“When discussing how computer science and interest in artificial intelligence contributed to the cognitive approach of creativity, we talked about the Turning test. What is the Turning test and how is the concept of the Turing test relevant to the study of creativity? (Hint: Eliza, GPS, and AARON)”

The students had a problem with this question. Even after multiple explanations and further hints, they still had a problem. Finally, I relented and gave them the answer to the question (and did not include this question on the final exam).

What follows is my answer with some modifications to serve the purpose of this post and make it more accessible to a broader audience.

“When Computer Science first became a discipline, computer scientists were examining problem solving and, in the process, started considering what it means to be intelligent.

Why?

One of the reasons was because computer computation involves a physical symbol system just as human computation/thinking involves a physical symbol system. If the output of a human symbol system can be said to be “intelligent” then it logically followed that the output of a computer symbol system also has the potential to be intelligent.

This logical inference was further supported when the computer scientists successfully created computer programs that could play chess and chess is something that intelligent people play. In fact, the subdiscipline of artificial intelligence is all about making computers or machines intelligent or at least behave intelligently.

But what is “intelligence”?

Alan Turing suggested what is now known as the Turing Test as an operational definition of intelligence. In its essence, the proposal is that if a human interacts with a computer and thinks they are interacting with a human, then the computer could be said to be behaving intelligently. The means by which the computer does so is irrelevant by this definition. After all, we aren’t entirely sure how HUMANS manage to behave intelligently.

I believe that the question of whether or not computers can behave intelligently is, at heart, the same question as whether or not computers can behave creatively.

Let us say that a computer creates something novel such as an original painting (which AARON can do) or proves a mathematical theorem (which GPS could do).  A human (who doesn’t know that a computer created this new thing) views the product and is asked if the product is creative. If they say that it is creative, even if they say it is only a little creative, can’t the computer be said to have passed the creativity version of the Turing test? Does it matter HOW the computer creates something new? After all, we don’t know how we do it.”

As I said earlier, the students had problems with this question even though we had discussed this issue in depth at several points during the semester.

I believe that one of the reasons they had problems with this question and my answer to the question is that ALL the students in the class vigorously denied the potential for creativity in the computer program.

When it was pointed out to them that they thought the output of the program (in their case, a painting created by AARON) was creative when they thought it was a human that created the painting, their response was that the computer wasn’t creative. Rather, it was the person who had written the program that resulted in the painting who was creative.

I pointed out that the programmer in question did not actually tell the computer what to create. He only supplied the computer with background knowledge about form, design, composition, etc. and the computer used that information to create a novel product. I then argued that this is exactly how humans develop novel products.

In response, the students fell back on the argument that when humans create, they do so intentionally and consciously. The computer did not know what it was doing, THEREFORE, it was not being creative.

I think it is a beautiful thing that the students in the classroom fall back on the equivalent of Searle’s Chinese Room argument to refute the hypothesis that computers can be creative. That they do this with no prior knowledge of the Chinese Room argument or a full understanding of the premises underlying the Turing Test is also pretty special.  According to Searle’s Chinese Room argument, it suggests they have not provided an intelligent response to the claim that computers can create.

In Dartnall’s introduction, he made a comment concerning “Emmy,” a program written by David Cope. Emmy takes as input several musical compositions from a single composer and, using a series of simple rules and transformations, creates novel musical pieces in the style of that composer. Reports from those who have heard these novel creations say the pieces created by Emmy are difficult to distinguish from those of the composer on which the pieces were based. In addition, they have been described as “soulful” and “delicately finessed and preternaturally beautiful.”

Dartnall questions whether Emmy, or any other computer program for that matter, can be considered to be “creative”.

When I hear things like this I always wonder why this is an issue.

If a person can not distinguish the creative output of a computer program from the creative output of a person, and if the output would be deemed “creative” if created by a person, then why is the same output created by a computer deemed “not creative”?

Why does the source of the idea matter?

I shouldn’t actually phrase this question as if it was an novel question because it isn’t. The first time we started asking ourselves questions of this sort was when we started exploring whether it was possible for a computer to be “intelligent”.

Turing proposed what is now called the “Turing Test” of intelligence. In short, Turing’s test of intelligence states that if a human in conversation with a computer actually thought they were conversing with another human, the computer could be said to pass the Turing test for intelligence. “If it acts intelligently, then it is intelligent.”

I think the same could be argued for creativity. If a computer creates music (or a poem or a painting…) that is judged to be creative by a person unaware of the source of the creation, then it IS creative.

The most common counter to the argument that I just made for the potential for creativity in computers is that a computer is not creative INTENTIONALLY. This idea of intentionality is also used as an counterargument to the proposal that computers passing a Turing Test exhibit intelligence.

This argument points out that humans are intentionally creative. They create new pieces with full knowledge of what it is they are trying to do. In contrast, a computer program is taking input and, using the rules supplied by it’s creator, is producing novel output. It knows not what it does…therefore, it is not creative.

This suggests that creative output without the intention of being creative is NOT creative.

At best, the programmer is the one being creative because the programmer wrote the program that created the novel products.

I think that the fact that David Cope was able to write a program that was successful at generating creative output could be used as an argument in FAVOR of the creativity of the output.

Specifically, in creating EMMY, David cope applied his understanding of the cognitive processes involved in creating music within a particular style.

He specified the input that was needed into the system, the representation of the input, the analysis of that input and subsequent use of the analysis and input representation to create new compositions.

The fact that EMMY successfully creates novel compositions in the style of the input it receives suggests that Cope understands the processes underlying composition to the extent that he is able to model the creative process.

Researchers understand the processes involved in bipedal walking enough to create bipedal robots that can walk. Do we deny that these robots are walking just because someone programmed them to walk? Do we deny that these robots are walking because they didn’t INTEND to walk? No. We would agree that the robot is actually walking. Why can’t the same be said for a computer that has been programmed to exhibit creativity?
Perhaps the reason people are willing to say robots are capable of being programmed to walk but are unwilling to say that robots are capable of being programmed to create is because our ability to create is considered to be one of the characteristics that distinguishes man from beast.

But that is the topic of a different discussion.

More information on creative music programs:

A Wired article exploring the research of Gerhard Widmer who is expanding on the work of David Cope

An article examining the use of intelligent tools in music