We start with this story, taken from Wikipedia, called 'River Crossing Puzzle'. There you can find out more about the background and other 'propositions':
The wolf, goat and cabbage should cross the river, but the farmer can only take the wolf or goat or cabbage with him in a small boat. How can he prevent the wolf from eating the goat or the goat from eating the cabbage. The
only thing left to do is to take the goat first. But then it gets more difficult, right?
But if you are given a little help, it becomes easier. You just have to realize that the farmer can also transport back. So we won't give you a solution to the overall problem at this point, but trust that you can now easily figure out
whether he has to take something, what or who he has to take back with him, so that there is no unpleasant situation.
Our question is a completely different one, namely: does a computer program that solves such tasks already belong to artificial intelligence? Even scientists involved with this might say 'yes'. But we want to answer that with a
Alan Turing (1912 - 1954) created the test named after him to check the existence of artificial intelligence. To put it simply, a person talks in written form for a certain period of time, once with a person and then again with a
machine, each in a different room. What percentage of the answers does he correctly assign to humans and computers?
You guessed it, we don't think the two examples are comparable, the solution to the problem above is a snap for a computer. To pretend below that it is a human being we not only find it much more difficult, but also from the
sum of the problems to be solved belonging to another league.
Of course, we have a very specific goal in mind for artificial intelligence, which is to one day enable us to drive completely autonomously, i.e. level 5 as promised. Try to imagine for just a moment what can happen on a city
street and compel you to react. As an introduction to this complex topic, we give the example of fuzzy logic.
Modern electronically controlled automatic gearboxes no longer have a range selector for performance or economy mode. The control device determines, through the available sensors, whether the driver prefers to drive in
the performance range or in the economic efficiency area. Now, how does the control device decide? A passenger would probably be more able to judge, after travelling a certain distance, whether the emphasis is placed on
economy or on performance. To transfer the rules of human thinking to the computer, is the job of Fuzzy logic.
Therefore, the Fuzzy logic inserts, through an affiliation function, a variable (value between '0' and '1'). Then, using a certain affiliation function, the temperature 59.5°C, would be 'hot', using a different affiliation function, it could
be categorised as 'warm'. The black diagram (above on the right) shows, in addition, various dependencies. The affiliation is shown mostly in the form of a triangle (blue), a trapezoid (red) or a normal distribution (brown). In
between them, intersections occur where the values applied to the X-axis can be assigned to the one or the other area with different affiliation values. In this case, with the trapezoid, in the center of a certain X-area, the
affiliation value 1, is reached.
In our example with the range selector, certain accelerator pedal positions with varying affiliation values, are assigned to certain areas (a small amount of gas, a little more gas etc.). In the same way other variables are
included, e.g., the speed at which the accelerator pedal is operated. In the end, several variables with their affiliation values are calculated against each other, and the result is - although it concerns 'blurred' values, quite
distinct, either performance or economy.