Is It Deja Vu All Over Again? (Part VII)
What’s It All About, Alfie? (Part II)
Read this with an open mind, and do not be too hasty to draw conclusions!
- This post is written, partly in good humor, but also
- to make a point (or two), as well
SUMMARY & Introduction
Now it might seem outlandish to use the words ‘artificial intelligence’ in the same phrase with witchcraft, magic, and alchemy, all of which seem to relate to a much different historical time than the present. In fact, however, all of these words have appeared together in other writings, and not without some rational basis. In fact, there is such an interesting and perhaps even real connection between at least some of these words, that we propose to add two more, with some justification. The first word we would add is the word ‘Golem’, which describes an entity that does exactly and literally what it is instructed to do, without any possible thought or concern for what the consequences may be. The second word is ‘sorcery’ (as in the Goethe poem, The Sorcerer’s Apprentice). The message from all of this is Beware, There May Be Unexpected Consequences!
Now that we have laid the groundwork for discussion, let us proceed to some explanation of why these words are found here, and we’ll link it all to a partial answer to another question posed earlier, namely, is artificial intelligence a science, and if so, what kind of science is it?
Let’s Begin the Discussion with Witchcraft and Magic
To begin, and to quell the fears of some, let us point out that we are not going to try to relate artificial intelligence to witchcraft in any form. Even we do not think that there is any really legitimate connection (although we can’t rule out the possibility that there are those who might think that there is).
And so we move on to the next word, magic! Some of you may have seen or heard the recent Google demonstration of a computer dialing a restaurant for a dinner reservation, and fooling the concierge on the other end of the phone line into believing that he/she was speaking to a human being. In fact, it sounded like a confident and hip young man with all the recognizable mannerisms in his voice. It even succeeded in making the reservation!
Indeed, many who heard or watched that demonstration were disturbed by the idea that apparently a computer could do a pretty good job of imitating a human conversation, by actually interacting with an unsuspecting human. However, there followed an immediate outcry from many who feared the possible implications that might result from this kind of (magical?) ability, especially if used for possibly devious purposes.
Are we dealing with magic? Not with the old Merlin kind of magic, but certainly with the illusion that something has transpired which, in fact, is not what we think it is. And that is what we like to call magic today. Many who witnessed the demonstration were quick to see an unexpected consequence. If “magic” of this kind is real, and becomes part of our everyday experience, and not something out of the ordinary, we may face the prospect of confusing the illusion of reality with reality, itself, and that dear reader may be either good or really bad (think about it).
According to Google, which is now making the demo app available for making appointments, the app used “canned” phrases linked to words spoken by the person on the phone and matching stored keywords which might be expected when making a reservation or appointment. Speech synthesis (which was actually not needed in this case) has been well researched and developed since about the 1970s, so there really isn’t much new here, just an application of faster and more powerful computers which can react in real time. The app is similar in behavior to the familiar “phone chat” apps that offer help when you call customer service. If they can’t find a stored phrase to match keywords in what you appear to be seeking, they generally default to a response similar to “Let me connect you to someone who might be able to help” and then the line goes live to a real person.
A Golem is Probably Familiar, and The Principle of Inevitability
Now, to continue from a different point of view, history is full of references to an entity called by various versions of the word “Golem” (depending on the language involved). What is a Golem? There are many interpretations, but one of them is that a Golem (originally a being created out of clay) is not necessarily intelligent, but if given specific instructions, the Golem will carry them out exactly and literally as prescribed.
Does this sound familiar? It is, in fact, a very good description of the modern digital computer (or even the modern robot). In fact, the instructions presented to the machine must be supplied by a (syntactically) valid program. If the program fails to enable the desired result, then the corresponding computation or action also may fail to reach the desired goal, and in that case, there may be unexpected consequences.
Thus, what may appear to be magic performed by a computer, is in fact in every case, simply a machine doing what it has been programmed to do, regardless of whether the program is capable of reaching the desired objective. As long as the program is syntactically correct, whatever follows from it will result, and it may not be what was intended. The same may be said for the robot. We might refer to this as the Principle of Inevitability (or in more familiar terminology, “garbage in, garbage out”).
The computer has no free will and is incapable of doing anything not perfectly described by a program (which may be exceedingly complex). The program itself is always ultimately crafted by one or many human beings in order to accomplish some purpose that is within the realm of possibility for a computer or robot to handle. Even if the program has been “trained” to accomplish some goal, it still remains a program (a set of instructions) being executed by a machine, and nothing more.
To believe that the computer actually possesses human-like intelligence, is simply an illusion which we might call today artificial intelligence, but which also conjures up the familiar term in our common parlance, “magic”.
This brings up another important point. Many people today do not have a sufficient basis of knowledge to provide a context for understanding what a computer might be capable of doing or how it is being done. To those people, magic is about as close as they are going to come to a viable explanation.
We might mention that there is a need today for better transparency, or reporting, on the part of the AI community to provide insight into what they are doing and what tools are being used, etc, otherwise they run the risk that everything they do is suspect, and we are going to have much more to say about this aspect of AI in future posts and Dialogues.
Alchemy is Empirical
The next word is “alchemy”, and it referred originally to the efforts, in the middle ages, to find a way to transform lead into gold. The motivations for this are obvious, and unfortunately no one ever succeeded in doing it (as far as we know). However, it is the process of alchemy that is of interest to us.
The process of alchemy was not based on a sound foundation in science, but rather on the simple idea that lead and gold appear to be different substances, although of similar weight, and that there really ought to be some way of turning lead into gold. The process was one of trial and error (and perhaps even some magic “incantations”?). In particular, it would not have been unheard of for some alchemist hidden away in a dark underground chamber, or laboratory, deciding that perhaps a little sulfur combined with lead in just the right way would result in a heavy yellow metal which just might turn out to be gold! Probably nothing in the scientific knowledge base of that time would have ruled out this possibility. But why are we talking about this, when artificial intelligence is our topic?
We now refer the reader to our first Dialogue concerning neural nets and other matters, in case more background is needed for what follows. What is briefly pointed out in this first Dialogue is that neural networks do not have a perfectly sound underpinning in science, and that working with them is very much a trial-and-error process. Try this, and if that doesn’t work, try something else, and if someone else has already figured out what seems to work, copy what they did.
Now if artificial intelligence, and in particular, the use of neural networks is a science, then we have asked the question, what kind of science is it? The answer to that query is often provided by the practitioners themselves, who will likely tell you that it is very much an empirical science (not completely guided by sound fundamental principles, but rather by trial-and-error). We shall have much more to say about this in future posts and in the Dialogues, but for now, we leave matters where they are, and move on. We continue to leave as an open question, for the moment, the question of whether AI is actually a science.
Sorcery May Have Unexpected Consequences!
Our last word, which we added to the list, is sorcery and it is one well worth considering, and relates to much of the current news and speculation about artificial intelligence. In particular, we mentioned that the example of the word sorcery that we will refer to is found in the Goethe poem by the name “The Sorcerer’s Apprentice”, or in German, “Der Zauberlehrling”.
In case you aren’t familiar with the poem, the sorcerer possesses power to command inanimate objects to do his bidding. He is leaving his workshop and asks his apprentice to tidy up the place while he is gone. The clever apprentice has an idea to make his job easier. Using an incantation he has learned from the sorcerer, he commands a broom to go and bring pails of water to clean the floors. The broom turns into a stick figure with arms, legs, and a head and rushes off with a pail to bring water and pour it onto the floor. Having done this, it continues to go and fetch more water, until a flood of water begins to fill the workshop. At this point the apprentice wants to stop the broom (a Golem?), but unfortunately he does not know an incantation to do this. He finally takes an axe to the broom and slices it in two. Unexpectedly, the two halves each transform into brooms capable of carrying water and the situation gets even worse as they both bring pails of water as quickly as they can. Finally, in an act of desperation, the apprentice calls out for the sorcerer to return. The sorcerer returns and orders the brooms to a corner, instructing them to take orders only from their master (who does understand what he is doing).
It is not hard to read into this poem an analogy with what some people might think of or envision as artificially intelligent robots , and even perhaps “evil” robots that might just go out of control and inflict serious consequences on humanity. Most of such thinking is surely far-fetched,. However, with careful programming of a robot by a malicious “master”, similar scenes with robots might, in some sense, actually occur. And if they do, will their “master” return to stop them? Today we hear many warnings of unexpected consequences, and they are being taken seriously by many.
Perhaps one of the greatest dangers today would relate to vehicles possibly being controlled by poorly understood autonomous processes which may not be perfectly reliable or predictable (and we will have much more to say about this). Perhaps we would do well to keep Goethe’s thoughts in mind, and perhaps he was just being perceptive of things that might come when he wrote The Sorcerer’s Apprentice. The poem, in turn, was based on an even older fable.
The lesson to be learned from all of this is perhaps that in dealing with things that are poorly understood, or may not be perfectly reliable or predictable, there can be unexpected consequences. Norbert Wiener himself, the father of Cybernetics, issued similar warnings in the early years of the computer revolution (see our blog post here).
An Historical Note
As an interesting historical note, Goethe was also known to be interested in alchemy! It may also be of interest to note that the word “robot” first appeared in 1921, and may have been based on the notion of the Golem!
More to Come
We hope you have enjoyed this post (written slightly in jest), but have also found here food for thought!
Please stay with us as we continue to delve into the mysteries of “The Monster That…” in our future posts and dialogues. Until then, we’ll wish you well and hope to see you again soon!