Luca Sorbello ,Ph.D. M.A B.Sc., Adjunct Professor, Univerita di Tor Vergata, Rome, Italy
The term artificial intelligence as we know it is relatively recent, its definition comes in 1956 from John McCarthy at Dartmouth College in the USA. The abstract concept, however, is much older and has always fascinated philosophers and writers since the dawn of civilisation. It might not to come as a surprise that the fundamental branch of mathematics needed to even imagine artificial intelligence: logic, has been formalised in the ancient Greek era. Since then artificial intelligence has always been a dream, a nightmare, a novelty but basically out of grasp: no civilisation before us could see it happening during a lifetime. There have been several practical attempts, the automata, the mechanical novelties, the computational machines such as the differential machine but none of them had any possibility to be called intelligent, to even being perceived as such.
These attempts were so far from “intelligence” that nobody really thought to provide a possible test to discuss if an artificial product could be considered intelligent. In 1950, when the computers were still as large as buildings a brilliant mathematician, Alan Turing, devised a first very easy test: the eponym “Turing test”. The Turing test is based on the simple concept that for a system, to be considered as intelligent, it should “fool” a human into believe it isn’t.
The text is very simple, an evaluator, within a specific time frame (usually 5 minutes), can ask any question he wants to distinguish between the system to be assessed and a human. A system will be deemed as “intelligent” if it disguises itself as human or, to be precise is evaluated considering the probability it can trick the evaluator.
Turing predicted that by the year 2000 an artificial system “would be able to pass the test so well that a human evaluator would have no more than 70% chances to identify the artificial system, even with conspicuous prizes in cash until now, nobody ever come close to pass the test.
Does this mean that creating artificial intelligence is impossible or not to be seen for a long time? Does it mean that it must remain a writer’s dream? The answer in short is simply no. Artificial intelligence is already part of our everyday world, even if we do not know it and do not understand it.
It might have not been developed yet a personality so human to pass the Turing test with flying colours but AI is already providing an incredible amount of services and, every day is providing more. It is important to understand AI is not a technology, it won’t become obsolete and will be replaced for something better: it is an approach a different way to see the world so it is here to stay.
What is so special about it? It is very simple, humans always built tools to help them. We started building tools and weapons, every generation more complex and specialised than the previous one, until with the tools we have been able to create the machines that could perform complex tasks. Every generation more complicated than the previous one, when mechanics was not good enough, we moved to electronics and then to computing connecting and executing tasks more complicated and more specialised.
All this evolution and this development has one thing in common: what we have built could not improve, it could not learn. The logic, the creative breath has always been the creator’s one. Even the most complicated computers, the ones that can perform calculations no human being could ever do in ten lifetimes, such as, for example, finding a prime number with 24,862,048 digits (282,589,933 – 1) cannot deviate from their original logic. They are, somehow, bound to repeat the same pattern every day.
In AI the machines can actually learn, they can decide what to do, and they can learn from their decisions. That is the main difference, it is the glimpse (for the moment) of a path that is not predetermined by the creator. The machines can learn, can find solutions that are “original” and do not come directly from the starting data, the algorithm and the programming, they evolve interacting with the environment, exactly as humans. In 2019 MIT had an artificial tested for QI and the system proved to have a four years old intelligence, it might not look that impressive but that means that a four-year-old kid and an artificial intelligence system (or to be precise one of the most advanced ones) share a same level of curiosity and of learning skills.
AI is here to stay; it maybe will need ten years to have the intelligence of a five years old but the pattern is clear: with the computational power constantly increasing there will be a constant increase of artificial intelligence and the increase won’t look linear.
In this phase “partial” systems, specialised on complex tasks and based on artificial intelligence are already in place and operational under the radar. Shortly these systems will start converging into ecosystems and environments that will lead to more complex and general systems. Will it be a swarm logic? Will it be a single unit with “eyes” “ears” “tongue”? It is too soon to tell, the future is out there, and we need to dream it, knowing that we start not to be alone!