In traditional ML computer systems, the processing is very simple, see following steps:
calculate numbers between different components in different steps by using various algorithms
then evaluate the end result with a pre-defined decision as 1(=true) or 0(=false).
ML is calculating the intermediary steps more or less arbitrary and a human says OK or KO.
This process holds true to many successful tasks, e.g. identification of an object type.
As an example, in order to determine if a picture contains a cat,
the system is feeded with different images, while for each image
a human decides whether it is or it is not a cat, namely as 1 or 0.
Therefore, this kind of training is called ML Machine Learning.
But obviously it has a lot of disadvantages, such as:
one neural net can only be trained for only 1 very specific object/task,
while the amount of data, aka pictures, is really huge, means millions.
Tipalo biomimetic model
Our biological inspired technology enables a total different approach.
We create many different neural nets, which work like biological neurons.
First of all, they learn fully autonomous and the AI learns only by
reacting with the corresponding body to different external stimuli.
So, our biomimetic model needs sensors and actors connected to a certain body,
which on its side, has a pre-defined structure, according to its species.
Depending on the environment, where the body is, it will customize itself
by sensing the contained object via its sensors and so, it learns by itself.
The reaction to the external stimuli is called feedback, which implies
reacting to an object will cause a reaction, which on its side will cause
a judgement, such as it is ok or not ok to do it, this way the embedded organism learns.
This is our biomimetic model, which we digitally implement to create a general purpose AI.