to use the photons. The photonic neural network where light replaces electricity can sense if something touches the light fiber or cuts the laser ray. Outside effects. Pressure causes curves in laser ray trajectories if they travel in optic fiber. And the sensor sees that thing. That improves data security. The system can also sense things like seismic wires and changes in electromagnetic fields. 

The binary system can use photons in two ways. The different light wavelengths like blue, and red can be zero and one. The system can determine that a certain lux level is one.

Below a certain lux level is zero. The photonic computer can use CCD chips or photovoltaic cells as receivers. The photonic microchip keeps the temperature in the system lower. 

"Distributed Acoustic Sensing (DAS) is an advanced technology used for infrastructure monitoring. It detects tiny vibrations along fiber optic cables that can stretch for tens of kilometers. DAS has become essential for applications like earthquake detection, oil exploration, railway monitoring, and submarine cable surveillance. However, these systems generate vast amounts of data, creating a major challenge: processing it quickly enough for real-time use. Without rapid data processing, DAS loses effectiveness in scenarios where immediate responses are crucial."(ScitechDaily, This AI Uses Light Instead of Electricity and It’s Mind-Blowingly Fast)

"To tackle this, researchers have turned to machine learning, particularly neural networks, as a way to speed up DAS data processing. While traditional electronic computing with CPUs and GPUs has greatly improved over time, it still struggles with limitations in speed and energy efficiency. Photonic neural networks, computing systems that use light instead of electricity, offer a breakthrough solution. They have the potential to process data far faster while using significantly less power. However, integrating photonic computing with DAS has proven difficult, mainly due to the complexity of DAS data and the need for precise signal processing." (ScitechDaily, This AI Uses Light Instead of Electricity and It’s Mind-Blowingly Fast)

https://scitechdaily.com/this-ai-uses-light-instead-of-electricity-and-its-mind-blowingly-fast/

From weather broadcasts to social algorithms.

"Aardvark Weather is an AI-driven system that dramatically reduces the time and computing power needed for accurate forecasts. Unlike hybrid models, it replaces the entire forecasting pipeline with machine learning, outperforming traditional systems using a fraction of the data. Credit: SciTechDaily.com" (ScitechDaily, Scientists Just Built an AI That Predicts Weather in Minutes – And It’s Beating the Best)

Researchers created an algorithm that predicts whether to beat the best.

The AI can handle thousands or even billions of objects at the same time. That makes it the ultimate tool for predicting things like solar storms and weather. The system can collect databases from certain things. Then it compiles observations about the thing that it should predict. 

If the AI should predict solar storms. It must collect information about solar activity like changes in its luminosity, and whirls and particle flow that predict solar storms. 

In the same way, weather prediction requires information about the airflows, temperature, and all other things that we can connect to storms and other types of weather.  The new AI- or algorithm makes it possible to create models with a very high accuracy. Those high-accurate AI-based systems can make models about magnetic fields, air flows, and many other things that can help to make the material, and other types of research. 

And then what about "Psychohistory"? That is a mathematical model of the human behavior. The system could predict a large human group's behavior. 

But maybe. Someday that hypothetical system could turn so accurate that it can predict a single person's behavior. 

The requirement for that kind of algorithm is that the system can collect trusted and confirmed data freely. 

Isaac Asimov introduced "Psychohistory" in his SciFi novel "Foundation". 

We can think. That Psychohistory is the mathematical model of social behavior. The idea is that similar algorithms that are used to predict weather can be used to predict human behavior. In this model. The system can predict the large human group's behavior in certain situations.

The system uses the Boltzmann constant and some other types of formulas to create models about human group's behavior. The system compiles things that it sees with data that it found from the net. The prediction forms when data. The system collects from sources, like historical documents is compiled with information that the sensors send to the large language model LLM. 

Then it can compile that behavior with data that it collects from the news and environment. The idea is that the system acts like an astronomical or weather-predicting AI that can calculate large gas mass behavior. The system cannot predict the single gas atoms place in the universe. But it can predict large gas masses like galactic supergroup behavior. 

But then we must realize that our knowledge of neural networks is higher than in Asimov's time. The system can collect data about neural connections in the human nervous system. Then it can compile that thing with people's nervous system that the system knows. The idea is that a person with a certain social background should behave in similar ways. If those social backgrounds are identical that means the nervous connections must be identical. 

But the fact is that. The system must also follow the social behavior that it can make profiles of people. That profile tells about the person's way of behaving in certain situations. So the system must collect data matrix about people. 

Then it must compile that data matrix with people who it observes. The problem with this type of behavior prediction cannot be trusted it must have the full and confirmed dataset that it can compile. Social algorithms are new things that can predict things like riots.  

https://scitechdaily.com/scientists-just-built-an-ai-that-predicts-weather-in-minutes-and-its-beating-the-best/

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

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

Quantum algorithm vs. binary algorithms.

Image: Quanta Magazine

The biggest difference between quantum and regular algorithms is that in quantum algorithms or in quantum networks data is connected to physical items. Sometimes people misunderstand the terms algorithms and networks. Networks are things. That transport data. 

Algorithms are programs that compute data. The term quantum algorithm can mean the calculation program that handles very long decimal numbers. That can involve tens of millions of numbers. The biggest problem with quantum computers is temperature. 

Those systems require so powerful coolers that they are factory-size systems. The compact-size quantum computers are on the door. But that means they are room-size monsters. Those systems require supercomputers that can operate their operational systems. A good and powerful alternative for supercomputers is the morphing neural network. 

That makes it possible that the qubit's values can be zero and one. This allows the system can make many operations at the same time. The binary network's values can be 0 or 1. The things called morphing neural networks can also make binary computers make many things at the same time. The system shares missions between different computers. 

The question is what is the most powerful computer in the world? The answer is interesting. The most powerful or fastest computer depends on the formula that the system should make. If we want to calculate simple calculations like 1+1 the most effective system is the credit-card size calculator. 

The thing. That decreases the quantum computer's power is that making the quantum entanglement there data travel in those systems takes time. That makes the quantum computer slower in the simple algorithms. Quantum computers are most powerful when the system must handle multiple variables and complicated formulas. 

The thing is that the quantum computer is not the best tool in the world. If we want to make simple calculations. 

There are calculations that the that the normal computer makes the universe's entire lifetime. 

And the quantum system makes that thing in minutes.

But then the morphing neural networks are tools that can make many things faster than the regular computer. The morphing neural network is a group of binary computers. That means they can share complicated series. With each other. 

The AI-based binary systems can jump over the zero points of Riemann's conjecture. So AI is the game changer in all of those things. 

Another thing is that. The new high-power binary systems are not like traditional binary systems. 

There the data goes in different wires. And that solves the "zero" problem. 

The zero problem means that the system must separate breaks in the data row from zero. 

And the system must also know. If the system is switched off. 

The system must also separate two zeros from each other. That's why there can be a different wire that shows when power is on and off. And data can travel in different wires.  The system must give serial numbers to ones and zeros. If they travel through different wires. It can sort them into the right order. 

Or there can be two low states in one wire. That allows the system to separate zeros from breaks. Low states like  3-5V are zeros. And 3-0 V is the break. And it accelerates the system's speed. 

And then the final question: which system breaks the RSA algorithm fastest? The morphing neural network or quantum computer? The RSA encryption uses Riemann zeta function. The morphing neural network can begin the code-breaking in the many points in the number series that Riemann zeta function creates. The system can use a number row that was created before. 

The code-breaking operation is not the same as calculating more numbers to the Riemann's series. Or a series of binary numbers. The idea of Riemann zeta function is that the formula generates only binary numbers. That thing means that it should protect the data. 

But if there are zero points that formula can generate also other than binary numbers. And the AI-based encryption systems can jump over those points. The Riemann's series can be programmed to the morphing neural network. Each computer takes the bite or sequence of that number row under the handle. 

https://www.quantamagazine.org/quantum-speedup-found-for-huge-class-of-hard-problems-20250317/

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

The small language models are reflex algorithms.

Researchers are interested in small language models. The reason for that is simple. Large language models include thousands of billions of parameters. Those parameters require lots of computer capacity. And those computers use lots of electricity. Training for the LLM costs billions of dollars in computer and electric bills. And the biggest problem is that the LLM requires a big data center. 

Normal companies have no money or other resources to run the LLM on their own servers  And in the ICT world those tools can cause data security problems. The LLM that runs on the Microsoft or Open AI servers runs on machines. That is under the potential competitor's control. The LLM is a good tool but it requires lots of power. 

If we want to operate robots independently using the LLM. We must be sure. The robot has an internet socket connected to the central computer. In that model, the robot sends orders first to the computer center. There the LLM transforms those orders into actions for the robot. That system is useless if something disturbs it. 

The robot cannot keep the connection in electromagnetic fields. Robots are planned to be used in high-risk environments like nuclear accidents and military work. The remote control is easy to jam, and that's why researchers in military and civil fields search for systems that can be compact and locally operated. 

One of the problems with LLM is this. Those systems search data from the entire internet. That makes them good tools for making things like doctoral theses. But if the AI-controlled robot uses that model it requires lots of energy and those things are slow. 

The robot requires the reflex algorithm. The RISC systems have a limited number of databases. That makes them compact and effective. The RISC system called the small language model, SLM is the tool that can make robots safer. And allows them to operate independently. 

Those reflex command bases can involve responses to things like: "Would you step away from the door". In the same way, the jet fighter cannot ask for advice from the computer centers if it sees an incoming missile. 

The robot must not make contact with the computer center. Every time, when it must react to some everyday things. When a robot hears something it must realize that it must not react to everything 

So, if we want to create an AI that can operate in a complicated environment we must modify the LLMs and create a lightweight version. The lightweight LLM or small language model SLM has only a couple of million or even less than a million algorithms. Sometimes those SLMs called RISC-language models. RISC (Reduced instruction set computer) systems are like pocket calculators. They might be more limited than large systems. But they are fast-reacting and they do their job very fast. 

Those lightweight language models can run on regular servers. They react very fast because they have only limited action libraries. The small language model can be installed on the aircraft's computer. And it can act as an assistant for the pilot. 

https://www.quantamagazine.org/why-do-researchers-care-about-small-language-models-20250310/

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

The electronic warfare in Ukraine shows how important. Is to develop GPS-independent homing systems.

The ground-launched small-diameter bomb, GLSDB, is a good weapon on paper. But its Achilles heel is the GPS homing system. That system is quite easy to jam if the enemy knows the frequencies that the GPS uses. I think. That civilian GPS was in weapons delivered to Ukraine. 

In the war scene, one of the GPS systems.  In enemy hands, GPS can cause catastrophe. Because that gives those systems radio frequencies to enemy ECM systems. Another thing. What can cause failure is this. Maybe Ukrainian troops tried to shoot them outside the allowed area. That is the Ukrainian territory. 

And the safety system destroyed those weapons. But the ability to jam the satellite transmissions is the thing. That should be noticed. The GLSDB types of weapons require new and GPS-independent navigation systems. 

The combination of the gyroscope and optical AI-based seekers can make that system GPS-independent. In real life, a missile must know the shooting point. And then it must know the direction where it must fly. After that, the missile must recognize the target. 

The AI can make that kind of weapon a new way to survive and travel through the air defense. The AI that can recognize missiles that shoot against the incoming missile can make the missile wobble. That decreases the AA missile's ability to point the missile. Weapon research is the race between weapons and counter-weapons. In the Gulf War in the year 2003, the GPS-guided bombs were the ultimate tools. But the thing that weapon researchers should predict is that.

After the Gulf War information about that bomb was public. Russians knew the GPS navigation system's role in the battlefield. And its ability to aim bombs.

And they had 20 years to develop a counter weapon against the GPS-guided bomb. And after the Gulf War, everybody knew that weapon. But for some reason, there was no navigation system. That could replace the GPS and other satellite navigation systems if they are under jamming. For some reason, the developers didn't realize that the missiles and bombs that used only the GPS were vulnerable. They should predict that the GPS is quite easy to lock if the enemy knows its frequencies. 

And when we think about things like Iranian drones or slow cruise missiles that attack against Ukrainian targets we must realize that Ukrainians are lucky. Those slow drones are easy to pick if they are located. 

The AI that recognizes the target and then makes the missile make the evasion movements can make those slow things more deadly. In the worst case, the AI is the thing that allows the drones to communicate with each other. That means that when one drone sees the AA station it attacks against it.  The drone swarms act as an entirety. The fact is that the AI algorithms and the new electronics can make the old, large-size missiles deadly tools. 

It's lucky. That Russians didn't change cluster warheads to their large AA missiles. Those things can be more deadly than the warheads that they had when those missiles shot against Ukrainian positions. 

https://www.eurasiantimes.com/atacms-delayed-but-glsdb-is-headed-to-ukraine/

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

The Chinese AI won human pilot in air combat.

(SCMP, China’s red-eye AI just killed human pilots’ last hope to win in air combat: researchers)

When we say that humans beat AI, we can always ask, can the average person make that thing? Can the average chess player win AI if the world champion can? Or can an average pilot beat the AI in air combat? If the best pilots can? 

Chinese authorities are interested in AI's military applications. One of them is the advanced autopilot. That can operate as air combat duties. Advanced AI can beat humans in air combat. And that fits into the Chinese state policy. The AI and LLMs are tools that can revolutionize air combat. 

And this is one of the reasons why those things are under development. The main role of the Chinese military is political trust. 

Every time the pilot is alone in a plane. It's always possible that. The person starts to rebel and attack against the Communist Party HQ. This is one of the reasons why drone technology is under development. 

Being alone in the cockpit allows pilots to rebel. 

The kill switch can solve that problem.

It's always possible that hackers get an activation code for that switch. In that case, hackers can destroy multiple aircraft. 

If a pilot flies the mission from the ground. Next to military police. 

This authority can arrest the pilot. If something goes wrong in the mission. 

Also, things like the Tiananmen case caused suspicion that the security troops might not want to operate as government orders. 

(SCMP, China’s red-eye AI just killed human pilots’ last hope to win in air combat: researchers)

In the news, The Pentagon connected to robot armies. 

However, trust is the reason why China and its government research robotics. And especially its military applications. 

And we can say that. China is the top government in AI and its military application research. 

The drone aircraft can operated remotely. Or it can operate independently. 

The remotely operated drones are quite easy to jam. That makes them vulnerable. But if a robot fighter operates independently. The last ones are the result of microchip advances. The AI can give new abilities to regular drones and cruise missiles. The drone can make evasive maneuvers. The thing that makes the AI-controlled robot systems superior is that the AI doesn't matter things like G-force. And it can make maneuvers, fatal to humans. 

Without human control, the opponent's last hope is that the control algorithm will not work as it should. The automatic bombardment systems probably already exist. The aircraft or drone must only know the drop point for the bomb. 

And large-size drones can also drop GPS or optically homing bombs. The GPS is an effective system, but it's easy to jam. This is the reason why the replacement for that system is under development. 

The homing system uses an image-based homing system as the Javelin missile. The system has the image of the target. And then the bomb can fly into it.

The system can use a hybrid system. A combination of inertial navigation, a modified TERCOM system, or GPS for that thing. The inertial (gyroscope) is immune against ECM. 

The aircraft or drone can fly to the drop point even if ECM jams the GPS. Using that navigation. Then, the system can drop the bomb. 

That flies into the target first using the inertial. And optical seeker. 

The system flies to the target. Searching for structures that are similar to that stored in its memories. 

When its camera sees a target. A weapon starts to glide against its target. That system is made to operate. Even, if the ECM cuts satellite connections. 

 https://www.scmp.com/news/china/science/article/3300557/chinas-red-eye-ai-just-killed-human-pilots-last-hope-win-air-combat

Humanoid robots are coming to homes.

Who likes cleaning? And who would outsource that thing to some butler? The problem is that butler must get a salary from their jobs. Sometimes, those people have data security problems because they can tell things about their employers. The answer to those problems can be the housekeeping robot. The housekeeping robot can make things like food, and clean house.  

If that kind of robot has the right database, it can also make repairs. Those things are productive. The new flexible robot can make almost everything that humans can make. The robot needs an internet socket and the right databases to get new skills. The humanoid robot is a good tool because it can use the same tools as humans. The humanoid robot is a physical extension of the large language model, LLM. That robot increases the LLM's ability to get information. It can also make the LLM able to interact with the physical world and make physical works. 

Or maybe quite soon this kind of robot can build entire houses. By using the right modules the robot can make almost everything that its owner asks. The human-shape robot can also operate as a security officer. It can also interact with humans and the internet. That thing means that the robot can have a projector that allows the user to use it like a walking phone booth or internet socket. 

Those robots can share their data using satellite communication. So if they have the right power sources like nuclear batteries. They can operate for over 100 years. The problem with nuclear batteries is that in the wrong hands, they are dangerous. 

Of course, robots are also at risk. They can operate as soldiers and assassins. There is always a risk that somebody cracks the robot's security code. And then programs it to operate as an assassin. When we think about things like authoritarian governments. Those home-assistant robots can also watch things that people do in their homes. When we create something new like human-shaped robots we make also multipurpose tools. Those tools have the ability to change the world. 

https://www.freethink.com/artificial-intelligence/humanoid-1x