János Végh

János Végh, Ádám J. Berki

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.

János Végh

The experience shows that cooperating and communicating computing systems, comprising segregated single processors, have severe performance limitations. In his classic "First Draft" von Neumann 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 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, their achievable payload computing performance is conceptually modest. The type of workload that AI-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. 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; derives an appropriate interpretation and handling for Amdahl's law. It shows that, in that interpretation, Amdahl's Law correctly describes ANNs.

János Végh, Ádám J. Berki

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.

János Végh

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.