Super Turing Machines
LARSX has extensive patented experience in machine learning and advanced ASIC design including analog and mixed signal systems. The company focus is developing Super Turing Machines, implemented using high frequency patent pending analog recurrent neural networks. The company pioneered the patent pending "SPECTRE" Super Turing machine architecture, which has many reductions to practice.
The LARSX company focus is to license and develop SPECTRE Super Turing field of use ASICs/modules.
SPECTRE is a super Turing analog recurrent neural processing architecture. SPECTRE semiconductor analog neurons operate at 3 THz asynchronous unclocked frequency. SPECTRE direct band gap semiconductor neurons are are wireless with dynamic interconnect through modification of gate level structures. SPECTRE neurons dynamically self-organize into diverse recurrent neural networks.
SPECTRE ASICs/MCMs are implemented using GaAs/GaN semiconductor processes. SPECTRE multi-chip modules require one Joule of energy to operate. Operating at Terahertz frequencies, SPECTRE module Super Turing recurrent neural networks can easily exceed over 500 Quadrillion operations per second.
SPECTRE Super Turing machines obsolete digital computation because SPECTRE implements complex imaginary number computation with high fan in and fan out, mimic brain neurons. Consequently SPECTRE Super Turing computations exceed current digital computers since digital machines are restricted in the Turing limit to calculations involving rationale numbers. SPECTRE degrees of problem solving freedom far exceed digital computation. Unlike current digital machines, whose processing and memory are separate, SPECTRE analog recurrent neural networks combine memory and processing seamlessly.