Do machines think, or do they know what they do?

Cogito ergo sum (I think, therefore I am) 

René Descartes

Sometimes philosophers say that intelligence is being you. That means if a machine thinks it's intelligent. And when the machine is intelligent, it's itself. That is one of the key questions of robotics. The machine is a thinking creature. That is the thinking loop. So, should machines have rights? When we step on the machine's foot. It can say "ouch". But we can make the same effect with a switch and tape recorder. 

The tape mechanical switch activates the tape recorder there is the recorded word "ouch". And we might not think this system is intelligent. But that is one very simple example of so-called pseudo-intelligence. 

Pseudo-intelligent machines can make quite impressive things. They can ask for key cards and say "Hello Tom" when the person logs into the system. They can involve many complicated reactions. But those things are fixed in the program code. The real intelligence means that the system can learn new things. 

But then we must think ourselves. What happens if we would spend our entire life in some house? We have not seen cars and shops before. And then we just step out of that building and start to walk on the streets. What would the world look like? Will we do something funny if we don't ever visit the shops? What happens if we don't know what a tree is? Robot programmers face that situation all the time. 

We don't automatically know what traffic, cars, trees, and shops mean. We learn those things from an early age. And we might think that those things are self-evident. But they are not. We learn those things from our parents. We have natural phobias. We are afraid of spiders and snakes because those things can be dangerous. 

Sometimes in the past, some of our ancestors noticed that those animals and insects could bite painfully. And that thing is stored in our DNA. Pseudo-intelligent systems can make many things. 

But those systems cannot learn anything new automatically. They need something to teach that thing to them. Or when a programmer teaches computers the programmer programs it. 

Thinking machines are becoming a reality. But before we say that machines think or do not think, we must understand what thinking means. Or we must create a description of what the word "thinking" means. "I think therefore I am,": said French philosopher René Descartes. That means thinking makes a creature exist. And being existing means that creatures can affect us without us asking permission. 

That means we are lords of that creature. We are afraid of machine rebels. And we are afraid that machines take our jobs. But before we blame machines, researchers, and engineers we can ask, would you do those jobs that are transferred to AI-controlled robots? When we talk about thinking robots and artificial intelligence. We talk about things like chess robots, learning computer games, and other stuff, that can transform their knowledge into real-world robots. 

When somebody plays a combat simulator on the net, the simulator can transfer those tactics to the military combat simulators. In that case, players teach the real-world combat robots to handle situations. The research about thinking machines began from chess-robots and it still continues. But when the computer first time won human in chess that was great news. The fact is that the news was about the best chess player on this planet. 

And the AI can win over an average chess player a long time before. So, are we fair? If there is one person on this planet, who can win AI can we say that AI loses to humans? 

Perhaps, we should rather start to think about the complexity of intelligence. And we must stop thinking about the extreme cases. Is the world champion the average person in games? That is one question that we should think about. 

Thinking is the ability to connect information, stored in memory to the information. That the sensors bring to the system. That means thinking is like making puzzles. The machine thinks the same way as the human. Databases are the memory of computers. 

It connects collected data to the database. Combine it with the memory. And make new data entirety.

As you see. Theoretically, AI is quite easy to make. 

The only problem is this:  How to route data to the right database? The database involves information on how to react to each situation. While we discuss with computers or large language models (LLM) we have lots of time to think how to make the questions. And we have plenty of time to analyze and look at that data. 

But when some robot operates on streets there is no time to analyze situations. In real life. 

Situations come to the front of the system very fast. The environment can look different than in simulation and that is a problem with robots. 

The process that we call "active denial" can solve this problem. In that case, the system has a filter that eliminates every unnecessary data packet that the system doesn't need. The control system removes things like houses. And other things from the computer that desire reactions. However, the active denial system must recognize the houses. 

Also, it is harder to make descriptions of situations that come from the camera. That describes how the machine must react to some words. If the robot connects observation with the wrong database, that can cause a catastrophe. 

When we make things like robot cars that car faces unpredicted situations. 

Normally orders for unpredicted situations would be. That the car just stops or slows its speed. But, there is one case where the car should not do that. That is a case of robbery. 

But then we must realize one thing. When we meet a machine that is more intelligent than we are, we face one problem. The intelligent machine can hide its intelligence level. That means intelligent machines can lose us in chess. If it is afraid that we will shut down the server that runs the AI. The AI is a tool that can make many things. However, AI requires a physical body to make an effect on the physical world. Without a robot body, the AI cannot move things and clean yards. The robot body is the tool that makes AI capable of making physical things. 

The Chinese hackers can steal the HFSS the world's most effective R&D program.

"Yaoguang software has been available for free for the past few months. (Representational image)" (InterestingEngineering, China releases ‘world’s most powerful’ weapon design software, 15x faster than US)

Chinese developers released new software that fits the weapon research. That program is similar to U.S. software that U.S. developers used for simulating magnetic catapults and other kinds of military stuff. Chinese developers gave their software name "Yaoguang". The program is a Chinese copy of the HFSS the most effective material and electromagnetic simulation program in the world. 

Developers made The Yaoguan similar complex simulations as HFSS. But it's faster. 

That can mean that the Chinese use better supercomputers than Americans. Or they use something differently than HFSS users. 

Or the code in the program is tuned better. Those simulations require high-power computing. But Yaoguang makes those complex electromagnetic reflection simulations faster. The HFSS uses three hours for those simulations. The Yaoguang does it in 12 minutes. 

(Interesting Engineering)

And that makes this problematic. It's possible. That the Chinese hackers stole the HFSS source code. And then they can make changes to that software. That requires the necessary programming skills. Those changes can involve more advanced memory control. That makes it more powerful. 

"Yaoguang takes just 12 minutes to perform radiation simulation analysis on the multi-band antenna used in the new phased array radar while Ansys HFSS, a powerful electromagnetic industrial software in the United States, takes three hours to complete such tasks." (Interesting engineering, China releases ‘world’s most powerful’ weapon design software, 15x faster than US)

If the developers connect an HFSS-type development tool with a large language model. That makes it the most powerful tool in the development process. 

The weapon-research software is always dangerous. People have seen how AI can boost weapon development research (R&D). One user has lost the account because of the use of Chat GPT for weapon-making. That person developed the model for an automatized gun. The Open AI didn't tolerate that thing. But the AI is ready for military purposes. The policy is not a good and effective way to deny weapon research. 

The policy can change. Somebody can steal the source code. And then make changes to it. There are hackers in the east, who are ready for that project. They can operate for governments. But people like Mafia can also operate in this kind of project. 

The developers can create "the evil twin" of the famous chatbots. People like Russian oligarchs and mafia have money. They can hire developers to make that kind of software. The fact is that people are afraid that China or North Korea can steal the Chat GPT source code and make necessary changes to it. For making the catapult and other simulations the AI requires access to programs that can handle magnetic simulations. The program that the Chinese published is more effective than HFSS. 

https://interestingengineering.com/military/china-weapon-design-software

https://www.techtimes.com/articles/309031/20250111/openai-says-no-ai-weapons-sentry-gun-powered-chatgpt-now-shut-down.htm

https://www.randsim.com/software-solutions/electronics/ansys-hfss/

Atom-size microchips are coming.

"When circular polarized light is irradiated onto a thallium lead alloy, the majority “up spins” of electrons flow in the right direction (spin-polarized current). Credit: Taniuchi et al 2025" (ScitechDaily, Revolutionizing Electronics With Single-Atom Spin Currents)

Theoretically, researchers can create transistors using three carbon or metal atoms. The base, collector, and emitter would be atom-size structures. 

The atoms that spin are controlled and can input electricity to that system. Two, or maybe three-atom-size structures can form switches and routers. 

That thing can make ultra-small microchips possible. Atom-size routers and switches are coming to microchip technology. The system's base is in a system. That controls the atom's spin. And that can unlock new and very fast processor technology. Miniature, nano-size microchips are tools that make many things possible. 

Nano-size microchips can control nanomachines. The problem with nanomachines is their small size. The outside remote controller can control those tools. However, they cannot operate in electromagnetic fields which limits their ability to operate for medical purposes. The small microchips allow those systems to operate as well as in versatile missions. 

They can drive as complicated code as larger computers. The new microchip technology opens new routes for medical and military applications. 

There are two versions of independently operating machines. First uses outside computers to control it. The system uses remote controllers that are connected to outside computers. The independent operating machine that uses an internal command unit is independent of remote controllers and data transmission channels. That makes it hard to affect them from the outside. 

And that is one of the wildest things that we can imagine. The futuristic liquid robots that can change their shape are groups of nanomachines. Those systems seem solid structures. But they are a large group of small silicone nano-robots. 

Nano-size microchips are also tools that can operate very highly intelligent robots. The system that fits in the human-size robots can have far more calculation units than in the highest-power supercomputers. And that makes them drive very complicated codes and algorithms. Those robots can be far more intelligent than even humans. 

https://scitechdaily.com/revolutionizing-electronics-with-single-atom-spin-currents/

Sam Altmann says that superintelligent robots are coming.

We are ready for the next step in the AI. The fast internet and deep neural networks turn superintelligent robots into reality fast. Nobody predicted how fast this advance could be. 

The AI-controlled robot can use large language models remotely anywhere where it can reach the network. The network allows that robot to communicate with supercomputers. If a robot or its computers have access to supercomputers and LLM they can give very intelligent impressions of themselves. The impression is like the cases where LLM is connected with microphones and loudspeakers. 

The robot must have cameras as eyes. Microphones and loudspeakers for vocabulary communication and text-to-speech/speech-to-text applications. The system transforms speech into text and then dumps it for the LLM. During that process, the system must have an error correction protocol. 

That allows the robot to communicate with the LLM. The vocabulary correction makes the speech more understandable to LLM. If the LLM doesn't understand what the user says it can give the list of choices, what the LLM thinks the user wants. That makes it possible to give commands to robots in a noisy environment. 

If the LLM doesn't recognize those words it can give a list of what the word looks like, and then ask to select from the menu would the word be A, B, or C. Or if there is no match the user can give the command again using better articulation. 

Satellite communication allows the robot to communicate with fixed supercomputer sites. Anywhere on the planet. The only needed thing is the system has access to a satellite. 

The LLM can itself be in the computer center. But the robot bodies can also create neural networks by connecting each other. Into one entirety. This kind of network-based application benefits the shared calculation capacity. That makes those systems handle information very effectively. 

But when we think about superintelligent robots the robot itself is not intelligent. A supercomputer that controls those things is intelligent. Or the network-based solution that connects those robots in their entirety is intelligent. However individual robots are not intelligent. Their computing capacity is low. But united they are strong. 

In robotics superintelligent robots are tools that can follow spoken commands. 

They are tools that can have the same abilities as humans except they are not living. The controller can give spoken commands to the system. The weakness of that model is the background noise. The solution for that thing can be voice recognition that allows the system to deny unauthorized commands. 

Active denial or active filtering is the tool that denies robots and LLM to take unauthorized commands. The AI can also ask for confirmation after each command. That thing means that the system introduces what it should do and waits the acceptance before it starts to act. The command can be "clean the yard" and when the robot starts to operate it can clean its entire area without different orders. 

And those robots can communicate with the user. Those robots are only tools that follow the supercomputer orders. We can think that those systems are like advanced mobile telephones. 

https://bigthink.com/business/sam-altman-the-superintelligent-robots-are-coming/

AI turns the R&D process very fast.

 

Image by Gemini

Maybe in the future, the R&D process will be fully automatized. Polish SciFi writer Stanislaw Lem introduced that model in his novel "Peace on Earth" the model where the AI-developed machines will turn a threat to Earth. The system would include two systems. Super simulators and selective simulators. 

Supersimulators are AI-controlled computer-based systems where the system creates virtual models of the aircraft and other tools. The selective simulator is the AI-based system that hosts the opponent's digital twins. 

In the R&D the super simulator creates a virtual model. Then a selective simulator uses its digital models to fight against it. It's possible that if the super simulator and its model win. It delivers that model to the selective simulator. And after that, it gets data on why those digital competitors win or lose. Then it tries to develop a new model that beats its original model. 

AI shows its power in the R&D process. The AI developed in Dubai created the rocket engine in three weeks, while NASA engineers made the same system over decades. 

That is one of the areas. Where AI shows its power. The system requires specific and well-made orders. Then, AI and deep neural networks do the job. Deep neural networks can connect free, internet-based data collected from simulators and laboratories. 

Deep learning systems can develop materials. And shapes of other things. Like aerodynamic structures in other laboratories. 

AI can involve those processes in other projects. Highly advanced flow simulations are tools that can make many things virtually, that previously required physical models. Because the system uses mainly virtual tools it's cheaper than traditional processes. The system can make even millions of simulations and calculations in seconds. 

When AI controls the 3D printing systems it can search things like bubbles inside the materials. That makes those systems more flexible and more effective than previous systems. The fast R&D process is the key element in many new tools. The ability to make the rocket-engine type of tough tools that require precise and high-quality work. That kind of system is dangerous in the wrong hands. 

The AI-based R&D process makes it possible to create things like stealth fighters in a new and effective way. 

AI-generated structures and materials including computer codes are things. That some people like Kim Jong-Un want. 

X-ray spectroscopy, scanning tunneling microscopes aerodynamic and radio reflection simulations are tools that make it possible to create very advanced systems very fast. 

https://interestingengineering.com/photo-story/ai-designed-rocket-engine

https://leap71.com/2024/06/18/leap-71-hot-fires-3d-printed-liquid-fuel-rocket-engine-designed-through-noyron-computational-model/

In data security. The weakest link is the user.

Quantum tools make new materials and information possible. Information security and quantum technology are more advanced than before. Even the best systems cannot protect information. If people don't care about the basic rules of data security. 

Unauthorized persons have access to the rooms. Where there is access to quantum computers that destroys data security immediately. Another thing is that if the personnel are not prepared for social hacking.

Social hacking can mean cheating, burglaries at home or even hostage-taking. Violent gangs can also use violence to extort passwords. 

That makes the system vulnerable. Knowledge is power in security. 

The knowledge about virtual actors and other things like hidden cameras makes the systems a world more secure. If intruders can put cameras in the office. 

 If the camera sees the screen. It makes even the best quantum security useless. The hacker sees everything that happens on screen. And that is a big risk to data. 

Hackers can try to lock the user's access to the system by typing the username. When a user makes the new password the hacker sees that from the screen. 

That is one version of the threats to data security. The second thing that endangers data security is long-range communication. Long-range communication uses traditional binary transmission. Enhanced secure versions can use different wavelengths and different routes to enhance data security in long-range data transportation. 

The system can cut information into pieces and send those pieces to different routes or radio frequencies. That makes data systems more secure than regular systems. Data segments in the system must have serial numbers that the receiving system can collect so that data flows in the same order as it is transmitted. 

Because information is bound with physical particles it makes quantum systems more secure than traditional networks. The problem with this thing is it doesn't work for long distances. 

Information transportation that is connected with a physical transporter doesn't alone make anything secure. Even in the best systems we must use strong passwords and be sure that the entrance to the office is secured.  

If the user doesn't use strong confirmation in the system. That makes quantum systems the same way vulnerable as regular computers. The weakness is the quantum systems is the binary port. The binary port is the machine that connects quantum systems with things like input-output tools. 

The binary port controls the quantum system and data that travels between long distances is always vulnerable to attacks. AI and social hacking are tools that can also cause vulnerabilities. Social hacking is a problem in all data systems. And if a person forgets their cell phone or even small paper there is a new password on the table. 

That destroys even the best security protocols. The ability to take pictures of the screen and papers that people forget endangers security as effectively as traditional hacking. 

Fascinating prime numbers.

Is there some non-random sequence in prime number series? 

Prime numbers are divisible by only one and by itself. That thing makes prime numbers the prime tools for ASCII (American Standard Code for Information Interchange) encryption. The ASCII means the numeric code of keyboard symbols. In that process, computers count ASCII numbers using prime numbers. That allows the encryption computer to hide messages from outsiders. And that makes the prime numbers so important. Prime numbers have also fascinated mathematicians throughout history. 

Famous mathematician Carl Friedrich Gauss (1777-1855) spent 15 minutes per every day in his career. Calculating prime numbers. Another mathematician, Bernhard Riemann (1826-1866), created the "Riemann's Zeta function" for calculating prime numbers. The Zeta Function zero points should not include any other than prime numbers. That's when researchers use normal numbers. But there are zeros in complex numbers. 

"The real part (red) and imaginary part (blue) of the Riemann zeta function along the critical line Re(s) = 1/2. The first non-trivial zeros can be seen at Im(s) = ±14.135, ±21.022 and ±25.011." (Wikipedia, Riemann zeta function)

"This image shows a plot of the Riemann zeta function along the critical line for real values of t running from 0 to 34. The first five zeros in the critical strip are clearly visible as the place where the spirals pass through the origin." (Wikipedia, Riemann zeta function)

Numbers from 1 to 49 placed in spiral order. (Wikipedia, Ulam Spiral)

and then marking the prime numbers:(Wikipedia, Ulam Spiral)

There can be a couple of nonprime numbers in Riemann's Zeta Function zero-point series in the case that the zeros. Riemann's hypothesis is one of the biggest unsolved mathematical problems. The problem is this: is there some kind of non-random order of prime numbers? Is there some kind of non-random sequence in the prime number series? If that series exists and somebody finds it, that thing can revolutionize number theory. 

But then we can think about another way to find the prime numbers than calculating. Mathematician Stan Ulam created the Ulam Spiral. If we want to calculate prime numbers using the Riemann Zeta function that takes time. The Riemann Zeta function is the tool whose weakness is that it forms prime numbers in reading lines. Faster computers must only calculate that series that slower computers used. And that breaks the code immediately. The Ulam spiral is the tool that should respond to the need to select certain prime numbers without linear models. 

"The Ulam spiral or prime spiral is a graphical depiction of the set of prime numbers, devised by mathematician Stanisław Ulam in 1963 and popularized in Martin Gardner's Mathematical Games column in Scientific American a short time later. It is constructed by writing the positive integers in a square spiral and specially marking the prime numbers." (Wikipedia, Ulam Spiral) 

"Ulam and Gardner emphasized the striking appearance in the spiral of prominent diagonal, horizontal, and vertical lines containing large numbers of primes. Both Ulam and Gardner noted that the existence of such prominent lines is not unexpected, as lines in the spiral correspond to quadratic polynomials, and certain such polynomials, such as Euler's prime-generating polynomial x2 − x + 41, are believed to produce a high density of prime numbers. "(Wikipedia, Ulam Spiral) 

"Nevertheless, the Ulam spiral is connected with major unsolved problems in number theory such as Landau's problems. In particular, no quadratic polynomial has ever been proved to generate infinitely many primes, much less to have a high asymptotic density of them, although there is a well-supported conjecture as to what that asymptotic density should be."(Wikipedia, Ulam Spiral) 

"In 1932, 31 years prior to Ulam's discovery, the herpetologist Laurence Klauber constructed a triangular, non-spiral array containing vertical and diagonal lines exhibiting a similar concentration of prime numbers. Like Ulam, Klauber noted the connection with prime-generating polynomials, such as Euler's" (Wikipedia, Ulam Spiral) 

https://www.3blue1brown.com/lessons/zeta

https://www.quantamagazine.org/mathematicians-uncover-a-new-way-to-count-prime-numbers-20241211/

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

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

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

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

The Gemini AI tells Riemann's hypothesis in this way.

What is the Riemann zeta function?

The Riemann zeta function is a special mathematical function that is closely related to the distribution of prime numbers. It is defined over the complex numbers and has many interesting properties. The zeta function can be used to study the distribution of prime numbers among positive integers.

Zeros

The zeros of the zeta function are the complex numbers for which the function's value is exactly zero. These zeros are distributed throughout the complex plane, and their location has deep connections to number theory.

Trivial zeros: The zeta function has infinitely many so-called trivial zeros at negative even integers (-2, -4, -6, ...).

Non-trivial zeros: These are more interesting. The Riemann hypothesis states that all non-trivial zeros lie on the so-called critical line, where the real part is exactly 1/2.

Riemann Hypothesis

The Riemann hypothesis is one of the most famous unsolved problems in mathematics. It states that all non-trivial zeros of the 1  zeta function lie on the critical line.   

Why is the hypothesis important? If the hypothesis were proven to be true, it would be a huge breakthrough in number theory and would have significant applications in many other areas of mathematics.

Why has the hypothesis not been proven yet? The Riemann hypothesis is a very difficult problem, and proving it would likely require entirely new mathematical tools.

Summary

The Riemann zeta function is a complex and fascinating mathematical object whose zeros are closely related to the distribution of prime numbers.