The size matters in cosmological models.

“Two images from the Quijote simulations used in this study. The panels show the same region of the Universe, but in different cosmological models. The top image corresponds. To the standard ΛCDM, adiabatic cold dark matter model, while the bottom image shows a universe with massive neutrinos and modified gravity. “(ScitechDaily, AI Learned the Rules of the Universe and That Became a Problem)

The differences are subtle, but they reveal how changes in the underlying physics can affect the formation and distribution of cosmic structures. Credit: Francisco Villaescusa-Navarro (ScitechDaily, AI Learned the Rules of the Universe and That Became a Problem)

The term ACDM can also mean : the associated critical data model. That is the critical tool, when the sensor. It transmits information to the AI. 

AI can help cosmologists, but it can also become a problem. 

The method researchers call transferable learning can help them develop new models in cosmology and many other things. The term transferable learning. Means when the system learns something. It can apply. That learned thing. To other similar cases. So, when AI sees similar curves in some other cases. It can use things that it has already learned. To that other problem. This means that. The researchers must not always. Begin the training process. From the beginning. 

The AI can search for similarities for the new thing in its memory. And if there is a match. That thing means that the AI. It can use that model for reaction. This should make AI more effective. The problem is this. The AI selects its sources using statistics. And that can make it hard to bring new data for the AI. Old research. They are very often-used sources. If somewhere is the new data. Before, nobody used the new data as a source. Old data dominates search engines. The AI is an excellent tool. When it must collect and analyse data from the galaxy movements. 

But in cases like supermassive neutrons, the AI is in trouble. The AI is the best in business. When it must analyze precise information. Things like galaxy clusters and their movements are precise information. But in cases like supermassive neutrinos. The AI is not very good. At things where it must create models for new physics. When AI must observe phenomena. It can interpret them as the same. Even if they are different. Or in the cases. 

There are some observations. Objects’ temperatures change. The AI might not know that the object’s temperature can change virtually. Because if something travels between the telescope and the object. That means. that the brightness or temperature. That reaches the observer changes. The AI might not notice things like clouds. In the Earth's atmosphere. Or other surprises when it observes some targets like Cepheid variables. If the system doesn’t know about that thing. It can recognize the Cepheid variable as a new star. If it doesn’t know that the star is a Cepheid. 

When AI tries to analyze a certain point. That thing is very hard to do. But when AI must analyze. A very large entirety. The AI becomes more effective. The AI sees things. Like movements of galaxy clusters. And it can make. An analysis of the changes in those movements. We can use fuzzy logic to analyze how the star clusters move in the galaxy. But then we face a problem. If we try to predict. The movement of the galaxy. In its supercluster. That is hard. 

We must know the entire system to make. A complete analysis with high precision. 

The problem is in perspective. The thing that seems large on Earth. Seems very small in the scale of the Sun. And the sun seems very small in the scale of the galaxy. When the scale of the system turns bigger. The forces in the system are also stronger. In big systems. The phenomenon scale is larger. But they affect more slowly. From our perspective. The forces that travel between galaxies take millions of years to reach other galaxies. The distance between the Andromeda galaxy and the Milky Way. It is 2.6 million ly. So light travels 2,6 million years from that galaxy to the Milky Way. And that means that any force traveling between those galaxies needs 2,6 million years for that trip. 

When we try to create a model. Of how one small sand bite behaves in a river. We must know many things. Like changes in the forces that affect the sand bite. But if we want to predict how the sand bottom behaves in the river. We can make that calculation very easily. When we think about galaxies. Stars are like sand bites on the bottom. 

One star’s behavior is hard to predict. But the entirety is quite easy to  calculate. And then we can go to bigger systems. In galactic superclusters, the galaxy is like sandbite on the bottom of the river. The force that affects the entire galaxy. Must be much harder than the force that affects sandbite. But millions of galaxies. They send. A very much. Energy. Many sudden things can happen in the galactic superclusters. Those events might not. Seem.

Like a very sudden thing. But an eruption in the core of the galaxy can start in milliseconds. Shockwave travels across the galaxy at the speed of light. So, if the star is at a distance. Of two light-years from the eruption source. The shockwave of radiation. It travels to that star. So, if Sagittarius A erupts violently in the core of our galaxy, the Milky Way. The radiation travels to Earth 26.000 years. The distance between Earth and that supermassive black hole. It’s 26.000 ly. The material, or plasma shockwaves, travel far behind that radiation shockwave. And the distance between plasma and wave movement increases all the time. 

 But. If things like supermassive black holes are in the trajectory. That makes them collide. That thing is very hard to change. When we face things like galactic superclusters. Things that happen on that scale seem very slow. But forces that put galaxies. To turn their trajectories into travel. At the speed of light. The force. That affects things. Like, turn their trajectories. Must affect a certain time with a certain force. 

If we want to create an AI that analyzes galactic clusters star by star. We cannot make that thing. In the galactic scale, it suddenly happens. Violent eruptions. Those eruptions can break the entire model. In the scale of superclusters, events like supernovas don’t have enough force to affect the macrosystem. But a supernova could destroy things like dwarf galaxies. But if the supernova explosion happens in dense star clusters. That shockwave. Can. Launch other supernova explosions. 

https://scitechdaily.com/ai-learned-the-rules-of-the-universe-and-that-became-a-problem/

https://en.wikipedia.org/wiki/Lambda-CDM_model

https://en.wikipedia.org/wiki/Sagittarius_A*

The new era of the air force is here.

Above: YQF-44

The main problem is to find. The most effective mix of manned and unmanned systems. 

The YQF-44 is the new type of drone. That system uses AI-based air combat capability. These kinds of systems can operate independently or with manned systems. Basically. An automated bombing system is very easy. To make. The system can use a radio beam. And the drone flies in it. Another radio beam. That crosses the path rays. Launches bombs at the right point. More advanced AI-controlled systems. It can use a Terrain Contour Matching (TERCOM) technology to determine its location. And then that system can determine its location using terrain photos. When the image. Its camera sends. Matches with an image. When that is stored in computer memory, the system can launch bombs. 

Above: X-62A with chase plane. Lockheed Martin (Internet)

The QF-16 is basically an aerial target. The Shield Corporation, along with Lockheed-Martin (General Dynamix), transformed that thing into a system.  That can go. Into air combat duties. The official name of the F-16 robot is the X-62A “Vista”. 

The more advanced algorithms allow the drone to operate in any mission that a human pilot can. It's possible to assemble the AI into any aircraft on Earth. That can make it possible. To use old jet fighters. Like robot versions of the F/A-18 (QF-18) and A-10 “Thunderbolt” (QA-10) as attacking decoys. The old jet fighters. Serve as air combat practice. Target missions. There are various old jet fighters operating in that role. The newest planes are F-16 robot versions, QF-16. Maybe. Also, the old A-10 and F/A-18. Are serving in that role. There is a possibility of installing the AI into those old aircraft. 

That can attack enemy targets using weapons and kamikaze tactics to make those missions. Those systems can fly. With other systems, they can force the enemy to shoot. And uncover their locations. And those systems can also operate as patrol systems. That can attack enemies. Theoretically. It is quite easy to transform the target drone into an attack aircraft. That uses AI-operated technology. To recognize their targets and attack them. 

X-BAT is a drone. That rewrites rules on the battlefield.

Shield AI is promising a fully autonomous fighter jet with its X-BAT. Shield AI

© Shield AI

-Shield AI unveiled what it called a fully autonomous multirole fighter jet.

-It uses Hivemind, the AI pilot it previously used to pilot an F-16 through a dogfight.

-The company says the jet can take off without a runway, from container ships and remote islands.

(MSN.com)

X-BAT is an autonomous, large-sized drone. That unmanned combat aerial vehicle (UCAV)- or killer drone is also independent of runways. That drone uses. Vertical take-off and landing (VTOL) technology.  The fully autonomous drone. That doesn’t require runways is the new tool that the military requires. The drone itself is developed by Shield Corporation. This created a robot version of the F-16 jet fighter. That AI-controlled jet fighter. Achieved success in air-combat simulations.  And that made Shield Corporation. Continue its work to create autonomous combat systems. The AI-controlled attack aircraft can make risky missions. 

And another thing is that. The new, effective AI-controlled systems. Allow the robot fighter to engage in air combat against human pilots and other drones. The difference between manned and unmanned aircraft. Is this. Unmanned aircraft can also. Make kamikaze attacks. If it sees a target. That value equals the loss of the aircraft. The unmanned combat aircraft can also carry nuclear weapons internally. And their air-combat capacity makes those next-generation missiles more powerful than they were before. 

The cruise missile. Those that can shoot down other aircraft can also be equipped with direct energy weapons. Like lasers and EMP weapons. Those kinds of weapons don’t need very large systems, and they can guarantee that future nuclear cruise missiles can reach their targets. Or they can escort manned and unmanned systems. It’s not sure. Does the USAF or the NAVY choose X-BAT? But that technology is also useful in other UCAV systems. 

https://www.airforce-technology.com/projects/qf-16-full-scale-aerial-target/

https://www.lockheedmartin.com/en-us/products/x-62a-vista.html

https://www.msn.com/en-us/technology/artificial-intelligence/a-new-autonomous-fighter-jet-just-broke-cover-it-s-powered-by-the-same-ai-brain-that-flew-an-f-16-through-a-dogfight/ar-AA1OXx15?ocid=BingNewsSerp

https://www.rudebaguette.com/en/2025/10/combat-drones-new-ai-challenges-aviation-norms-operating-without-pilot-or-runway-redefining-future-military-tactics/

https://www.slashgear.com/1680874/ai-piloted-f16-against-human-dogfight/

https://www.twz.com/air/our-first-look-at-yfq-44a-fighter-drone-prototype

https://en.wikipedia.org/wiki/Anduril_YFQ-44

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

https://en.wikipedia.org/wiki/General_Dynamics_X-62_VISTA

The mosaic model can be a novel approach to creating AI.

   The mosaic model can be a novel approach to creating AI. 

The regular model for creating AI is the large language model, or LLM. That means that. Developers must have control over large data masses. And large code structures. All large code structures are complicated. Because they involve a lot of code. And there are always some kind of errors in the code. That means, when the number of code lines increases, there are more possibilities to make errors. Making the AI is like making a statue. When the system turns more sophisticated. And more advanced, the development process turns slower. When we want to make statues, the first steps are always fast.

 But when we are closing the goal, finishing details takes more and more time. The finishing process just before the statue or product is ready for delivery to customers. Takes most of the time in the development process. The same way. Finishing and detecting errors from the data structure takes most of the time in the programming process. The problem is this: the LLM requires many event handlers. And each event handler is an independent algorithm. That means LLM is a large number of algorithms that operate as a whole. 

The problem is that the LLM rises like dough. That means there are more and more algorithms that must operate as one system. And when the code mass. And the data mass that the system operates on is both large. There can be lots of errors. The AI is a tool that can detect errors in other programs. But the system is not very effective if it must search for errors in its own code. The outsider algorithm can make error detection. AI needs a model. That it can compare its code to. With the code model. Actually, the outsider AI makes the comparison process with two other language models. 

And the problem is: when a new AI model comes. There are no models that the developers can compare their code to. Without models, it is impossible to compare even billions of code lines, which can involve minimal errors. But when there are billions of code lines, and let’s say. There is an error in 1000 lines; searching those codes is a long-term process. That requires carefulness. 

The modular, mosaic model helps expand the AI. 

The mosaic structure or the morphing neural network. That can involve multiple independently operating small language models. It can make it possible to make a flexible and elastic system. Every brick in the mosaic structure is an independently operating server that runs an independent small language model, SLM. Each SLM is a module that involves a certain skill. If the system doesn’t need a module. It can put the server that runs the module into sleeping mode. That makes it more energy efficient.  When developers want to give a new skill to the AI. They can make a new SLM and then connect. That thing to the entirety. 

Researchers say that the future belongs to small language models, SLMs. SLMs are lighter; they involve less code, which makes the development process easier to handle. The SLMs can also operate as a whole. This means: each SLM can operate independently. But those things can also form the virtual entirety. That entirety is like a mosaic. Each SLM has certain skills. When the user sends signals to the system, the SLM knows. If it has that skill or capacity. If the SLM doesn’t have a match. That system forwards the message to another language model, which has the database. That involves data about the SLMs and their skills. And the system sends that information to SLM, which can make the job. 

The idea is that each bite of the mosaic can be written independently. Which means developers can make new modules for AI. And then collect them into  a mosaic. The mosaic-type developing means that. Programmers can control the code better. That also decreases electric use, because if those modules or parts of the mosaic are not required, the system can put its servers to sleep. A mosaic model or morphing neural network structure allows developers to make an elastic and flexible system. 

AI-controlled drone swarms are entering battlefields.

   AI-controlled drone swarms are entering battlefields. 

The next step in drone technology is the development of AI-controlled drone swarms that utilize a LEGO-based system architecture. That cloud-based modular technology offers them flexibility and a multi-mission ability. Those drones can search enemy targets, support fire control, and make kamikaze attacks.  They can carry various sensors. Like Geiger meters, gas detectors, microphones, and other systems. 

That means when those systems require more calculating capacity, they unite their processors' capacity to work as a cloud-based morphing neural network. When the solution is made. Those drones, or their computers, can be separated, and then those systems, or each drone, can operate independently or as part of the entirety. 

Researchers took this idea. From a hypothetical alien model, where aliens can be like insects. When those insect-aliens require intelligence. Those things turn together. And then they share missions with each other. 

That idea is transformed into the drone swarms. Those drone swarms can call other swarms to solve complex problems, and then they share their missions or roles with the drones that participate in that swarm. Those drones can operate as one entity. Some drones can attack air defense. In that case, even small drones can be a more dangerous tool than anyone expected. They can hunt enemy commanders from the streets. The drone can slip into buildings through windows or ventilation. And even through sewer systems. If they are able to operate underwater. 

A large drone swarm can create a layer over the battlefield. Those drones can interconnect their sensors, which can transmit enemy moves to the commanders. Drones can also attack targets. That they recognize. The system uses images stored in the drone’s memory. 

One drone can drill a hole into the wall using explosives, and other drones can drive themselves into that hole. The drone can wait until its AI recognizes the target, and then attack it. The drone swarm can land on the roofs of trains. And then make their attacks. Large-sized cruise missiles or unmanned boats can carry drone swarms to target areas. Those drones can perform those missions without communicating with the command center. That makes them immune to normal jammers. 

The EMP system uses  high enough power EM impulses that it can destroy physical electronics. It can also destroy AI-controlled drones. The drone swarm can be released to the operational area just before manned aircraft comes. Drones can search for anti-aircraft artillery,  missiles, radar systems, and radio transmitters. They can attack ammunition storage. 

They can also make a radar and IR shield between aircrafts and ground-based systems. In those cases, small drones can search ground-based systems. And attack them. They can also disturb defense. Using jammer systems, aluminium bags, or IR lights that cover stealth planes behind them. Drone swarms can also close airfields, and they can fly into the aircraft’s jet engines. If drone swarms hover above runways, they can deny aircraft takeoff and landing. The small damage to the aircraft’s window or structure destroys the stealth-fighter’s stealth capacity. The drone swarm can communicate with aircraft. And stratospheric and orbital  satellites using laser systems. And those systems can also deliver those drones against the targets. 

https://www.msn.com/en-us/lifestyle/shopping/ai-powered-drone-swarms-have-now-entered-the-battlefield/ar-AA1LHwJg

The new computers are morphing neural network systems that mimic quantum computers.

   The new computers are morphing neural network systems that mimic quantum computers. 

"By linking smaller superconducting modules like building blocks, researchers at the University of Illinois Urbana-Champaign achieved near-perfect qubit performance. Their modular approach could open the door to scalable, flexible quantum computers of the future. Credit: Shutterstock" (ScitechDaily, Scientists Build Quantum Computer That Snaps Together Like LEGOs)

The fact is this. The regular binary computers can also operate like LEGOs. When a problem becomes too complicated for one computer. That computer can call more calculation units or computers. To operate on the problem. The system can call for assistance over the internet. That means when the computer doesn’t get an acceptable answer, it calls more computers to work with that thing. 

The new innovations in quantum computing represent a significant step toward a more efficient and effective way to calculate things. The reason why quantum computers cannot be stuck is this. They are like a tower of binary computers. Every layer or state in a qubit operates as an individual quantum computer, and if one of those states is stuck. 

Another state or layer comes and releases that state. When we think about the power of quantum computers.  We must remember that they can drive multiple programs. At the same time. Or they can cut and share complicated problems over those layers, and the AI-controlled quantum computer operational systems can act like LEGOs. 

Those systems can operate and run multiple different programs at the same time, but if that system sees something very complicated. That system will collect more and more quantum states and quantum units together to solve those problems. If the quantum system does not find an acceptable answer. That system connects more and more quantum units and quantum states to operate with complicated questions. 

So, in the case when the system doesn’t need very much power. That can allow all its units to work separately. With different problems. But when the system requires more power. The central system orders those systems to save their duties. And then start to work as a whole on that complicated problem. Things like drone swarms can use similar technology. That system can call all units to work on things. Like routes that those drones can choose. And then the system breaks entirely and shares those solutions to individual drones. 

The second big advance will be a room-temperature quantum computer. 

"Figure: (upper panels) Scanning-electron-microscope image showing a charge-density-wave device channel in the coupled oscillator circuit. Pseudo-coloring is used for clarity. Circuit schematic of the coupled oscillator circuit. (lower panels) Illustration of solving the max-cut optimization problem, showing the 6 × 6 connected graph, circuit representation of the six coupled oscillators using the weights described in the connectivity matrix, and values of the phase-sensitivity function. Credit: Alexander Balandin" (ScitechDaily, UCLA Engineers Build Room-Temperature Quantum-Inspired Computer)

The UCLA engineers built a quantum-inspired computer. The system will use a morphing neural network technology that mimics the quantum computer. That can change the world. The room-temperature quantum computers are tools that will revolutionize computing. When we think about things like quantum dots in virtual quantum systems, those quantum dots are the binary computers that operate like states operate in quantum computers. That makes those computers very powerful tools. Because those systems are immune to errors and stucks. 

"Scientists have built a physics-inspired computing system that uses oscillators, rather than digital processing, to solve complex optimization problems. Their prototype runs at room temperature and promises faster, low-power performance. Credit: Shutterstock" (ScitechDaily, UCLA Engineers Build Room-Temperature Quantum-Inspired Computer)

If some of those computers are stuck, some other computer releases that system. Because that system is morphing. That means all its participants can operate independently with different problems. But when a problem reaches a certain state of complexity. The system sends a message that all computers must unite their force to work on that problem.  

If a researcher makes a quantum computer that operates at room temperature, that system can be superior to the regular binary systems. There are so-called virtual quantum computers that operate in data centers. In those special neural computing systems. Each physical binary computer works as an individual quantum state in a quantum computer. The system operates entirely. It tries to mimic a real quantum computer. 

The system can operate like a quantum computer, but the qubit states are replaced. By using a physical binary computer. Those systems act like a quantum computer. That system’s Achilles heel is that it needs. A lot of power. If we want to make a virtual quantum computer. That qubit has 129 states, which requires a system with 129 binary computers. And that causes very big electric bills. Those systems release heat. This means those systems require powerful coolers and other things that protect those machines. 

https://scitechdaily.com/scientists-build-quantum-computer-that-snaps-together-like-legos/

https://scitechdaily.com/ucla-engineers-build-room-temperature-quantum-inspired-computer/

Do conscious AIs need to have the capacity to feel pain?

  Do conscious AIs need to have the capacity to feel pain?

How can we be sure that the AI feels pain? The “pain reaction” can be sensor-activated tape. When somebody touches a robot in a certain way. It can say, “It hurts”. This same reaction is possible to create. Using a tape recorder and an electric switch. That means the AI can have multiple reactions that mimic human reactions. So if the bot car has certain sensors and somebody kicks it, that system can say. “Please, don’t kick me”. 

Do the AI need to feel pain? That’s a good question, because if we want to create artificial superintelligence, we need to create a machine with consciousness. And that is very dangerous and difficult. Researchers and philosophers don’t agree on what consciousness is or what it entails to be you. Consciousness is being you, but how does that thing form? We say that some part of consciousness is the sum of information that our senses send to the brain. But that’s the first thing that we agree on. 

The second thing is that consciousness is the thing that is connected to memory. Our experience modeling our way to react to something. But the thing that we see the world as being at the middle of is a mystery. This thing is consciousness. But how does that thing form? We can make machines that can mimic feelings very easily. When somebody presses the robot’s hand with a certain power. That launches a reaction. Where the robot says, “This hurts”. The fact is that a similar reaction is possible to create.

By using a scale and a cassette recorder. When certain pressure impacts the scale. The hand presses the power switch. That conducts electricity to the recorder that plays the message to people. This thing means that. Robots that mimic pain are easy to make. But if we want to make a conscious robot, we must realize one thing. The robot can mimic human reactions very easily. If the robot is left behind, it can say, “Please don’t leave me”. Or, when somebody kicks the robot's shin, it can say, “This hurts”. 

In that case. Reaction is like reflexes. When something happens and there is a match in the system memories, that activates a reaction. If some merchandise’s weight is higher. Than a certain level. The robot can report that the merchandise is too heavy. The human-shaped robot requires the scales because it may carry too heavy things. Or presses too strongly. That breaks the robot or merchandise. 

The robot can use its own touch sensor that activates that reaction. Or the surveillance cameras can send that information to a robot. Those reactions are preprogrammed reactions. For a certain situation. Reactions are like tapes that certain actions activate. This doesn’t require deep consciousness. 

These kinds of reactions don’t mean that the robot really fears that situation. Another thing is that. If somebody tries to shut down the AI. There can be a program that only authorized persons can perform that operation. The system can only transform to use a secondary power line. That means the AI can refuse to shut down its main unit, even if it's not conscious. Consciousness makes AI dangerous. A creature with consciousness protects itself. That is only one way to see those things. The AI doesn’t need consciousness to refuse to shut itself down. And that means. It’s hard to state. If the system is conscious. Or if the system just follows reactions that were programmed into it. 

The ability to control the mind can be an ultimate training method. But it can also open the darkest visions of dictatorship.

The ability to control the mind can be an ultimate training method. But it can also open the darkest visions of dictatorship. 

Brightest innovations have darkest sides. That is the thing. That we should somehow learn. This is the thing. That people like Fritz Haber are shown to us. The same person who created ammonia synthesis and made fertilized irrigation possible created chemical weapons that killed millions in the First World War. In the same way. Things like artificial intelligence and brain-implanted microchips are tools. That can make many good things. But those systems can also control things like killer robots. And that means we must realize that all of those systems have two sides, bright and dark. 

The ability to control consciousness and decode thoughts is a tool. That has always fascinated people. The CIA and other intelligence agencies studied those things for a very long time. And an idea is to create a spy who doesn’t know being a spy. The person would walk into the office. And read documents and other things. And then that person goes to the safe place. And after the mark, that person can repeat everything that this person saw and heard. The idea is to use humans as walking recorders who don’t know that they are spies. 

The BCI system can make this thing possible. Brain circuits store information in the microchip, and then that system decodes that data in a safe place. 

That makes them pass things like lie detectors and even sodium amytal interrogation. In some other visions, the consciousness control can be used to hide top-secret technology. What if the operator can make missions using highly classified systems? And then that person will not remember anything. And any details about the missions and equipment are lost in their memories. The ability to control memories is a tool that can have many roles in the future. Controlled memories. Or, so-called synthetic memories are tools that can be used to train complicated things very fast. 

And that thing can save human lives. If a person is in trouble in remote areas. And the only person who can give first aid is person who lacks the needed skills, those systems can offer those skils in a very short time. Those systems can teach complicated algorithms to people, and they can make many things that we don’t even think about possible. Brain implants can be used to teach people things like karate. When we think about things like mind manipulation. We must realize one thing. In the wrong hands, those systems are extremely dangerous. 

When Brain-Computer Interfaces, BCI systems operate with people, they interact with the computer directly through the brain lobes. That means the mind cannot separate data, or information that it gets from the BCI, from information that the person’s own senses will send. And that is the danger of the BCI, that it is a very powerful tool. The BCI can cause a singularity where people are connected with the AI and each other. By those kinds of systems. In the worst cases, the system destroys personality and turns people into a homogenous mass of information. The biggest problem with the BCI-controlled computers is that. A person will not separate what is real and what is a cyber dream. That means the person can simply forget to sit on the computer and then die because of a lack of food and drinking. 

The biggest problem is this: the person. Hackers can target the users of wireless BCI connections. This thing makes remote extortion possible. The same system also offers a way to communicate using brain waves. That thing is called technical telepathy. Technical telepathy is thing that allows a person to control robots and other things using brain-implanted microchips. Those robots send information to the controller’s brain that handles that information. Like it comes from the senses. These kinds of systems are created to control bionic prostheses. But they can also control robots and other things. 

https://www.livescience.com/17449-matrix-inception-brain-manipulation.html

https://www.popularmechanics.com/military/research/a63149552/pentagon-psychic-spies/

https://www.sciencedirect.com/science/article/pii/S2589004223007526

AI’s advancement turns slower.

   AI’s advancement turns slower.

Growing accuracy requires more complicated code. That causes a situation where AI’s advancement slows. When developers created some pseudo-AI tool in a couple of hours in the 1980s, those programs required about 10-50 lines of code. Those programs asked “what's your name?” and then they output the name that the user gave. Then they might ask, “Is the sun shining”? And then the user could answer “yes” or “no”. Then the program replied with something that gave good things in the user’s mind. Today, AI algorithms require billions of code lines. And that causes a situation where the advancement slows. 

So, when accuracy grows in program advancement slows. 

AI’s advancement turns slower. When its accuracy grows. And that means the AI follows the line of the limits in mathematics. Term limits mean the equation that’s the curve approaches zero endlessly. But that curve never reaches zero. If we translate this mathematical equation into an AI model, we can say that when AI approaches human level, the advancement slows down. Maybe. The AI will never reach a complete human level for programmers, but it will reach a level that is almost human-level intelligence. So the base elements in the AI are easy to make.

Then, researchers should find something more accurate. At the same time, they must find new, complicated ways to train their AI. And that means there is a need for more complicated algorithms. Those algorithms require more power, more time, and more accuracy. This means that the programmers use more and more time. That developers can  create more complicated code for algorithms. And how they should react. When accuracy grows. The sector of their algorithm can work turn smaller in the same time. When the need for accuracy grows, the speed of programming slows. 

And the next thing is that the error detection must be at a level where the system can be trusted. Another thing. What slows AI’s advancement is the calculation power. The system needs the entire data center for every query. That means the system needs so much calculation power that developers have no money to buy the systems that they need. Complicated code requires high-power, very high-accuracy systems. And when the system requires lots of capacity for the smallest duties. 

That causes a situation where the developers don’t have time to use AI as they need. When some details keep the system busy, it has no time to drive new code. Complicated code requires a complicated- and long-term error-detection process. The human coder cannot check even billions of lines of computer code. The entire human lifetime is not enough for that process if the human coder wants to make that thing without automated tools. 

New drones are a challenge for security.

  New drones are a challenge for security. 

Geran 3 drone

The new jet-engined Shahed drones called Geran-3 cause problems for Ukraine's air defense. That is one of the things that the Ukrainian war has taught us. The rapid development of the drone industry can render countermeasures that were useful and effective a couple of weeks ago ineffective in current situations. Things like optical fiber-controlled drones are hard targets for defenders. 

Those things are almost immune to the normal jamming systems. The EMP pulses that can destroy electronic components or laser systems that can cut the command fiber are tools that can affect those drones. But the use of the non-coherent EMP means that their own drones and other vehicles in the EMP area are also in danger. The laser system should detect the optical fiber before it cuts that fiber. 

And fiber-controlled drones can also be put in chains. That improves their attack range. The optic fibers would be pulled through another drone, and the drone can make a chain that is hard to detect or avoid. There is also a possibility that drone swarms are controlled by using laser communication. The laser drone can be a fiber-controlled system. And it can send laser pulses to other drones wirelessly. 

The laser-controlled drones have one problem. They require straight eye contact or a fiber that transfers the command signal. Regular lasers have one problem in drone control. Those lasers must be aimed. Using high accuracy. Blinking holograms can help with that problem. The optical communication is harder to jam. But the laser rays cannot travel through the smoke and fog. Those things also cover drones from defenders. 

 The Shahed-drones are now dropping anti-tank mines. But there is a possibility that the jet-engine-powered versions of those drones can also carry anti-tank weapons. Those anti-tank or anti-radiation missiles can increase those drones’ ability to cause damage. And another thing is that those drones can use those missiles against the power stations or radar installations. 

And they can also disturb the air defense. That is the problem with weapon development. The solutions that were effective against propeller-engine Shahed-136 drones are not effective against jet engine versions of those drones. The jet-engine drone can fly much higher than those propeller drones. The next step will be  AI-controlled drones that can operate independently. The drone will fly near its target using inertial navigation. Or it can use TERCOM, the AI follows the terrain area, and the system compiles that data with images. Those are stored in its memory. 

When the drone is close to its target, the system starts to search point, that it must destroy. The system uses photographs that are stored in its memory to detect the target. And then the system orders the drone to dive. And make a kamikaze attack. Basically. Those drones can use similar systems that Javelin uses to detect targets. Image-based target recognition can be used in all types of missiles and drones. 

https://www.armyrecognition.com/focus-analysis-conflicts/army/conflicts-in-the-world/russia-ukraine-war-2022/russias-jet-powered-shahed-238-drones-introduce-new-challenges-to-ukraines-air-defenses

https://www.twz.com/air/russian-mig-29-fitted-with-an-interceptor-drone-is-a-laughable-mess

https://www.twz.com/air/russias-shahed-long-range-drones-are-now-dropping-anti-tank-mines

https://en.wikipedia.org/wiki/HESA_Shahed_136

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

MIT's new robot learned by watching.

MIT's new robot learned by watching. 

The new breakthrough in AI and robotics is a tool that can learn like humans. Those robots can look at the monitor and then learn and repeat those things. That means the robot is easier to teach, and no programming is needed. The ability to learn like humans makes robots more versatile. But the thing is that those robots are only one part of the AI-based tools. 

The robot can act as a medium that transmits data that it collects to the AI-central computers. And basically, the AI can learn things by using any camera system in the world. That gives the ultimate possibilities for the AI. If we think that the AI learns to drive, it's possible to use the robot as the medium. 

The robot sits in the classroom and drives a car following orders that the instructor gives. Then the system can transfer those models to the robot vehicles. Or, maybe the driver robot includes the service if somebody buys a robot car. The man-shaped robot can carry things like luggage to the car. And the same robot can wash the car and clean the house if its master wants. 

The problem with a human-shaped robot is that it can do the same things as a human. The human-shaped robots can learn things simply by looking at things on TV. That makes it possible to use the regular movies to train robots in civil and military actions. The thing is that the robot's shape must not look like a man, so that it learns things. The robot aircraft, or drone, or robot car can use the same methodology as a human-shaped robot to learn new tricks. 

Digital twins can make it easier to teach AIs. The digital twin can be the character that a person drives in computer game-type simulations. The digital twin can be the tool that helps to see how the system controls the real robot. The digital twin can also make it possible to advance reactions. That physical robots can face in their missions. The digital twin makes it possible for that system to advance AI. And test the AI and program's ability to respond to challenges. 

The AI can use things like aircraft or satellites to record maneuvers that potentially hostile actors make. Then the AI can drive those maneuvers in the digital simulation. Or simulate a battlefield and create counter-maneuvers to that thing. In the same way, the AI can collect data from the flight profiles of the aircraft, and then the system starts to create counter modes for that maneuver. This is the thing that makes the AI an effective tool in civil and military operations. 

https://www.rudebaguette.com/en/2025/08/that-robot-learned-just-by-being-watched-mits-terrifying-ai-builds-3d-control-map-from-video-alone-no-sensors-or-programming-needed/

https://scitechdaily.com/ai-twins-could-help-save-the-planet-but-only-if-we-fix-them-first/

AI and cartels.

   AI and cartels. 

Increasingly, more actions are being outsourced to AI. The stock market is a vast arena where individuals can utilize more or less sophisticated algorithms. The algorithm can see things. Like, changes in courses, like raising the value faster than a human. Then the AI-trainer must have the point where to sell and where to buy. The system must follow the small losses of the values and then decide to buy or sell. 

The AI must buy cheap shares and sell them when their value is high enough. The AI can follow more targets than a human. And it can see how the share values advance. If those shares are losing their value fast. And a person doesn't watch that curve, which causes losses. The AI will never get boring. And it ever sleeps. The case of the investor finds something else interesting: the screen on the table can cost millions in seconds. 

The AI can be a tool that solves those problems. But algorithms can exchange information illegally. The cartel means a situation in which a person uses an inner circle or non-public information to make profits from the stock market. 

And one version of cartels is a case. When traders make a contract to drive investments to certain targets. When somebody buys shares. That raises their values. So the group just dumps money into some target. The AI can also accidentally ask for advice from another AI. 

There can be a leading algorithm. That leader algorithm makes investments, and then other algorithms follow that thing. The fact is that,  even if some algorithms or AI chatbots look different, they can operate on the same platform. The AI chatbots can be the same, even if they seem different. And that thing means the AI can start to drive investments to certain objects. One AI-based solution can operate millions of investors' investments. 

This kind of AI can theoretically aim for even billions of dollars to the same target. And that thing makes it possible to raise the value of the shares. The AI is the tool that can react faster than humans. Criminals can create the AI-based investment cartels using quite simple algorithms. The AI that can control millions of objects in the bors can guarantee that a person cannot lose money in that game. The AI can invest small sums of money in a large group of targets. The system can buy and sell shares far faster than a human. So the person will always get profits. 

https://fortune.com/2025/08/01/artificial-stupidity-ai-trading-stock-market-behaviors-price-fixing-collusion-wharton-study/

https://finance.yahoo.com/news/artificial-stupidity-made-ai-trading-110500308.html

Can we trust AI?

   Can we trust AI? 

Artificial intelligence is a tool that doesn't create information from emptiness. That means the AI uses data and information that already exists. If something or somebody changes information from the data sources that the AI uses, the AI gives wrong answers. The AI can use old-fashioned information. In the same way, a human can take an old book of facts from the shelf. 

That is one thing that we must remember. The second thing is that most AI chatbots are owned by companies whose purpose is to bring profits for their owners. That causes a situation where some of the answers that the AI gives are customer-friendly. That means they are made to please customers, or people who pay for those chatbot services. 

However, one of the most important aspects of using AI is that it is actually utilized incorrectly. But the question is, why is it used? Because it makes work more effective. That means people who work with the AI assistants have no time to think and check the answers that the AI gives. The prime focus in AI use is that it makes people do more work. 

The modern business ecosystem has evolved over time, when all products were physical. That means the business ecosystem measures the effectiveness of the worker by simply calculating the number of work performances in a time unit. And that means the number of code lines that a programmer makes determines how effective that person is. And that means there is no time to analyze and think about the product that the AI makes. 

This is the problem with AI. It makes people more effective, and it should leave time to make analyses about the product. But the problem is that if the AI allows the person to work at double speed, that gives the company leaders the opportunity to fire half the workers. That decreases costs. 

When we talk about how AI makes things less effectively than a top-level programmer, we can find one very interesting question. If we analyze that argument closely, we face the claim that all programmers are top-level specialists. All programmers are top-level specialists. And another thing is that some workplaces are so hurried that they have no time to advise junior programmers. Sometimes old workers think that the young comrade is a competitor, and that makes the workplace atmosphere poisonous. 

The person might not dare to ask for advice from workmates. And that causes firing. The thing is that the AI is a tool that can help people. But we must remember that this tool allows us to analyze and think about new possibilities, only if we give it that chance. But we must have time and willingness to check the data that the AI gives. And if we see the AI only as a tool that allows us to cut the personnel costs, that is not the way to make our working life easier. 

If we just copy-paste texts that the AI makes. And don't even try to analyze that data, we don't increase trustworthiness in the company. That means we deliver the decisions to the AI. AI can also use old-fashioned datasets. Or it can use fake or falsified data. That means the AI makes mistakes in the same way as humans. It simply takes the wrong book from the shelf, and then that causes destruction or catastrophe. 

The AI is not immune to fake information. The problem with propaganda home pages is that their page rank are raised by drumming it with net queries. That makes those homepages easier to find. The problem with those pages is this: they use .com or some other neutral terminal identifiers. So, those homepages are not easy to connect to China or Russia. They are operated from servers that are in Western countries. The operators use a remote control to control those servers that operate in Western sockets. 

If the AI uses regular web browsing applications, that means it can select a propaganda page. The popular pages are always at the top of the page lists. And that thing means that the AI selects popular homepages. That decreases the AI's trustworthiness. The AI requires training. That training means that the users must set limits on what kind of sources those systems use. 

The AI in the helmet can revolutionize aviation and spaceflight.

 The AI in the helmet can revolutionize aviation and spaceflight. 

"Illustration of AI Model ChatGPT Navigating a Simulated Spacecraft in a Competition." (Rude Baguette)

Think about AI systems.  Like Chat GPT, as a robot controller. That tool will bring an ultimate boost for the robot that the controller can command using spoken language. But what if that tool is integrated into the aircraft and spacecraft? That can integrate multiple systems under one roof. And when commanders say that there is something. That must be removed; the user must not do anything but give the command. 

The system selects the closest robot that has the strength to make that movement. Robot groups can operate like ant swarms under AI control. The system can include flying units that can search for things for land-moving robot ants. 

What if the space suits are equipped with an AI socket that the system can call an assistant robot if there is something wrong? The future spacecraft can have things. Like assistant robots to make their missions go as they should. 

"The XQ-58A Valkyrie demonstrator, a long-range, high subsonic unmanned air vehicle completed its inaugural flight March 5, 2019 at Yuma Proving Grounds, Arizona." (Wikipedia)

"A XQ-58 Valkyrie deploys an Altius-600 unmanned aircraft system"(Wikpedia)

The AI that can fix errors in maneuvers can be a tool that makes many things safer. The real thing is the AI-controlled operating system in the combat aircraft. The system can integrate with multiple sensors. 

Such as satellites, other aircraft, and ground vehicles. Those sensors can be optical, radar, or acoustic sensors. The thing is that the AI-controlled systems can give flexibility to mission planning. And that kind of system can also make maneuvers in cases where the pilot has no time to react to every threat. The next-generation fighters are rather flight groups than individual systems.  

There are cruise missiles and other kinds of systems flying with those fighters. The drone can be a full-size fighter that can make autonomous attacks against targets. Or that robot fighter can also act in a kamikaze role. The new cruise missiles can also have the ability to refuel themselves in the air. Those missiles can have countermissiles, and they can make feint and instinct maneuvers, and change their speed and flight profiles during operations. And those missiles might have the ability to return to base if they are not needed. 

They must have the ability to communicate with drones and drone swarms. Those, maybe hypersonic aircraft, can drop drones to target areas, and those drones' missions can be to search targets, disturb air defense, and take out the electric supply. Those drones can also destroy things like fuel supply, and the AI-based systems can also target and assassinate enemy commanders. Those systems can be extremely dangerous in the wrong hands. 

https://automatedresearch.org/weapon/area-i-anduril-altius-600m-and-700m/

https://www.flightglobal.com/military-uavs/taiwan-receives-first-anduril-altius-600m-loitering-munitions/164081.article

https://www.rudebaguette.com/en/2025/08/chatgpt-at-the-helm-of-a-spaceship-could-change-space-travel-forever-after-stunning-early-successes/

https://en.wikipedia.org/wiki/Kratos_XQ-58_Valkyrie

The new humanoid robots can be more dangerous than anyone thought.

  The new humanoid robots can be more dangerous than anyone thought.

The new robots serve customers. And they can act as the most advanced sex-dolls that we ever imagined. The AI-controlled robots can operate independently. And the problem is that those robots are from China. The Chinese government is one of the most authoritarian governments in the world. So those leaders should realize the value of those human-shaped robots for surveillance, spying, and assassination tools. The robot "James" or robot adjutant that follows some general can transmit every single thing that the general says to the control center. The thing that makes those robots problematic is that those systems have no limits on how they handle their missions.

They have no conscience as we do. And that means those robots can be even more traitorous than some fateful lady is in gangster movies. There are movies. There, people are replaced by robots. And those Chinese robots are one step closer to that way. In the wrong hands, those robots are a bad daydream. The butler robot can also act in all other roles where humans can. Those butler robots can also act as surgeons, if their command algorithms are changed. That robot can make it possible to live in remote places. If some humans require help, those robots can acquire surgeons' skills just by asking the program.

But in the same way, those robots can act as eliminators if the person they serve does something that is not allowed. The robot can make everything that humans can. The same robot can clean house, fix cars, or even act as a surgeon or commando if the programming of those systems is changed. The skills of robots depend on their programming.

And that is one challenge to data security. If an unauthorized person changes the algorithm, that can have horrible results. Law enforcement and intelligence agencies are interested in the AI-controlled robots that they can use in covert operations. Those robots will not change their side. And they follow orders all the time. The computer program or algorithm includes orders that the robot must follow.

The computer program, or the algorithms, differ from humans in that the computer program is the ultimate authority for computers. Computers and robots never resist their programming. They excuse their orders only because that ability is programmed into them. But if a robot does not have the ability to resist, it will do everything that its master orders.

For working perfectly, those robots require lots of calculation capacity. They can make contact with other robots, like robot cars. And those systems can share their data processing ability if they cannot participate in the robot's operation otherwise. The system can also communicate with data centers. In that case, the robot is only a relay station. That is the interface between humans and the data center base, central computers.

https://interestingengineering.com/innovation/ai-sexdoll-china-starpery-technology

https://www.rudebaguette.com/en/2025/08/worlds-first-humanoid-robots-quest-for-a-phd-in-opera-and-drama-sparks-fears-of-ai-replacing-human-creativity-in-arts/

How to teach AI?

 How to teach AI? 

If we want to make AI that operates smoothly in everyday life, we must make one thing clear to ourselves. The thing is that the AI must handle many variables so that it can operate independently. And it's hard to program those things into the AI’s source code. Even the best programmers forget something. So the answer can be the robot that sits in the classroom and learns like a human. This helps to create all the necessary variables for the AI. The robot acts as a medium between humans and the AI. 

If we want to make self-developing AI, we must use two AIs. The first AI makes a requirement of what it needs for the other AI. Then another AI creates source code. Then, that system must connect the AI to the source code of the first AI. And after that, the system reboots the servers. Before that, the system must simulate how it controls the virtual robots. 

The AI’s self-development happens as in natural organisms. The AI must detect what it needs, and then the biggest problem is how the AI can create code that involves the data it needs. Source code is the DNA of the AI, and there must be something that tells the AI what a certain code sequence makes. The answer to that problem would be the digital twin, the other AI that it can use to simulate what certain source code makes. The best way to get the code is simply to ask for it from some other AI. However, that means the other AI must have access to that source code. 

That requires that the AI tell its problem to other computers. And those other computers give it the code that it needs. But there are legal and other kinds of problems for that kind of self-developing model. So, how do we improve our skills? If we don’t know something, we search for data on the internet or visit the library and read about it in some books. Or, maybe we can take some courses on the topics. 

The thing in AI self-development is how the AI describes the problem to itself? When we talk about teaching AI, we can imagine a situation where the robot is sitting in the classroom and listening to how the teacher explains something. This is one version of the AI training. The AI can also search how some coders make solutions and then copy that code into its digital twin. If we want to teach AI to drive cars, we can use the same method as we use when we learn to drive cars. The driver must drive a car that records the actions for the AI. Then the AI asks necessary questions, and the human accepts that process if it's safe. 

We must remember that there are many variables that the AI should notice in the traffic. It would be very hard to describe all of those variables and traffic signs for the AI by writing them into its source code. The answer would be that the programmers make the base model for the AI. And then AI fills that model. And one of the most interesting ways to learn how to drive is to put the human-shaped robot into a driver’s school. The robot must know that it must avoid humans, know the traffic light, and other regulations. Then the robot that the AI uses for its developing platform or transmitter that transmits data to the AI will finish that training.