AI is difficult and complex, because, even if we know today how a single biological neuron is working alone,
there is no explanation how biological neural nets are creating and developing higher forms of organization

Here are the "treasures" of our company as soft IP - Intellectual Property

 1. AI knowledge - status ready

  a. AI theory, taking the biological brain as a template
   b. AI model,  creating a digital biomimetic model of a brain, based on ANS classification, e.g. templates
   c. AI structure, quantifying the different brain tissues into different types of neural nets
   d. AI framework by AI levels, which imply a digital brain also develops itself in time
   e. AI levels, comparison with human brain development at different ages, e.g. at the end of year
      - 1 month, 6 months, 1 year
      - 3 years, walk + talk + play
      - 6 years, kindergarten
      - 10 years, primary school
      - 14 years, secondary school
      - 18 years, high school
      - 23 years, degree
      - 28 years, job experience
      - 30 years - dedicated social experience

2. AI semiconductor technology in development, beta release expected in 2021

  a. AI operating system -in development, beta release expected in 2021
           simulate brain functionality as a network of neural nets connected to
           sensors, actors and internal organs, forming an ANS for an embedded system

      - written entirely in VHDL enabling massive parallel processing
      - external interfaces - Ethernet only
      - internal interfaces - HBM2, evtl. DDR4 in addition
      - applications - neural nets only, as self-programmable dedicated Tipalo neural nets
      - self-programmable dedicated networks on chip - connecting all neural nets including the i/o neural drivers
   b. ANS - brain functionality with specialized regions, as ANS template + time development
      - neural driver input             : sensor cam as eye, e.g. cerebral cortex for visual system
      - neural driver output           : actor screen as face, e.g. cerebellum for body reaction
      - neural net applications     : reflexes as pre-defined connectivity, e.g. brainstem
      - self-learning mechanism : at different levels, e.g. hippocampus
      - accumulated knowledge  : depending on sensors and actors, e.g. neocortex

3. AI semiconductor implementation

  a. FPGA SiP with integrated HBM2 : XCVU37P in development, beta release expected in 2021
         - different FPGA boards  

      FPGA software,   in development, beta release expected in 2021 
  - encrypted FPGA bitstream with VHDL operating system

     ANS software, in development, beta release expected in 2021
         - encrypted data as ANS with corresponding applications, according to species, see ANS classification
         - encrypted data as instincts and pre-defined knowledge, according to species
         - encrypted data as knowledge accumulated via the self-learning mechanism
         - encrypted functionalities according to training program and environment

b. 3DSoC - on hold

This step is on hold, until the 3DSOC technology is proven + available to all commercial customers worlwide !
         - 3DSoC - three Dimensional monolithic System-on-a-Chip, with 3D Carbon nanotubes logic and RRAM
         - this revolutionary technology is currently transferred into production exclusively to Skywater foundry
         - we are already in contact with Skywater, in order to check a conversion from FPGA SiP to 3DSoC

For further and detailed information on 3DSoC, kindly visit Skywater corporate site, see link below: