It is commonly accepted that ‘the brain computes’ and that it serves as a model for establishing principles of technical (first of all, electronic) computing. Even today, some biological implementation details inspire the implementation of more performant electronic implementations. However, grasping details without context often leads to decreasing operating efficiency. In the cases of those major implementations, the notion of ‘computing’ has an entirely different meaning. We provide the notion of generalized computing from which we derive technical and biological computing, and by showing how the functionalities are implemented, we also highlight what performance losses lead the solution. Both implementations have been developed using a success–failure method, keeping the successful part-solutions (and building on top of them) and replacing a less successful one with another. Both developments proceed from a local minimum of their goal functions to another, but some principles differ fundamentally. Moreover, they apply entirely different principles, and the part-solutions must cooperate with others, so grasping some biological solution without understanding its context and implementing it in the technical solution usually leads to a loss of efficiency. Today, technical systems’ absolute performance seems to be saturated, while their computing and energetic inefficiency are growing.

The operation of the brain has inspired von Neumann’s famous paradigm. How did the data- and event-controlled analog brain inspire the model of the instruction- and clock-controlled digital computer? One of his great ideas was to store the computer’s knowledge in the computer’s memory. Why loading a ready-made software program into its memory is the only way to teach the computer’s hardware to do something? Given that the instruction execution of the processor is so fast, why is the software-implemented training by machine learning so painfully slow and power-wasting? Is using another computing model, component, or algorithm with faster operation risk-free?

Neural information theory represents a fundamental method to model dynamic relations in biological systems. However, the notion of information, its representation, its content and how it is processed are the subject of fierce debates. Since the limiting capacity of neuronal links strongly depends on how neurons are hypothesized to work, their operating modes are revisited by analyzing the differences between the results of the communication models published during the past seven decades and those of the recently developed generalization of the classical information theory. It is pointed out that the operating mode of neurons is in resemblance with an appropriate combination of the formerly hypothesized analog and digital working modes; furthermore that not only the notion of neural information and its processing must be reinterpreted. Given that the transmission channel is passive in Shannon's model, the active role of the transfer channels (the axons) may introduce further transmission limits in addition to the limits concluded from the information theory. The time-aware operating model enables us to explain why (depending on the researcher's point of view) the operation can be considered either purely analog or purely digital.

In all kinds of implementations of computing, whether technological or biological, some material carrier for the information exists, so in real-world implementations, the propagation speed of information cannot exceed the speed of its carrier. Because of this limitation, one must also consider the transfer time between computing units for any implementation. We need a different mathematical method to consider this limitation: classic mathematics can only describe infinitely fast and small computing system implementations. The difference between mathematical handling methods leads to different descriptions of the computing features of the systems. The proposed handling also explains why biological implementations can have lifelong learning and technological ones cannot. Our conclusion about learning matches published experimental evidence, both in biological and technological computing.

Neuroscience extensively uses the information theory to describe neural communication, among others, to calculate the amount of information transferred in neural communication and to attempt the cracking of its coding. There are fierce debates on how information is represented in the brain and during transmission inside the brain. The neural information theory attempts to use the assumptions of electronic communication; despite the experimental evidence that the neural spikes carry information on non-discrete states, they have shallow communication speed, and the spikes’ timing precision matters. Furthermore, in biology, the communication channel is active, which enforces an additional power bandwidth limitation to the neural information transfer. The paper revises the notions needed to describe information transfer in technical and biological communication systems. It argues that biology uses Shannon’s idea outside of its range of validity and introduces an adequate interpretation of information. In addition, the presented time-aware approach to the information theory reveals pieces of evidence for the role of processes (as opposed to states) in neural operations. The generalized information theory describes both kinds of communication, and the classic theory is the particular case of the generalized theory.

Experience shows that cooperating and communicating computing systems, comprising segregated single processors, have severe performance limitations, which cannot be explained using von Neumann’s classic computing paradigm. In his classic “First Draft,” he warned that using a “too fast processor” vitiates his simple “procedure” (but not his computing model!); furthermore, that using the classic computing paradigm for imitating neuronal operations is unsound . Amdahl added that large machines, comprising many processors, have an inherent disadvantage. Given that artificial neural network’s (ANN’s) components are heavily communicating with each other, they are built from a large number of components designed/fabricated for use in conventional computing, furthermore they attempt to mimic biological operation using improper technological solutions, and their achievable payload computing performance is conceptually modest. The type of workload that artificial intelligence-based systems generate leads to an exceptionally low payload computational performance, and their design/technology limits their size to just above the “toy” level systems: The scaling of processor-based ANN systems is strongly nonlinear. Given the proliferation and growing size of ANN systems, we suggest ideas to estimate in advance the efficiency of the device or application. The wealth of ANN implementations and the proprietary technical data do not enable more. Through analyzing published measurements, we provide evidence that the role of data transfer time drastically influences both ANNs performance and feasibility. It is discussed how some major theoretical limiting factors, ANN’s layer structure and their methods of technical implementation of communication affect their efficiency. The paper starts from von Neumann’s original model, without neglecting the transfer time apart from processing time, and derives an appropriate interpretation and handling for Amdahl’s law. It shows that, in that interpretation, Amdahl’s law correctly describes ANNs.

Experience shows that cooperating and communicating computing systems, comprising segregated single processors, have severe performance limitations, which cannot be explained using von Neumann's classic computing paradigm. In his classic "First Draft," he warned that using a "too fast processor" vitiates his simple "procedure" (but not his computing model!); furthermore, that using the classic computing paradigm for imitating neuronal operations is unsound. Amdahl added that large machines, comprising many processors, have an inherent disadvantage. Given that artificial neural network's (ANN's) components are heavily communicating with each other, they are built from a large number of components designed/fabricated for use in conventional computing, furthermore they attempt to mimic biological operation using improper technological solutions, and their achievable payload computing performance is conceptually modest. The type of workload that artificial intelligence-based systems generate leads to an exceptionally low payload computational performance, and their design/technology limits their size to just above the "toy" level systems: The scaling of processor-based ANN systems is strongly nonlinear. Given the proliferation and growing size of ANN systems, we suggest ideas to estimate in advance the efficiency of the device or application. The wealth of ANN implementations and the proprietary technical data do not enable more. Through analyzing published measurements, we provide evidence that the role of data transfer time drastically influences both ANNs performance and feasibility. It is discussed how some major theoretical limiting factors, ANN's layer structure and their methods of technical implementation of communication affect their efficiency. The paper starts from von Neumann's original model, without neglecting the transfer time apart from processing time, and derives an appropriate interpretation and handling for Amdahl's law. It shows that, in that interpretation, Amdahl's law correctly describes ANNs.

Classic science seemed to be completed more than a century ago, facing only a few (but growing number of!) unexplained issues. Introducing time-dependence into classic science explained those issues, and its consequent use led to the birth of a series of modern sciences, including relativistic and quantum physics. Classic computing is based on the paradigm proposed by von Neumann for vacuum tubes only, which seems to be completed in the same sense. Von Neumann warned, however, that implementing computers under more advanced technological conditions, using the paradigm without considering the transfer time (and especially attempting to imitate neural operation), would be unsound. However, classic computing science persists in neglecting the transfer time and is facing a few (but growing number of!) unexplained issues, and its development stalled in most of its fields. Introducing time-dependence into the classic computing science explains those issues and discovers the reasons for its experienced stalling. It can lead to a revolution in computing, resulting in a modern computing science, in the same way, as it resulted in modern science's birth.

Today’s computing is based on the classic paradigm proposed by John von Neumann, three-quarters of a century ago. That paradigm, however, was justified for (the timing relations of) vacuum tubes only. The technological development invalidated the classic paradigm (but not the model!). It led to catastrophic performance losses in computing systems, from the operating gate level to large networks, including the neuromorphic ones. The model is perfect, but the paradigm is applied outside of its range of validity. The classic paradigm is completed here by providing the “procedure” missing from the “First Draft” that enables computing science to work with cases where the transfer time is not negligible apart from the processing time. The paper reviews whether we can describe the implemented computing processes by using the accurate interpretation of the computing model, and whether we can explain the issues experienced in different fields of today’s computing by omitting the wrong omissions. Furthermore, it discusses some of the consequences of improper technological implementations, from shared media to parallelized operation, suggesting ideas on how computing performance could be improved to meet the growing societal demands.

Machine learning intends to be a biology-mimicking learning method, implemented by means of technical computing. Their technology and methods, however, differ very much; mainly because technological computing is based on the time-unaware classic computing paradigm. Based on the time-aware computing paradigm, the paper discovers the mechanism of biological information storing and learning; furthermore, it explains, why biological and technological information handling and learning are entirely different. The consequences of the huge difference in transmission speed in those computing systems may remain hidden in "toy"-level technological systems but comes to the light in systems having large size and/or mimicking neuronal operations. The biology-mimicking technological operations are in resemblance to the biological operations only when using time-unaware computing paradigm. The difference leads also to the need of introducing "training" mode (with desperately low efficiency) in technological learning, while biological systems have the ability of life-long learning. It is at least misleading to use technological learning methods to complement biological learning studies. The examples show evidence for the effect of transmission time in published experiments.

Computing science is based on a computing paradigm that is not valid anymore for today's technological conditions. The reason is that the transmission time even inside the processor chip, but especially between the components of the system, is not negligible anymore. The paper introduces a quantitative measure for dispersion, which is vital for both computing performance and energy consumption, and demonstrates how its value increased with the changing technology. The temporal behavior (including the dispersion of the commonly used synchronization clock time) of computing components has a critical impact on the system's performance at all levels, as demonstrated from gate-level operation to supercomputing. The same effect limits the utility of the researched new materials/effects if the related transfer time cannot be proportionally mitigated. von Neumann's model is perfect, but now it is used outside of its range of validity. The correct procedure to consider the transfer time for the present technological background is also derived.

The paper explains why Amdahl's Law shall be interpreted specifically for distributed parallel systems and why it generated so many debates, discussions, and abuses. We set up a general model and list many of the terms affecting parallel processing. We scrutinize the validity of neglecting certain terms in different approximations, with special emphasis on the famous scaling laws of parallel processing. We clarify that when using the right interpretation of terms, Amdahl's Law is the governing law of all kinds of parallel processing. Amdahl's Law describes among others the history of supercomputing, the inherent performance limitation of the different kinds of parallel processing and it is the basic Law of the 'modern computing' paradigm, that the computing systems working under extreme computing conditions are desperately needed.

The computing paradigm invented for processing a small amount of data on a single segregated processor cannot meet the challenges set by the present-day computing demands. The paper proposes a new computing paradigm (extending the old one to use several processors explicitly) and discusses some questions of its possible implementation. Some advantages of the implemented approach, illustrated with the results of a loosely-timed simulator, are presented.

Progress is updated in an ongoing research effort to develop new analytical capabilities for the conceptual assessment of rotorcraft life-cycle cost. Recent developments in both modeling and implementation are reviewed. New parametric equations are presented for cost elements related to development and operational phases, and a software tool for graphical illustration of cost estimates is demonstrated on selected aircraft. The demonstration utilizes case study aircraft concepts which emphasize the model’s new features for consideration of electric battery-powered aircraft and assessment cases in government acquisition where validation of estimates against legacy fleet aircraft is required. Near-term plans for additions and updates to the cost model are offered in conclusion.

The increasing interest in understanding the behavior of the biological neural networks, and the increasing utilization of artificial neural networks in different fields and scales, both require a thorough understanding of how neuromorphic computing works. On the one side, the need to program those artificial neuron-like elements, and, on the other side, the necessity for a large number of such elements to cooperate, communicate and compute during tasks, need to be scrutinized to determine how efficiently conventional computing can assist in implementing such systems. Some electronic components bear a surprising resemblance to some biological structures. However, combining them with components that work using different principles can result in systems with very poor efficacy. The paper discusses how the conventional principles, components and thinking about computing limit mimicking the biological systems. We describe what changes will be necessary in the computing paradigms to get closer to the marvelously efficient operation of biological neural networks.

Today we live in the age of artificial intelligence and machine learning; from small startups to HW or SW giants, everyone wants to build machine intelligence chips, applications. The task, however, is hard: not only because of the size of the problem: the technology one can utilize (and the paradigm it is based upon) strongly degrades the chances to succeed efficiently. Today the single-processor performance practically reached the limits the laws of nature enable. The only feasible way to achieve the needed high computing performance seems to be parallelizing many sequentially working units. The laws of the (massively) parallelized computing, however, are different from those experienced in connection with assembling and utilizing systems comprising just-a-few single processors. As machine learning is mostly based on the conventional computing (processors), we scrutinize the (known, but somewhat faded) laws of the parallel computing, concerning AI. This paper attempts to review some of the caveats, especially concerning scaling the computing performance of the AI solutions.

Using extremely large number of processing elements in computing systems leads to unexpected phenomena, such as different efficiencies of the same system for different tasks, that cannot be explained in the frame of the classical computing paradigm. The introduced simple non-technical model enables to set up a frame and formalism needed to explain the unexpected experiences around supercomputing. The paper shows that the degradation of the efficiency of the parallelized sequential system is a natural consequence of the computing paradigm, rather than an engineering imperfectness. The workload is greatly responsible for wasting the energy as well as limiting the size and the type of tasks the supercomputers can run. Case studies provide insight how different contributions compete for dominating the resulting payload performance of the computing system and how enhancing the technology made the computing + communication the dominating contribution in defining the efficiency of supercomputers. The model also enables to derive predictions about the supercomputer performance limitations for the near future and provides hints for enhancing the supercomputer components. The phenomena show interesting parallels with the phenomena experienced in science more than a century ago, and through their studying, a modern science was developed.

The abstraction introduced by von Neumann correctly reflected the state of the art 70 years ago. Although it omitted data transmission time between components of the computer, it served as an excellent base for classic computing for decades. Modern computer components and architectures, however, require to consider their temporal behavior: data transmission time in contemporary systems may be higher than their processing time. Using the classic paradigm leaves a growing number of issues unexplained, from enormously high power consumption to days-long training of artificial neural networks to failures of some cutting-edge supercomputer projects. The paper introduces the up to now missing temporal behavior (a temporal logic) into computing, while keeps the solid computing science base. The careful analysis discovers that with considering the temporal behavior of components and architectural principles, the mystic issues have a trivial explanation. Some classic design principles must be revised, and the temporal logic enables us to design a more powerful and efficient computing.

The payload performance of conventional computing systems, from single processors to supercomputers, reached its limits the nature enables. Both the growing demand to cope with "big data" (based on, or assisted by, artificial intelligence) and the interest in understanding the operation of our brain more completely, stimulated the efforts to build biology-mimicking computing systems from inexpensive conventional components and build different ("neuromorphic") computing systems. On one side, those systems require an unusually large number of processors, which introduces performance limitations and nonlinear scaling. On the other side, the neuronal operation drastically differs from the conventional workloads. The conventional computing (including both its mathematical background and physical implementation) is based on assuming instant interaction, while the biological neuronal systems have a "spatiotemporal" behavior. This difference alone makes imitating biological behavior in technical implementation hard. Besides, the recent issues in computing called the attention to that the temporal behavior is a general feature of computing systems, too. Some of their effects in both biological and technical systems were already noticed. Nevertheless, handling of those issues is incomplete/improper. Introducing temporal logic, based on the Minkowski transform, gives quantitative insight into the operation of both kinds of computing systems, furthermore provides a natural explanation of decades-old empirical phenomena. Without considering their temporal behavior correctly, neither effective implementation nor a true imitation of biological neural systems are possible.

Since the beginning of using parallelized computing units it is known that the actual performance is less than the possible nominal performance: the unproductive part of the computing performance remains “dark”. It is also known that the amount of dark performance strongly depends on the number of parallelly working processing units, so its role must be crucial for supercomputers, where in the coming exa-scale models millions of processors are utilized, as well as for the exa-scale applications they are running, like brain simulation and Earth simulation. Although the effects affecting parallel performance are known from the beginning, their relative weights have been considerably changed with the development of the field and strongly depend on the type of application. For large computer systems the “dark performance” represents a new major obstacle, in addition the former ones, like “heat wall”, “memory wall”, “dark silicon”, etc. The careful reconsideration discovers that in contrast with the general belief, supercomputer performance has an upper limit, and reaching that limit explains some strange and mysterious events, like canceling projects immediately before their target date or that the special-purpose brain simulator cannot outperform the many-thread simulator running on a general-purpose supercomputer.