The “Deep dream” of artificial intelligence: can machines be creative?

This post is also available in Dutch.

Over the years, the level of progress in machine technology has reached a point where we must begin to ask ourselves: is creativity an exclusive trait of the human mind?


Picture by Lara edited using

Humans vs Machines

Nowadays, machines outperform humans on a variety of tasks, which are learned through the frequent exposure of numerous examples. Robots have out-maneuvered human-opponents in the go-game, have recognized some diseases better than their human-counterparts, and have even been hired by some law firms. There is no way humans can outbest computers when it comes to speed and accuracy: within minutes, computers can go through terabytes of data. Computers are known for managing search tasks well, but in light of recent improvements, computers can now also carry out more complex tasks. Deep learning, an artificial intelligence (AI) technique, makes it possible for machines to address a set of similar tasks simultaneously by using “ensemble problem solving”. This is the first AI technique that is capable of solving problems of high complexity similar to what we humans must face daily. On the other hand, if a machine is able to manage complex tasks like we do, then what would be left for us humans to do?

Humans have the ability to solve problems that have never been faced before by producing knowledge from past and unrelated events. The spark; insight; inspiration—these are the words people use when they make a breakthrough. These breakthroughs can be as groundbreaking as solving the Theory of Relativity to the daily tasks of a project manger in developing business strategies and finding gaps in the product market. The creativity needed to deal with such tasks is a “human” undertaking, as opposed to searching through endless inventories of data with the purpose of finding a single item.

Creative machines

Today machines cannot produce “creative” output in the office as of yet, but they are breaking ground in some of the arts. Let us go through some of the recent apps to see if their output is really artistic. There are apps like Dreamify or Prisma that enable the transformation of photos into work that looks as if it was produced by a particular great artist themself. These apps do not apply a filter to the image but create a reinterpretation of the image using an AI technique called deep neural networks. This algorithm has been used to create new jazz compositions—but don’t be too quick to judge, AI jazz is still in its infantile stages! On a more positive note, if you begin to like AI jazz, then the band cannot ever disappoint you by breaking up. Neural networks have also written movie scenes, produced literary works in Shakesperianand have  even written pseudo-phrases or tweets resembling those of Donald Trump.

The point is that the so-called creative content developed by deep neural networks is always produced in the style of a certain artist, so there is nothing really authentic regarding the content. However all great artists like Dali, Louis Armstrong, and even Shakespeare began exercising their craft by imitating other artists and expanding on their work. Maybe imitation is a necessary precursor for creativity. In any case, keep challenging yourself to produce authentic work, or otherwise you will just become a tool in the hands of powerful machines.

This blog was written by Lara. Edited by Marpessa.

+ posts

Leave a Reply

Your email address will not be published. Required fields are marked *