Tipalo® - acronym for Time based pattern logic
This is Tipalo in a nutshell, see pic below
Company overview
INTRODUCTION
Tipalo GmbH is a Swiss LLC near Zurich, a fabless AI startup with own software to pioneer logic applications.
We develop COGNITIVE AI as digital brains with an Artificial Nervous System similar to biological brains.
Our AI technology is used for the autonomous control of "bodies", physical as buildings, vehicles, robots or virtual as avatars.
These "bodies" must be equipped with digital devices, that act as sensors, actuators and their own organs.
AI APPROACH
Our AI technology is self-learning by collecting information that is later converted into knowledge areas.
The Tipalo approach to AI is very simple: we take biological intelligence as a template, means the living brain.
It needs a different mindset to comprehend and represent parts of the real world and replicate the human mind.
Our understanding of intelligence is based on logic, our AI concept contains objects reacting in time and space.
Tipalo COGNITIVE AI as digital brain can control any type of connected body, real or virtual
EDGE AI USE CASES
Tipalo target market is COGNITIVE AI, a brain for a corresponding body and performing intelligence for:
Level 1, as managers for smart buildings + robots for indoor manufacturing and outdoor surveillance
Level 2, as pilots for autonomous vehicles of any kind, e.g. terrestrial, naval, aeronautics, space
Level 3. as autonomous robots, e.g. for workers in hazardous environments, for avatars in virtual landscapes
CLOUD AI FOR TELEPRESENCE
Tipalo brains can be connected via Internet to any embedded system or robot body, real or virtual,
by using a connection with high bandwidth and low latency enabling real-time, like 5G or fiber glass,
Furthermore, the accumulated experience of many individual brains can be aggregated on demand,
enabling a collective knowledge database for a certain body type, e.g. autonomous vehicles
Tipalo AI - technology characteristics
ACTUAL PROCESSING
Tipalo AI is used in real-time, using very low working frequency paired with very low power consumption.
1. Real-time processing, where the reaction time of the digital brain is 1K faster then the biological brain
2. Working frequency is at 50 MHz, while using direct VHDL implementation for concurrent processing
3. Power consumption is less than 100 W, using high capacity FPGA SiP with on-chip integrated HBM2
FUTURE PROCESSING
Tipalo AI intends to migrate the implementation from FPGA with HBM2 to own ASIC
1. The ASIC will be manufactured using FD-SOI technology, enabling radiation resistant chips, i.e. 22FDX
2. The ASIC will use only persistent memory, eliminating any type of volatile memory, i.e. SRAM, DRAM, etc.
3. The used emerging persistent memory will also be radiation resistant, i.e. MRAM, CeRAM, 1T FeFET, etc.
The unique and defining feature of Tipalo AI technology is AUTONOMY, at 3 levels
Level 1. BODY
The body is the first decisive argument for autonomy, as it enables to do everything on its own.
Cognitive AI requires a body hardware with sensors, actors and internal organs, like any living entity.
Level 2. MIND
Next level of autonomy are spiking neural nets, using distinct groups enabling interactive algorithms.
This implies that only some neurons are needed for a specific purpose, allowing explainable AI.
Level 3. KNOWLEDGE
Another level of autonomy is the software, as synapses between neurons, which implies no code.
Synapses are used as universal format for any purpose and enable the accumulation of knowledge.
Tipalo AI - technology overview
AI IMPLEMENTATION
Tipalo AI enables the simulation of different brain regions in real-time, all the time and stand alone,
while being connected to a certain body hardware, which is equipped with sensors, actors and organs.
The VHDL implementation of our AI model allows massive parallel execution, without any processors.
This implies we do not use any math/statistics/algorithms, while no bias/training/inference is required.
AI KNOWLEDGE
We use 3 intelligence levels, which have a template in nature: insects, fishes/mammals/birds and primates.
They reflect the amount of neurons and their synapses, from 1M with 16, via 1G with 256, to 10G with 1K.
Each level has predefined knowledge, to manage the body and the mind, together with some genetic memory.
The AI explores autonomously the environment and stores its own experience as accumulated knowledge.
AI PLATFORM FOR DIGITAL BRAINS
Tipalo develops strong AI with cognitive features creating hereby digital brains with:
1. Real-time operating system in VHDL, enables simulation of brain tissue via Programmable Neural Nets (PNN)
2. Self-Learning Mechanism (SLM) enables knowledge accumulation using Self-Associative Memory (SAM)
3. Artificial Nervous System (ANS), contains the brain landscape as map of all synapses, see connectome
AI LIBRARIES FOR NEURAL COMPONENTS
Tipalo develops AI libraries with configurable reference designs based on PNNs for different ANS levels as:
1. Neural drivers for interfacing sensors, actors and organs of any kind, e.g. visual, limbs, rechargeable battery
2. Neural applications for various knowledge areas using SLM, e.g. identification, locomotion, task workflows
3. Neural storage for different memory types based on SAM, e.g. short-, mid- and long-term memory
AI COMMUNICATION
Tipalo AIs can connect with each other, via cable or wireless, to communicate for several purposes:
1. positive identification, to enable the coordination of their actions for a possible common task
2. certain interaction, to enable the extension of knowledge by teaching each other new things
3. direct exchange, to enable the aggregation of information for a certain area of knowledge