It's an interesting theory but I can tell as a computer programmer and developer that AI is impossible.
If what you mean is TI (True Intelligence) and not AI (Artificial Intelligence) then I agree.
There's a reason why it's named as Artificial, because it means that it is fake. Artificial Intelligence means fake intelligence. It's not intelligence but it can be perceived as intelligence (by human) thus it's named as Artificial Intelligence.
The same as "Pseudo random" in computer lingo, it's not actually random but it can be perceived as random by it's observer (human) because true random doesn't exist.
And I will go as far as to say that even if the human race existed for all eternity we may never reach the ability to produce a true AI. Human beings will never be able to design an AI because
Agree, human will never have the ability to produce a true intelligence.
In order to have true intelligence, human needs to produce creativity within the machine.
In order to produce creativity human need to tolerate the machine to be inconsistent, to venture forth from it's given framework and boundary condition which beats the purpose of building the machine in the first place. And the mechanism to produce such inconsistency is unknown.
It relate to the topics of producing "true random" as mentioned earlier, as long as "pseudo random" is being applied then true inconsistency cannot be achieved.
Imagine having an iPhone that no longer wanted to function just because it's in the bad mood?
Or it suddenly invented a virus on it's own in order to gain control of another iPhones?
Another example is in the case of "AI for self driving cars" in a condition where 3 possible outcome need to be chosen:
A. Turn a sharp left and crash to the pedestrian. (probable results: death of pedestrians)
B. Stay on course and let the suddenly appeared motorcycle to crash to you (probable results: death of the motorcyclist)
C. Turn a sharp right and crash to the river (probable results: self annihilation)
The programmer of the AI need to set a condition (weight factor) which outcome will be "more preferable". Survival of the self? Safety of pedestrian? Safety of another road users?
Once such condition is set, the AI will behave constantly following the condition set by the programmer.
OR
The programmer of the AI can apply "Pseudo Random" to set the weight factor thus sometimes the AI will choose A some other time B or C based on the current date and time when it happened (thus labelled as "Pseudo") But such decision will render the AI to be "unpredictable" thus "unsafe".
- You cannot program what you do not know or fully understand. When a programmer programs he needs to have a blueprint. What are the mechanics of consciousness? What are all the processes that occur to make a being fully conscious? Science does not have a proper answer to any of the above questions. Therefore if science doesn't have the answer to this question how would a programmer know where to start and how to build a true AI without a blueprint of consciousness.
The true intelligence will need to have the capability to write it's own blueprint and discover it's own consciousness by itself.
- For you to develop a true AI you need a highly advanced programming language. None of the programming languages we have can make a true AI because all of the programming languages that exist will only allow you to create programs that are programmed by programmers. Everything is under the control of the programmer. A true AI must be developed by a programming language that can live, breathe and evolve.
The true intelligence will need to have the capability to write it's own programming language, to devise it's own mechanism of hardware, to go beyond current capability and limitation of boolean logic math, to go even beyond applying the quantum probability and pairing mechanism for computation.