Human machineComputers can perform lots of tasks in the blink of an eye. They are great at solving problems, doing repetitive things and sorting items. However, we generally don’t think they are capable of coming up with creative thoughts or artistic ideas. At the end of the day, computers are just an intricate network of silicon, wires and transistors, so, how could they possibly be creative? I’d like to show you a provocative angle here, which considers computers as creative agents that will contribute to the future of human kind.

We say that an item is creative if it is new and has some sort of value. Newness assures us that an object/idea is original and never heard of before. However, newness isn’t enough to define creativity, since it tells us nothing about the value of a certain item. If creativity corresponded to newness only, a random generator would be the best creative agent ever. Certainly this is not the case in real life, since we hardly regard random ideas as creative, because they lack something important. That something is value. A creative item should always show a certain degree of value.

Now, the question of whether or not computers can be creative can be formulated as whether or not they can come up with something new and valuable. Newness is the easy part of the process. It’s very simple for a computer to come up with something new, since it can exploit a (pseudo)random generator and an impressive computational power. On the other hand, value is a very elusive feature. First, value is a subjective concept: what’s valuable for me, isn’t necessarily valuable for you. Second, value implies a kind of structure and a set of constraints. These constraints are needed in order to select a preferred subset of the potential infinite outputs.

painting fool

Painting created by the computer program “The Painting Fool”

Humans do this process intuitively. Computers can’t do that, so they have to rely on programmers. Therefore, the central issue is how you evaluate a creative item. Today there are effective strategies to assess the value of an item created by computers. For example, we can compare the new item to some models previously fed into the machine. As a consequence, the computer develops some parameters to assess its own work based on a sample. This process is incredibly effective when we consider music. In this case, the database corresponds to a set of compositions passed to the machine. The computer extracts relevant musical features and uses them as a stylistic guide to evaluate its work. This is the case of the astonishing computer composer EMI created by David Cope, which composes pretty convincing music in the style of a specific composer. Listen to this Bachish piece to get what I mean!

There are also other strategies to assess the value of a music piece composed by a computer such as evolutionary fitness functions or grammars, but I’m not interested in technical stuff right now.

What I’d really like to stress is that we are at the dawn of a new era. We’re teaching computers to perform creative tasks such as composing, painting and writing. Although the results are still modest, however, we’re slowly elevating computers from the level of slaves to the level of creators. Rather than considering the unlikely apocalyptic scenarios of intelligent machines dominating the world, I’d like to focus on the benefits brought by creative machines. Creative computers will provide amazing artistic/scientific/technological items, will facilitate our life and will empower human creativity. Indeed, I think a creative alliance between humans and machines is not so far anymore.

However, lots of philosophical questions remain still unsolved.

  • Can machines really be creative?
  • Can machines really appreciate creative items?
  • Can machines really judge creativity?

Maybe we’ll never answer these questions, or maybe it’s just useless to keep asking such impossible questions.

What’s your opinion?