So fascinating it is to us; to watch chimps using rocks to break nuts open, or crafting a probe out of a leafy twig to extract termites from their mound. Dolphins use sponges to protect their noses from stinging sea creatures that they hunt. Ravens use a sharpened stick to spear juicy larvae from logs.
With a tinge of arrogance, we (as humans) are amused and possibly taken aback by other species doing what we have mastered – using tools. Of course, our distant ancestors started off in the same way as all the animals I mentioned, progressing to the point we are at today.
For the vast majority of people on earth, not a day goes by without some form of tool use. In the workplace; I am hard pressed to think of an example that does not use tools. ‘Naturist walk guide’ perhaps – but even then there are cases to be made for tool use. The venue itself might be deemed a form of tool. Without getting too pedantic, I’m sure my point is made – we heavily use tools.
You can probably see where I’m going with this. The point of using tools is to make a difficult task easier – or even possible at all. We have become so good at making and using tools, the idea of a tool being made that can think for itself is actually at least hundreds of years old.
We can take examples of this from mythical history. Homer, around 1000 years ago, wrote about a giant bronze robot, Talos, who protected the island of Crete by running around it three times per day and if he saw any intruders he would pick up huge boulders to throw at them.
This was (and still is) pure science fiction. Today we can imagine how Talos might be powered and animated but back then it was moved by ‘divine essence’ or some other non deterministic concept. Nevertheless, Talos might be viewed as an example of an early idea of what might be called ‘artificial intelligence’. Talos was a tool used by the people of Crete to protect the island, and he was so good at it they let him do it autonomously.
Coming up through the ages, other man-made intelligences of myth were: the talking heads. Philosopher Albertus Magnus and monk-scientist Roger Bacon were said to have created ‘sacrilegious’ heads that talked, in the 13th century. These were rumoured to be able to answer all your life’s questions.
Vaucanson’s duck. From where we get the saying ‘if its looks like a duck, walks like a duck, quacks like a duck…’. Jaques de Vaucanson (1709 -1782) built gold plated mechanical ducks that had many of the attributes of actual ducks including eating, digesting and excreting food. This wasn’t a great intelligence but it was an example of the emulation of a complex system.
So there are many reasons and motivations for creating artificially intelligent machines. But whether the purpose of such a machine is for practical labour, intellectual pursuits, entertainment or mere curiosity, the overarching purpose is as a tool to achieve an end – whether that end is to occupy a child’s attention or bore a tunnel through a mountain range.
What is intelligence? This question clearly needs to be addressed if we are going to be making an artificial version of it. Government agencies have probably abused the term somewhat, for example: ”We have intelligence on country X’s nuclear capability.” – the word is used here in place of ’knowledge’ or ‘information’ about… . In the definition we are concerned with, static data and information does not constitute intelligence. So what does?
This subject could, and does, fill volumes of bound books. Here I am going to give a brief outline that will give some direction to our work.
Artificial intelligence is generally divided into two broad subcategories: another philosopher, John Searle, proposed these terms in 1980: strong AI and weak AI.
Strong AI has not yet been achieved. This is the type of machine that we see in science fiction – the character ‘Data’ from Star Trek NG is an example, another is the sinister computer ‘HAL’ from 2001 A Space Odyssey. Strong AI has all the capability and functionality of a human mind – and possibly more.
Weak AI is in the realm of reality. Limited decision making programs, chess players, statistical analysers, chat bots, etc. These are good at the task that they are designed for, but a chess playing AI would not be able to discuss your best career path for instance.
The defining line between strong and weak AI becomes quite blurred and gradual as the weak systems gain more and more capability. Strong AI may well manifest as an emergent property of multiple, interacting weak systems, but here again, we are stepping into the realm of the unknown.
As a working definition of intelligence for our purposes I will put it this way: An intelligent system is one that is capable of solving a given problem with a high level of, or complete, autonomy.
Of course this definition is not absolute. ‘Solving problems’ is not the only function of intelligence. There is also dreaming, imagining, doing things for no particular reason, etc. From here though, the philosophical discussion is turtles all the way down.