I am writing this article because it seems to me that most people assume it is impossible for computers to be intelligent and the general consensus seems to be that if it was at all possible, it would have to be off in the distant future. I am not writing anything in-depth here, this is just some concepts I thought I’d throw out for public consideration.
Maybe there is more to machines than we are willing to accept just yet.
Intelligence in the eyes of most people seems to separate humanity from animals and machinery; but what is it exactly? According to dictionary.com, intelligence is defined as “capacity for learning, reasoning, understanding, and similar forms of mental activity; aptitude in grasping truths, relationships, facts, meanings, etc.”
This definition covers a lot of things, how can we break this down and apply it to artificial life?
Learning
Starting at the beginning of the definition we must first look at ways in which the human mind learns in order to understand the relevance of machine learning techniques.
Most people learn through a variety of methods, a one-on-one teacher, hands-on activities, visual memory (ex: reading or viewing illustrations.) Some people need to listen to a concept and have it broken down a few different ways in order for it to make sense, often this means multiple teachers.
Now, looking at these things from a mechanical perspective, all of the above mentioned learning techniques used by regular people can be summed up simply enough as “input.”
Most people and animals have 5 senses to work with for gathering this input. Without fancy gadgetry a computer typically only has one input method we can use to interact with it; the text that you can type into it with a keyboard. Obviously by default, humans have quite an advantage in terms of variety in learning but this advantage doesn’t automatically make people smarter or give them more potential.
The next step after receiving input is to analyze it, match it up with other forms of related input that has already been acquired and register a conclusion, or again in mechanical terms “output”.
Output has some interesting functionality to it, we store it in our memory until it is needed or updated, when it is needed, we call the output and pass the information to another person.
Computers have been able to do these things since the start of the digital age but let’s take a look into some of the other aspects of intelligence as it is defined.
Reasoning and understanding
How do we reason? Well, as people we always want what is best or “optimal” in every circumstance, our bodies tell us through touch when we stumble into painful and pleasurable stimuli. This is an automatic defense mechanism we were born with; it helps to keep us out of trouble. Most people generally don’t have to burn themselves too often before they realize that this is a bad situation and should be avoided wherever possible. At the same time, trying different flavours of ice cream will yield a favorite as there will be one that appeals to your tastes more so than others.
So to be summed up, our reasoning is based on a set of scales that weigh the good and the bad, in any given situation we will always aim for the best possible outcome, even when we do not have enough information to make an accurate measurement.
This can easily be reconstructed in computers as well, we’d simply have to input a range of positive and negative circumstances into the program and code it to find the optimal setting for any given situation. (I’m speaking generally here, doing this in a literal sense directly interacting with each element would take more time than our life spans would allow.)
Actually, this practice is used in simplicity already in almost every computer program available today, but it’s called “error catching”. Programmers will try to trap errors before they happen, telling the program that when this situation occurs something must be done to fix or compensate the problem.
Grasping truths, facts, relationships and meanings.
I’ve grouped these ones all together because as I see it, they can all be summed up the same way, and tie into the functionality I’ve already talked about. The only real difference here is the addition of a functionality to find the averages. If one person tells you that the movie “Untouchables” was awful and that you shouldn’t bother wasting your money on a ticket and yet, five other friends of yours tell you the movie was great and well worth the $10.00, you will factor in your own likes and dislikes according to what you know of the movie and then if you are impartial you will probably listen to the majority vote, (unless of course, you are strapped for cash, then you will wait until it comes out on DVD.)
In this scenario, the verbal input was slightly varied having a strong positive and a small negative factor. Weighing these factors you came to a conclusion that it’s probably not such a bad movie, then of course you weigh in the expense verses budget and if that gives a green light you may watch the movie, and then your own experiences will determine as to whether or not you enjoyed the movie. If you didn’t like it at all, you may discredit your five friends and apply more credit to the one guy for the next time.
Anything that can be explained logically and applied mathematically can be duplicated easily in a computer, after all, the computer’s base code is comprised on 1s and 0s a mathematical binary language that allows the computer to store and respond to data which is the primary stimuli for a machine verses touch sensory for people like us.
So what are the main differences?
Computers don’t have as many different kinds of input, or output as we do, though that can easily be changed over time, also, we’ve never put enough effort into making a computer just like us in these regards, in some ways we underestimate the complexity of the human mind and how it interacts with our bodies, and in other ways it’s over-rated. When you boil it all down, it’s a matter of perception and openness to the differences.
To ask the question, will computers ever think like us? It might be more important to first consider whether or not we would accept it even if they were.