Supraelectronic Analog Machines for adaptive AI

Speech Recognition | Natural Language Understanding | Computer Vision

Analog Computation Corp is a U.S. defense contractor with partners throughout North America and the EU. We focus on design and development of multi-chip modules implementing application specific super Turing analog hybrid (analog/digital) turnkey machines for use in space, defense, IoT and commercial industrial applications.

Our focus is development of analog systems that convert quantum and digital calculations into efficient analog circuits. This enables a high artificial intelligence that we refer to as adaptive AI.

Supraelectronic ANNs exceed current ANN implementation by using uncomputable Real number degrees of problem solving freedom. This enables supraelectronic ANNs to solve problems that current ANNs cannot, such as eliminating bias and enabling indefinite resolution and scalability. Critical use case examples include speech recognition, natural language understanding and computer vision.

Our analog super Turing machine computation utilizes non-computable real numbers that are combined with new adaptive non-deterministic analog device circuits and III-V composite semiconductors. The result is multi-chip module SiP machines implementing III-V ultra high performance solutions to DoD and industry challenges.

Machines are developed using our company proprietary automated III-V SiP semiconductor process engineering synthesis technology. Our turnkey SiP solutions are fabricated using III-V U.S. and EU semiconductor process foundries and partners. We excel at mission critical module development solving the most complex space and defense application challenges for both the U.S. and the EU.

What is supraelectronics?

Analog computation in the distant past was used to implement digital computation (using operational amplifiers for binary arithmetic, for example). Yet with the advent of the transistor decades ago, all this changed. Now transistors are used to efficiently implement operations including addition, multiplication, subtraction and division for both integer and floating point math.

Yet this is grossly inefficient on many levels in one area of computation, i.e. artificial neural networks (ANN) and machine learning. ANNs are the basis for much of the Artificial Intelligence applications and research today, from playing the game of “Go” to speech processing, image understanding, solving exponentially hard embedded computational challenges and far beyond.

Conversely, brain neurons are analog devices; they do not implement the machine operations of digital arithmetic to accomplish their functionality. 

Supraelectronics: Analog Computation Corp develops analog neural networks and machine learning semiconductor systems operating at three terahertz and beyond, to function like brain neurons, i.e. without digital addition, multiplication, subtraction, division or any form of memory based table lookup. Powered by less than a digital watch battery, resistant to radiation and self-healing semiconductor technology, our III-V composite semiconductor ANN chips and high density SiP modules depend only on the natural current flow of the semiconductor devices we utilize, which like brain neurons naturally implement the functionality of ANNs without any form of digital computation.

Our automated analog ANN proprietary EDA synthesizes sophisticated analog ANNs, which implement all forms of machine learning, including reinforcement learning, for our customers in diverse applications. Much differently from current ANN machine learning systems, our unbiased ANNs deploy and learn continually from experience in operation, resulting in adaptive AI.

~ ~ ~ ~

The Analog Computation Corp difference:

1) Wireless semiconductor neurons

2) Analog Real number resolution

3) Adaptive ANN chip synthesis

adaptive artificial intelligence

Our supramolecular ‘circuitry’ of hemoglobin

(See details in caption below)


Work performed in the life sciences spanned 2001-2015. We have ton of examples of supramolecular systems representations whereby the femtometer resolution combined with virtually indefinite scalability is entirely unique. According to DuPont: we have a potential "disrupting technology in the drug design industry” and we “have developed new algorithms and tools that allow for direct visualization of the sub-atomic interactions that makes [our] tools so unique.” And.. “The use of these visualizations have allowed us [DuPont] to have meaningful conversations about the results and to see the important features and relationships between the molecules. These specific visualizations are unique to [us], and provide a view of the intermolecular interactions that is not available with other tools.