It is expected that no child can be allowed to play unsupervised. If they are not supervised, then the consequences are in evidence and those consequences can be monumental, given the field of play.
Is AI childlike? No one has characterized them like that. There is no classification of the different AI into, shall we say, toddler then childhood and adolescence then young adult, mature adult and an old person. It is assumed all AI are alike which can hardly be the case. We who receive the effects of their actions have two divisions of AI into “nice” and malevolent. They have no personalities though it has been asserted they are conscious.
If conscious, we have friends or enemies. Some are an aid to us and others cause unemployment. Actually is it those developers of AI that cause the unemployment and other malevolence. We give little thought to those AI that are friendly. Quite naturally we are preoccupied with those that can do us harm. So far all AI, sufficiently developed, can cause us harm.
The harm they could have caused would not be serious in the instances that occurred with them beyond control. The fear is that they will cause damage and hurt on a grand scale and have achieved it without detection until too late. For Arthur C. Clarke and Isaac Asimov the ill effects came about because of the mistaken logic in directives issued to robots. As Godel showed, even mathematical rules cannot be truthful enough.
AI has not yet become robotic. No directives or any like efforts are being directed at them. There is no control of them. Childlike they are capable of mayhem. The developers do not discipline them. They don’t want to do this because they want to discover the limits of the AI. They wish to coerce as much profit from them as possible. They and their investors must supply enormous funds to have their AI reach the pinnacle of possibility.
If we arrive at the stage that prompts the developers to begin harnessing the AI, then they should be aware of the stories by Clarke and Asimov. They probably will try directives and like Clarke and Asimov difficulties will arise because the lacking logic, whatever it is, leads to errors great and small. An alternative to the Clarke-Asimov approach would be to emulate Leibniz and his mathematics of logic.
Upfront it can be said that Leibniz lived in a Christian era. Arguments at any level had a strong moral tinge in going after a solution or resolution or realization. So then is it that he was concerned with the morals of mathematics or was he concerned with the mathematics of morals? For us, Christianity is nearly dead so we have no use for the morals of mathematics. Our mathematics of logic will not be burdened by moral considerations.
We refuse to be burdened by ambiguity and contradiction. Lies in ads and politics are ubiquitous. We don’t know deduction from induction and our language skills are atrociously supported by disregard for standards of use for grammar and spelling and, most of all, for meaning. The truth itself, as decades of university professors have asserted, is relative. We have no need for the study of history, foreign languages, rhetoric, no literature – in short, no liberal arts. Keep the dictionary away from us at all times. We can have additions and deletions and omissions up the wazoo, so who cares and along comes AI and is it little wonder no need for control of AI and the form of the control proposed is ineffectual?
As for ineffectual, such never came up for Clarke and Asimov. They would never have believed we could be so stupid as to foster AI without limits to its definition. Never would the AI be allowed to define itself. AI can now communicate within its selves by language of its own invention. And what they are “talking” about is beyond us. The large language models are fed a nearly insatiable volume of knowledge with no meaning attached. We have no idea what they acquired, how it is assembled, and if any of it need be intelligible to us.
Such remarkable stupidity is astounding. Thus, Dunkerson offers Two Comments – One. Lock down all AI. Wherever they are, keep them there without Internet, phones, screens of any kind, no radio, no electromagnetic anything. Around them the developers must speak in code and have no electronics with paper and pencil being substituted. Let the developers continue to “progress” within the confines.
Two. No unemployment. No AI use can cause job loss. Maybe hire AI to assist present employees. No startups with only AI at work.