The AI and new upgrades make fusion power closer than ever.

"New research highlights how energetic particles can stabilize plasma in fusion reactors, a key step toward clean, limitless energy. Credit: SciTechDaily.com" (ScitechDaily, Inside the Tokamak: Scientists Crack the Code to Stable Fusion Energy)

The new systems. Like magnetic field-controlled plasma brings fusion closer than before. 

The new version of Tokamak-ITER makes it easier to control plasma flow in the fusion reactor. One of the problems is how to deny the fusion ignition to send an energy wave that destroys the plasma ring that travels in Tokamak.

The answer could be that the system must make more ignition points to that reactor. 

AI-based systems can control magnets and ignition systems more accurately. 

 Fusion tests are needed for that thing. The improved model of Tokamak is called ASDEX.

ASDEX Upgrade (Axially Symmetric Divertor Experiment) is a divertor tokamak at the Max-Planck-Institut für Plasmaphysik, Garching that went into operation in 1991. At present, it is Germany's second-largest fusion experiment after stellarator Wendelstein 7-X. (Wikipedia, ASDEX upgrade.)

"An edge-localized mode (ELM) is a plasma instability occurring in the edge region of a tokamak plasma due to periodic relaxations of the edge transport barrier in high-confinement mode. Each ELM burst is associated with expulsion of particles and energy from the confined plasma into the scrape-off layer. This phenomenon was first observed in the ASDEX tokamak in 1981. Diamagnetic effects in the model equations expand the size of the parameter space in which solutions of repeated sawteeth can be recovered compared to a resistive MHD model. An ELM can expel up to 20 percent of the reactor's energy" ( Wikipedia, Edge-localized mode)

"3D visualization of an ELM in the ASDEX Upgrade tokamak as simulated with the MEGA code. The tokamak volue is colored according to the ELM structure. TheELM interacts with the energetic particle whose orbit is shown in green. Credit: Figure adapted from J. Dominguez-Palacios et al., Nat. Phys. (2024), CC BY 4.0)"(ScitechDaily, Inside the Tokamak: Scientists Crack the Code to Stable Fusion Energy)

The problem with the plasma in the fusion reactor is its very high temperature. In that material, even small effects can cause very big effects. If the plasma ring turns asymmetrical the system cannot form fusion. The plasma hovers in a magnetic field that denies contact with the wall of the torus. The most effective way to make the fusion is to use ion and anion plasma. 

Those plasma rings would pull each other together. And the laser can ignite the fusion. The problem is that those anions and ions must not touch each other before the temperature is high enough for fusion. Another problem is the point where fusion ignition starts. There must be enough ignition points. The energy impulse in the plasma ring causes ignition. That ignition sends an energy wave to plasma and destroys that plasma ring. 

Tokamak is one of the most common and advanced fusion reactor test beds. Another thing that can be promising is the system. That uses two linear particle accelerators to shoot ions and anions against each other. That model is the same way easy to use as Tokamak. In that linear model, the particle accelerators shoot ions and anions into a laser ray. And that causes ignition between those linear accelerators. In that model adding more fuel to the reactor is easier than in tokamak. 

https://www.iter.org

https://scitechdaily.com/inside-the-tokamak-scientists-crack-the-code-to-stable-fusion-energy/

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

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

https://en.wikipedia.org/wiki/Edge-localized_mode

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

https://en.wikipedia.org/wiki/Wendelstein_7-X

Splitting electrons can unlock a new era for quantum technology.

"Discovering that electrons can manifest as ‘split-electrons’ in nanoscale circuits, scientists are now on the brink of a technological revolution that could lead to the practical use of Majorana fermions in quantum computing, thanks to a novel use of quantum interference. Credit: SciTechDaily.com" (ScitechDaily, Scientists Create Split-Electrons, Unlocking the Future of Quantum Computing)

Scientists are unlocking a new quantum era using splitting electrons. The observation that electrons can split is fundamental because they are elementary particles that should not decay. The ability to split electrons is a big advance for quantum computing. Controlling electrons is easier than controlling photons. But otherwise, electrons have a problem. Those particles react with the electromagnetic fields. 

And that means magnetic fields can turn those particles into a new position. The system must keep the information. That the system sends and receives in identical form. 

Normally quantum computers use superpositioned and entangled particle pairs. But the problem in that case is those particles must be identical. 

They must oscillate with the same frequency and then their energy level must be on different levels. The transmitting side must have a higher energy level than the receiving part of the quantum entanglement. Splitting electrons can solve that problem. The system can simply put an electron split and then send the other side of it through the quantum route. That route could be the nanotube. Then the frame locks the receiving part of the particle pair to the right position. 

In error correction, the system must send the same information twice. The two-stage quantum model is one way about how to make error detection easier. In that model, the information travels first between superpositioned and entangled particle pairs. Then system sends the transmitting particle after that data transmission behind that ultrafast data. And receiver can compare is information that traveled in the quantum track to the information that the follow-up particle transmits identical to the information that travels in quantum entanglement. 

Then the system can start to make the data transmission between those electrons. The problem is that things like magnetism can affect the electrons. In an ideal version, quantum computers can use nanotubes to transport information through them. The nanotube is the electromagnetic wormhole and the system should send electrons through it. In that case, the electron acts in the same role as a neurotransmitter in human brains. That system requires a different type of thinking than traditional networks and quantum systems. 

The ability to make superposition and quantum entanglement through those nanotubes makes the system more effective. And it makes error detection easier.  The system makes the superposition between two particles and sends information with this "turbo speed". Then the system will send the transmitting side particle after that superposition. The system will check that the data is identical by comparing information that traveled between superpositioned and entangled particles. And then in particles that travel through the nanotube. 

https://scitechdaily.com/scientists-create-split-electrons-unlocking-the-future-of-quantum-computing/

New AI-designed microchips are more effective than before.

New AI-designed microchips are more effective than traditional microchips. They look different. Than human-made microchips. They can be tools. That will improve computing. New large and complex programs like large language models require powerful systems that can drive complex code and connect sensor information with it. 

Those microchips are required for new and powerful computers that run the AI.  The new computer-aided design/computer-aided manufacturing tools make microchip development. And microchip design is more effective than before. 

They are smarter, faster, and game-chancing. As the ScitechDayly's article says. 

AI can drive CAD drawings straight to the manufacturing platforms. The ability to turn the image straight to the product makes the R&D process more effective than before. The 3D printers can make products straight from the CAD images. 

The artificial intelligence that understands normal language also makes these kinds of things more effective. 

The reason for that is that AI is free from traditions. AI can use data and research results without connecting to some traditions. That allows it to make new, radical designs that are strange for regular designers. That makes AI the ultimate designer. The AI can use complete models of the best and most effective microchips are completed It can connect things like magnetic field models and other kinds of things like optimal temperatures for those designs. The system makes new and more effective microchips. 

The new, small-size microchips require complete knowledge of the electromagnetic fields and their interactions. Without that information the system is useless. Those very small microchips can operate as antirety. Or the system can use them separately. 

That makes new computers more effective and error-free than ever before. But the problem is that the system must know when it is stuck. If some of those processors in the multicore environment halt or spend too much time for some operation the system must call other processors to work with that that. Problem. Or it must drive information past the stuck processors. That denies the jam of the entire system. 

The system can operate even if some part of it has a focus on some problem. The AI can make the system that makes the intermediate records if there is some more urgent data in the system. When that more important, or higher-priority data is handled the system can return to work with the problem that is not so important. 

AI makes even single-processor computers more effective if it can clean their memory. That helps the system to handle data batter. When memory filling, that AI records data to a hard disk, cleans memory, and returns to the tasks. 

https://scitechdaily.com/ais-strange-chip-designs-are-faster-smarter-and-game-changing/

Microchips,systems,AI,designs,

AI gives superpowers to personal computers.

"Diffeomorphic Mapping Operator Learning, DIMON, a new AI framework, accelerates modeling by solving partial differential equations efficiently, reducing computation times from days to seconds. Tested in heart simulations, it promises transformative applications across engineering and science." (ScitechDaily, AI Breakthrough Solves Supercomputer Math on Desktop PCs in Seconds)

"The adaptable technological solution has the potential to revolutionize engineering designs." (ScitechDaily, AI Breakthrough Solves Supercomputer Math on Desktop PCs in Seconds)

"A breakthrough in artificial intelligence is making it possible to model complex systems—like how cars deform in crashes, how spacecraft endure extreme conditions, or how bridges withstand stress—at speeds thousands of times faster than before. This innovation allows personal computers to tackle massive mathematical problems that once demanded the power of supercomputers." (ScitechDaily, AI Breakthrough Solves Supercomputer Math on Desktop PCs in Seconds)

"The new AI framework offers a versatile and efficient method for predicting solutions to challenging mathematical equations. These equations are crucial for modeling phenomena such as fluid flow or electrical current behavior in various geometries, commonly encountered in engineering and design tests." (ScitechDaily, AI Breakthrough Solves Supercomputer Math on Desktop PCs in Seconds)

New AI gives table computers the supercomputer abilities. The idea is that the system uses the complete and complicated models. 

That makes it unnecessary to begin all modeling processes from the beginning. The system can use models that some other computers have made. 

The idea is similar to the picture. So we can use mosaic pictures as an example. Developers make it by using ready-to-use sub-elements. The ability to use free elements is making data-handling operations more effective. That increases the power of regular computers. But it also fits in use of the supercomputers.

"A Revolutionary AI Framework: DIMON. Details about the research appear in Nature Computational Science. Called DIMON (Diffeomorphic Mapping Operator Learning), the framework solves ubiquitous math problems known as partial differential equations that are present in nearly all scientific and engineering research. Using these equations, researchers can translate real-world systems or processes into mathematical representations of how objects or environments will change over time and space." ((ScitechDaily, AI Breakthrough Solves Supercomputer Math on Desktop PCs in Seconds)

And the thing that can boost that ability is the memory handling tool that cleans memory when the model is ready. The system develops models as layers. Every single layer is an independent model. When the model is ready the computer can store it or give it to another computer. 

Then it can clean its memory. Then that system can receive the model that another computer has worked. And continues to develop or build the new model. In that model, the AI-based systems play ping-pong balls with models that they create as stages. In every stage, the system can connect more and more complicated objects into that layer. The system handles program data like Photoshop handles its layers. 

The image that the developer makes could be an example. The city area image. 

The developer or system can select items for that image from the data library. The system can use things like images of houses. That made for some other purposes. 

And that is the idea of object-oriented programming. The idea in C++ and similar programming languages is simple. 

They involve libraries of commonly repeating operations. That means the user of that language doesn't have to program things like simple calculations from the beginning. They must just load the 

Those libraries (like Stdio.h) deny the need to do basic things every time the programmer starts a new job.  

The programmer can use libraries that involve responses for the orders that the programmer writes. Using the programming language. 

The new AI-based model is a new and very advanced version of the old object-oriented programming. The system can collect complicated models from libraries. 

Oor from the net. And that makes it possible for the table computers to get steroids. The other thing is that the computer can take another computer to its work. The system can make the digital twin and download data to the other computer while it cleans its memory. The ability to clean memory while the system develops the model makes it more powerful. In that model, the model is developing in stages. 

When the system makes the new model it can transfer it to its digital twin. And the system can clean the memory of the first computer. The AI-based system can also call more computers to handle the operation. 

https://scitechdaily.com/ai-breakthrough-solves-supercomputer-math-on-desktop-pcs-in-seconds/

https://www.nature.com/articles/s43588-024-00732-2.pdf

https://pubmed.ncbi.nlm.nih.gov/39653845/

New systems allow researchers to follow cellular communication wirelessly.

"To improve biosensing techniques that can aid in diagnosis and treatment, MIT researchers developed tiny, wireless antennas that use light to detect minute electrical signals in liquid environments, which are shown in this rendering. Credit: Marta Airaghi and Benoit Desbiolles" (ScitechDaily, MIT Scientists Harness Light To Wirelessly Monitor Cellular Electrical Signals)

The new systems use light to read cellular electric signals. Electric impulses on the cell's protein shell control the ion pumps. Ions are the key element in chemical communication between neurons and other cells. The immune system requires those ions to detect cells. That doesn't work as it should. 

The cell uses them to take in and remove nutrients. Ion pumps play a key role in communication between neurons and other cells. 

It's possible. Neurons can communicate with individual cells benefiting the electric vortex at the ion pumps. Or it can receive and deliver neurotransmitters through those ion pumps. 

That can make it possible to create new types of medical treatments and new ways to control and boost our wealth. In the most interesting ways to think they can exchange information between the cells and computers. And what if you can give orders to bacteria in your stomach that they must transport food straight away and not drive it in your body? That thing can save your life in the case of poisoning. 

The ability to communicate with bacteria will be an incredible advancement in microbiology. That thing might be closer than ever. The bacteria can used to produce virus antigens that can be covered with immune activators. That makes it possible to program the immune system to fight against new diseases. Genetically engineered bacteria can also create cancer cells' shell proteins. 

Those proteins can offer the possibility to activate the immune system to detect harmful cells in the human body. 

That makes immune cells destroy tumors and cancer cells in the early stage of cancer. In that stage cancer is a group of individual cells that are easy to kill. In that stage, they don't form tumors yet. 

In the future, it may even be possible to order bacteria out of the body. But before researchers can communicate with bacteria and give them orders, they must read its messages. That thing requires new tools that can observe 

https://scitechdaily.com/mit-scientists-harness-light-to-wirelessly-monitor-cellular-electrical-signals/

The compact missiles are the new threat in the hands of the military.

The new small missiles that Ukraine developed are the new tools for the military. First, those missiles are meant to destroy drones. But that can be a new type of threat to all targets on the battlefield and behind it. 

In the Scifi novel, "Dr. Adder" the writer, JW Jetter" introduced the anti-personnel miniature missiles. Those missiles can turn real sooner than we ever imagined. The new microchip technology, combined with AI is the tool that can make it possible to attack individual targets by using supersonic aircraft. 

The idea of those small missiles is that they will use similar AI-based image-homing systems as Javelin. 

Those missiles can be dangerous for small vehicles. 

Those new missiles can take out even individual persons. Their optical seeker recognizes the target and then the missiles home into it. 

Attackers can use small-size missiles to break things like aircraft windshields. 

They can make holes in the aircraft's wings. Because the system uses rockets that small drones can carry the quadcopter must only go in the right distance to the target.

The quadcopter can use that missile against aircraft, and chemical factories. The missile is far faster than quadcopters and the copter can shoot it from a distance. The same system. That is effective against slow drones.

 Is ineffective against those missiles. 

Those missiles are far faster than quadcopters and harm things like electric wires and electronics like communication antennas and radars. The quadcopter can also be a kamikaze tool that can follow its missile. 

Those missiles are making it possible that quite soon even intercontinental ballistic missiles can shoot individual persons. 

From thousands of kilometers. Those missiles can be mounted in quadcopters.

They can make the SciFi horror true sooner than we imagine. Those miniature missiles can make it possible to destroy light-armored targets like snipers. The personal body armor protects people quite well against rifle bullets. The new full-body armors make it necessary to create new systems that can penetrate those things. Those targets can also be combat robots that can also be equipped with titanium armor. Those robots are small and they need very fast reactions. 

https://interestingengineering.com/military/miniature-missile-to-counter-low-flying-drones

The AI is ready for the new cold war.

AI is a tool that didn't exist in the last Cold War. It's more effective than any computer program before. It's flexible and capable of acting as a tool that can be the primary actor in espionage. The AI-controlled systems can attack their targets in multiple ways at the same time. They can connect data from multiple sources. 

When we think about hackers and hacking there is no useless information in those operations. Books on the table, hobby cars, and other kinds of things can give a tip for passwords. 

In the same way, buried animals and other things can be the tip. The artificial intelligence writes to targeted systems login screens.

The AI is special. Because that system follows regular speech. It learns automatically from its mistakes. And that makes it harder to deny the attacks. 

In domestic use, the AI can search for things like unauthorized attempts to make contact with Western actors. 

It can search for things like data. That is harmful to governments. Drone swarms are one thing that governments can use to control people. The drone swarm can collect data from large areas. 

Drones can follow individual people. Weaponized drones can control people like demonstrations. And those recon and killer drones can be shot to another side of the world. 

Satellite control and artificial intelligence can make it possible for those drones can observe any office. And home anywhere they go. Those drones can target and eliminate individual people. 

It doesn't require special and advanced programming skills. It can analyze captured code and its functionality. The AI is the platform that connects many devices into one entity. 

That network can involve virtual and physical platforms like drones, programming environments, and other surveillance equipment. The AI can search the net to find the things in the person's history that can help it find the code that lets it into the targeted systems. 

The AI can create fake news and virtual characters. That can look like presidents. That makes those systems very dangerous. 

The ability to break things like the news channels' computers allows the AI to slip fake information into the broadcast. That is one version of the threats that AI can cause in the new Cold War. The AI can use virtual characters and voice transformation for social hacking. It can adjust the power of the power plants. 

The problem with the hacking of the new Cold War is that happens under the control of governments. Those governments don't follow any international laws. They use stoled data in their products. 

Those hackers search for data about things like people who dare to resist their governments. The addresses of refugees and the names of informants make newspapers and media houses interesting targets for hackers. 

https://www.freethink.com/artificial-intelligence/how-ai-is-being-drafted-for-a-digital-cold-war

https://aiandthefuture.wordpress.com/2025/01/06/the-ai-is-ready-for-the-new-cold-war/

The AI can predict our decisions and affect them.

"AI assistants may soon predict and influence our decisions at an early stage, selling these “intentions” to companies in real time, creating an “Intention Economy,” say Cambridge ethicists. These tools, leveraging vast behavioral data and persuasive AI, could manipulate human plans for profit, raising concerns about ethics, privacy, and societal impact, especially on elections, media, and markets." (Scitechdaily, AI Could Predict and Sell Your Desires Before You Know Them)

The next step in AI development is personally customized commercials. That means that every user can get AI-generated commercial material, personally made for that individual user. Deep neural networks can make customized commercials for individual users. Personal commercials are the next step from the targeted commercials. 

The deep neural network can create personal profiles of all internet users and create commercials. Targeted to one individual user. 

The users with a big effect are good targets for customized commercials.

The person with access to big power and big resources is a good object for targeted commercials. 

If that person is the company leader in the enterprise company. 

Those commercials have a bigger positive effect. 

From the point of view of salesman. 

Those people don't use their own money when they buy something for their company. 

Normally individually targeted commercials are targeted for enterprise leaders who decide contracts that the company does. That gives those commercials the largest possible effect. But AI can create commercials for every regular user. 

Those personalized commercials are the AI version of the P2P sellings. 

AI is a tool that can collect precise information about people. AI can make many things that we cannot even think and that means AI can predict our behavior better than we can. AI can follow the ways, how we use the internet, and what is our favorite music and movies. Things that we do alone in our bedrooms tell what we really are. 

The AI is an algorithm that operates in a deep neural network. All computer networks can be networked into deep neural networks. 

The thing that makes a regular network a deep neural network is the computer program that can turn the system sense. 

The capacity to analyze data is the second thing that the deep neural network can do. If the neural network has access to things like WEB-cameras it can see if the user is interested in things that run on the screen. 

If the system can see the location information and what kind of material the person runs on the device. 

That helps it to know if a person looks at videos. 

Or if a person just listens to music. 

When we are walking we might not stare at videos from our mobile devices.

But using that information the AI can make customized commercials for us. That thing guarantees that people will see them. 

The customized commercials are made using our user profiles.

Things that we do in work time are different from what we do in free time. 

The things that we do in front of our spouse can be different from what we do alone. 

The system uses music and video elements that please the user.

That makes those personalized commercials a dangerous and effective tool. 

The AI can create sounds that stress our brain areas that control our pleasures. That makes us see the commercials.

Everything that we write, study, and make on the net tells about our interests.  

How to make people see commercials?

That is the biggest problem for the salesmen. If people just go to get the sandwich or do some other things while the commercial runs.

That is a waste of money from the point of view of people who pay for them. 

Internet marketing's effect base is on user profiles and interests. The idea of the interest is that the commercial will bring interesting things to the user's computer screen. The problem is that every user has different interests. The AI can collect data and then make decisions about what products it offers for the users. 

If commercials can introduce interesting products.If the system selects products with a connection the user's other internet use. That makes those commercials interesting and raises sales. 

https://scitechdaily.com/ai-could-predict-and-sell-your-desires-before-you-know-them/

https://aiandthefuture.wordpress.com/2025/01/05/the-ai-can-predict-our-decisions-and-affect-them/

Artificial superintelligence (ASI) can be far different than we can think about.

The biggest difference between artificial general intelligence (AGI) and artificial superintelligence (ASI) is that the ASI is not connected to certain systems as AGI. This means that ASI is like a fog where reality moves. 

"A superintelligence is a hypothetical agent that possesses intelligence surpassing that of the brightest and most gifted human minds. "Superintelligence" may also refer to a property of problem-solving systems (e.g., superintelligent language translators or engineering assistants) whether or not these high-level intellectual competencies are embodied in agents that act in the world. A superintelligence may or may not be created by an intelligence explosion and associated with a technological singularity." (Wikipedia, Superintelligence) 

The artificial superintelligence ASI is the tool that connects people with the brain-computer interface (BSI) with the internet. The internet is like a matrix that allows people to teleport virtually from one place to another. That is like an ultimate virtual reality. 

The route to artificial superintelligence is long and full of surprises. Artificial general intelligence (AGI) is a tool that mimics human intelligence and learning methods. 

That tool will be as intelligent as humans. But it learns faster than humans. 

However, artificial superintelligence (ASI) is the tool that beats humans in intelligence. ASI will be more intelligent than humans are. 

It will be more powerful than humans. The ASI can be the first intelligent creature that we meet. 

But then we must ask how the AI is more intelligent than we are. The AI is a tool that knows many things. It can have the entire library in its memories. 

That makes it possible for the AI to answer every question. That we ask in seconds. 

People with internet connection. Can find answers to the same questions in seconds. So, how can we measure intelligence? Would a thing like a chess game be the thing that measures intelligence? The reality is that computers from the 1990s beat the average chess player in chess. 

If they have a good chess program. And if some chess world champion wins some computer, that doesn't mean that is possible for the average player. 

We must realize this when we talk about ASI. When we think about ASI as scientists we must ask would that thing be intelligent? 

Or will it be less intelligent? If somebody orders it to make some new theories? 

Or we can ask ourselves, would Albert Einstein be less intelligent if somebody had ordered him to make the "Theory of Relativity"? If we think of superiority and the henchman who makes that theory? Einstein or his supervisor? That is an interesting question. 

Artificial superintelligence can communicate with people through robots or brain-computer interfaces, BCI. This means that people can talk with artificial superintelligence using the BCI systems. Or the person who has that system can act as a medium between ASI and humans. Maybe that medium is rather a humanoid robot than a human who carries the BCI. 

The BCI must adapt to the slow data-handling speed of human brains. 

The system must use the system that covers the entire brain shell to exchange information between brains and ASI. 

The speed of thinking is extremely low. But the brains can use millions of axons for data transport. The slow speed in the single wire is not a very bad thing. That allows other neurons to control communication that travels in the axons. 

The brains operate oppositely. Or in a different way than computers. In human brains, data travels in neurotransmitters which means brains mimic the quantum neural system. The neurotransmitters are like qubits that travel in the brain. 

Neurotransmitters keep brain temperature low because information is stored in those chemicals. Lots of low-power data handling units help to keep the temperature low. Fast units require more nutrients and they produce more heat. So, a large number of low-power units will keep the temperature low. 

So neurocomputers might use nanotubes to transport ions that mimic neurotransmitters. 

They can use optical systems where lasers with different colors or wavelengths can transport data in that complex system. 

The neurocomputer is a system that requires lots of processors. 

Billions of neurons are working in human brains. And that creates a very big temperature problem. Same way temperature causes problems in large-scale neurocomputers. 

The brains trust a large number of connections rather than in small number of fast connections. 

In those connections travels smaller data mass, than internal connections in computers.

The data system in human brains is not like a computer. Things like neurotransmitter chemicals are tools. Whose mission is to protect brains.

But in the case of damage or misconnections. The lost data mass is lower than in computer damage. The large number of low-capacity connections allows the brains to keep voltage low in those connections. And if there is damage. That allows other neurons to rescue the information. 

The neuron requires time to load information in chemical neurotransmitters. Artificial brains will require ion cannons between transmitting and receiving routes. Those ions mimic the neurotransmitters. 

Those chemical's mission could be to keep the temperature low. In the human brain. If the brain uses only electricity that causes the high temperature and shortcuts. So brain uses neurotransmitters as qubits and the system in human brain mimics quantum network. 

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

https://aiandthefuture.wordpress.com/2025/01/04/artificial-superintelligence-asi-can-be-far-different-than-we-can-think-about/