The current huge new wave of AI and Deep Learning has been recently revolving around the topic of children, how they play and acquire knowledge and intelligence. This is not necessarily a brand new phenomenon, as it has been known and appreciated before, including cognitive scientists who recognized such processes in small children are key part of intelligence. Alan Turing wrote and was fascinated how children learn and considered it to be key for machine intelligence. He mentioned ‘childhood’ of artificial intelligence, which would be a phase analogous to humans, where processes such as learning, reasoning and critical thinking are formed. A nice popular illustration of this concept is in the book “2001: A Space Odyssey” by Arthur C. Clarke and the famous movie by Stanley Kubrick based on it, where one of the main protagonists, HAL 9000 computer representing superior AI, is indeed trained from childhood by his creator Dr. Chandra.
Virtually all of the current AI has been preoccupied with this problem, as a de facto necessary condition for achieving true AI. Its latest incarnation is in a form of solution to producing enormous labeled datasets for training in supervised learning. Many scientists, both in AI as well as cognitive sciences, reject this notion that true AI inherently requires immense training sets, because children do not. And this is indeed a powerful point, as it is truly fascinating how quickly children learn, from seemingly chaotic external world, with very few guidances and aids.
But we should not go overboard and get carried away, as early children learning, including adults too, is only about common sense and concepts in real world. There are some abstractions too, such as justice, fairness and care and respect for others but it is interesting there are adults, including some very successful ones, are lacking in this respect and can be viewed as mild, or oven more serious, psychopaths and sociopaths.
But genius is never about common sense, it is instead about always probing and going beyond at the level of abstractions that strain credulity and understanding. Some become accepted and commonly understood such as e.g. limits and theory of gravitation, though others still seem incredible and hard to believe such as e.g. general relativity and quantum mechanics. Even for the smartest of geniuses, it takes a lifetime of thinking and seeking new ways to come up with their discoveries, almost invariably going against the grain of common sense. And even the commonly accepted abstractions such as calculus and limits are only well understood by highly educated and trained. How many people would know how to answer Xeno’s paradoxes even today? The answer is surprisingly few.
This phenomenon is not only about narrow abstractions such as mathematics, it is much more pervasive and widespread. Consider a simple and straightforward term such as ‘bank’ which should have an easily understood meaning , as an institution where one deposits their savings. But such meaning is actually rather shallow and naive as banks are much more than that, as anyone with deeper understanding of finance knows. A phrase such as ‘bank run’ is puzzling from naive understanding of how they work, as how can there be a run when they are supposed to hold what people deposit? The answer is that they do not keep all the savings, and this is where the real story only begins.
These layers of deeper and deeper meaning and understanding are completely universal and apply to everything. They are key to genius, as they are such because they are able to go in and understand things deeper than anyone else. And this is not what children learning and knowledge acquisition is about, it is really the opposite. Note that the example of a bank did not come from finance by accident. In contrast to fields such as mathematics, or even music, where there were many cases of geniuses making great contributions at early age, finance is rather unique and special in that it seems to require a lifetime of learning, practicing and observing just to get into it and start gaining deeper understanding. Geniuses such as Tesla, Einstein, Newton, Gauss, Gödel, Galois, Mozart and others made great discoveries and masterful acts early in their lives. But finance is full of misconceptions, deceits, disinformation and delusions that require much more time to penetrate and sift through.
This is why superior AI at genius levels will not be about training machines form early age as children and somehow just getting them to distinguish themselves as geniuses. Instead, it will be about getting access and being preloaded, preset and initialized at creation with immense bases of our accumulated knowledge. Did HAL 9000, or a future machine like it, know about all of mathematics, politics, history, economics, arts, finance, engineering? And if so, how did it acquire it all? Surely not by Dr. Chandra, or somebody like him, teaching it all of this knowledge, for such a process is completely infeasible. One may say superior genius AI will just read it all, but the process of reading is very slow and inefficient, as it takes not only a long time to carry out, but even longer to comprehend, assimilate and remember what was read. Will this great AI need to do the same?
Our answer is that it will not, instead all this immense base of reasoning and understanding will be created together with the AI itself. It is this base that is key for all intelligence, both human as well as artificial, at all levels including that of genius and beyond.