AI v. a kitten: or what is AI cloning?

Some time ago my niece  saved a little kitten. She found him on the street and took home. Now he is a happy strong 6-month old hunter.

The other day he went into a yard and saw a bird on the ground. The moment he saw a bird his whole posture immediately changed. He almost fell on the ground and started slowly crawling toward the bird. His years were trembling; he was slowly gliding along the ground; only his elbows were moving. Kind of like in this pictures.

How did he know how to hunt?

No one taught him.

He was taken from his mother and left on a street when he was very little and naturally my niece didn't teach this kitten how to do it.

So where does the hunting knowledge come from?

The answer is obvious. That knowledge has been stored, built-in in his DNA code. The DNA has the information, not just about the bones and muscles and how the body grows and structures; but also the information about the behavior in specific situations. There are actions that get ignited automatically because those reactions are built-in in the DNA, and hence no one has to teach when and how to activate those actions.

When a kitten is born, his DNA is not a “tabula rasa”.

When AI is “born” it is completely empty. People have to invent and then employ a specific training protocol to teach AI how to do some specific task – one specific task, like recognizing words, or recognizing pictures.

And when one AI is trained to do that, there is no way to just replicate this AI into another AI system. Another AI that needs to do the same that must be trained from scratch.

On the contrary, if we would take a DNA from a kitten and cloned it the new kitten would have the same “skills (i.e. reactions) the original kitten had.

AI has no DNA like that.

No one can clone AI. Every new functioning AI that has to function like already functioning AI has to be trained from scratch.

To my best knowledge, no none in AI field is even working on cloning AI. No one works on “AI DNA” that can be reproduced so the training for new AI system in the same field would not have to go from scratch.

The search does not give anything on this. 

There is a term “digital cloning” but it means using one specific AI to make copies of digital objects (sounds, images).

That is why using AI is too expensive. Every new AI that needs to be used in the same field with the same functions (i.e. image recognition) has to go through a long and extensive an expensive process of training. The development of that process requires highly skilled professionals. But even then many AI systems make mistakes or have some bias. 

That is why anything your read about AI is GREATLY exaggerated.

Things will not change much until the invention and perfection of AI cloning, that will eliminate the extensive training for AI working in similar fields.

But so far the field does not even have enough people who can developed and execute an effective AI training.

For more on what is actually going on in AI field: https://www.cognisity.how/p/arti.html

Dr. Valentin Voroshilov