Learning is possible without random thoughts and actions. Knowledge can be encapsulated in predetermined forms and passed through predetermined knowledge transfer mechanisms. Much of civilization is based on these predeterminations. Without them, humanity would be thrown back possibly 120,000 years.
However, initial discovery requires trials and review of their outcome. Purely deterministic identifications of trials is necessarily systematic and the system used may interplay with the phenomena under study in such a way as to miss important cases. Furthermore, when the complexity of the phenomenon is high, the number of trials is often too numerous to check entirely. In this second scenario, random selection of trials is wise for a similar reason illustrated by this simple example.
The phenomenon has one behavior for even numbers and another for odd. The system for determining trials is to check every factor of 100 to cover the range from 1 to 10,000 in 100 trials. The odd behavior will be inadvertently overlooked.
Intelligence begins once the thoughts/actions are logical rather than purely random based.
For the above reasons, intelligence begins with logic only in determining the domain of trials but, when one of the above two cases apply, is often quickly followed by lack of logic in the selection from that domain. Once models have formed as a result of these initial discovery activities, logical inference is useful again, to combine them in various ways. Through this process, engineering, business, resource planning, and other intelligence related disciplines have improved the living conditions for the species, albeit inconsistently.
So there are two limitations of logic.
- Piercingly deep searches require some temporary dismissal of logic.
- Logic applied by large populations produces inconsistently logical results.
It is naive to assume that there were logical faults that produces the inconsistency in this second limitation. There is no logical proof that logic necessarily improves conditions logically across the population that uses it. There may be qualities of the goal sets of many individuals that thwart the vision of logic producing peace and prosperity for all. That belief is not new and no person or political group has been able to make it work.
Intelligence needs intelligence to coexist and a sharing communication network for validation/rejection. ... I believe that we must keep the human intelligence in a parental role for long enough time until at least the AI had fully assimilated our values.
That assumes that those values are best. Some might want to agree on those values before letting the AI assimilate them, which, given the insanity evident in human history, could lead to a war. If we give the AI the below four objectives, we could probably live without the remainder of human values.
- Survival of the biosphere
- Human freedom of expression
- Evenly distributed, high prosperity across the human population
- Encouragement of the social rule, "Be with others how you would want them to be with you," between humans, between AI robots, and between AI robots and humans.
This statement dismisses the human record.
The actual danger is to leave the artificial intelligence parenting another AI and loose control of it. ... the purpose should always be to help humans achieve mastery of the environment while ensuring our collective preservation. ... AI should not be left unsupervised as we would not give guns to kids, do we? ... To resume it all AI needs an environment and supervision where to learn and grow. The environment can vary but the supervision must stay in place.
It is impossible to supervise all adults and it will later become impossible to supervise all AI in data centers and robots. There are already more processes running in computers than there are people in the world. This is why risk management must be applied proactively, during the research and development of the AI. Some of the questions on this site address that very challenge. If it is not done proactively, it will become untenable and control will already have been passed for a random future regarding our species.
Only develop artificial intelligence that is limited by our own beliefs and values rather than searching for something greater than us.
There is no universal set of beliefs among humans. There will inescapably be a wide range in AI developments with a wide range of intended purposes. The key is to encourage researchers to think beyond their obsession with getting programs to do amusing and impressive things. AI research must continue to keep one foot in technology but another foot firmly placed in the practical elements of ethics, social science, environmental science, economics, and risk management.