QAI SPECTRE™ "Brain Chip" Super Turing Automated Machine Synthesis EDA System
QAI LLC has developed its automated synthesis system that generates an optimal QAI SPECTRE™ analog recurrent neural network super Turing machine ready for fabrication.
The QAI SPECTRE™ EDA system is over one million lines of C++ code and runs on Windows 10 x64 platform. System requires minimal hardware background to use. The synthesis system requires two 4K HD monitors for optimal design, modeling and synthesis of analog recurrent neural networks. The resulting output recurrent analog neural network synthesized system is super Turing because it directly computes using irrational numbers with neuron high analog adaptive fan in and fan out, implementing the Siegelmann-Sontag hypothesis, which was the design goal for our "Brain Chip" synthesis system development.
QAI SPECTRE™ super Turing analog recurrent neural network machine performance exceeds current and future P-Space reducible (Turing) machines (including quantum computing machines) because P-Space reducible machines are restricted to rational number computation. QAI SPECTRE™ super Turing analog recurrent neural network machine directly implement irrational number computation resulting in vastly more degrees of problem solving freedom. This results in QAI SPECTRE™ super Turing analog recurrent neural network machines solving NP-Hard problems in polynomial time. In contrast Turing machines bog down or become stuck in local minima false solutions. QAI SPECTRE™ super Turing analog recurrent neural network machine tunnel through all local minima.
The output of the synthesis system is the application specific analog recurrent neural network 3D super Turing GDSII (OASIS under development) "Brain Chip" file of analog unclocked asynchronous GaAs 3.5 Schottky diodes and GaN 2.5 HEMT transistors running at 3 THz and up to 500 GHz respectively. The GDSII file implements a 4D layer multi-chip module on micro cooled diamond with 15 die per layer. Within the design synthesis environment, wireless sub millimeter wave diode and transistor (the analog neurons) dynamic interconnection enables many additional degrees of problem solving freedom due to virtualized high fan in and fan out of the GaAs/GaN analog neuron interconnect. The wireless neurons are implemented by physical modification of the gate foot metalization layer on a neuron by neuron basis. This semi-conductor process modification results in an EHF broadcast transceiver, which can be tuned to different EHF frequencies within our synthesis tool. Neuron wireless frequency modifications are reflected in the GDSII file.
Wireless neurons implement quantum diffusion signaling, much different than classical diffusion because once a wireless signal is detected by a semiconductor neuron, in receive mode, the signal is ignored by any other neuron, unless the signal is simultaneously detected by a plurality of neurons in receive mode. The result in a novel self-organization of the neurons, first demonstrated in our QED quantum chemistry modeling and design in 2001.
The GaAs/GaN “neuron” semi-conductor implementation is heat resistant, self-healing (due to the known GaN semiconductor properties) and radiation hardened for deployment in space or harsh environments where survivability is demanded. Power use is low at about 1 Joule of energy per MCM-D layer. Neurons can be configured in the synthesis system to have wireless connectivity between MCM-D layers. Performance exceeds 500 quadrillion operations per second per integrated circuit die recurrent neural network. The GDSII file generated by our synthesis system is next full EM simulated within the Cadence Virtuoso layout platform. Tuning to using the Virtuoso layout platform is reflected in a new GDSII file, which is imported into the Keysight ADS simulator for end to end simulation including exothermal modeling. A future version of our synthesis system will support import of the final ADS simulated recurrent analog neural network super Turing design. ADS design rule checked output is then sent to the pure-play or IDM foundry for fabrication (tape-out etc). The wafers are 4-6 inches with 90%+ yield. Foundry turn around, including wafer test, is roughly two days. The wafers are cut using an EM beam or diamond saw. The known good die are tested using die test Keysight broadband scope with frequency extenders and VNA.
QAI SPECTRE™ super Turing analog recurrent neural network machine performance exceeds current and future P-Space reducible (Turing) machines (including quantum computing machines). Design is validated via Keysight ADS end to end analog simulation of the super Turing machine analog recurrent neural network design prior to tape out.
Pricing for QAI SPECTRE™ super Turing analog recurrent automated synthesis system is 8,675,660.00 EURO per seat, including required yearly QAI SPECTRE™ support contract to selected partners. Visit our partner page or contact QAI LLC for more information.