What is art? Is it as simple as having a new idea? Expressing emotion? Creating emotion? Simply something visually pleasing; Is it an accident that when looked at or listened to makes others think about a specific subject or idea? Is it as basic as mistakes made during imitation? Can anything be art? Now we just have lots of questions, - but there's no harm in that.
"Can computers be as creative as humans?"
Asking the Right Questions
Well, "Can computers be as creative as humans?" seems to be an equally difficult question to answer because we don't know what art is. But creativity does seem to be a key component of art, and creativity is just thinking differently, right? Maybe creativity is the only component that matters. Art is subjective anyway. How does one learn to be creative then? A better question is.
"Can we teach a computer to be creative?"
Yes, I think that this is the question we are looking for. Can we teach a computer to be creative? Let's narrow our definition of a computer so we have a handle on what we are talking about.
A Little Bit Of Background
(Alan Turing The Father of Modern Computing. Go watch the movie imitation game if you want to know more.)
In 1950 Alan Turing published “Computing Machinery and Intelligence” in which he proposes “The Imitation Game” which is now known as the “Turing Test.”
What is the Turing Test you may ask? The test is based on the idea that a computer’s success at “thinking” can be measured by its probability of being misidentified as human. So, we are talking about Artificial Intelligence, a term coined in 1955. This broad field of study is often misunderstood due to the term AI getting thrown around and used as a "buzz word" a lot. But if we stick with a broad definition we all utilize AI almost daily.
The Creativity of AI
"Is AI thinking, and more importantly is it creative?"
To answer that let's look close to home. AI isn't in some distant lab hidden away. We all use and reap the benefits of AI all the time. From Google auto search recommendations to Facebook's facial recognition, from Pinterest image categorization to our Netflix cues. We are surrounded by AI. Not only is the rise of AI providing services, but it is also learning from the data it collects. Machine learning at its most basic is the practice of using algorithms to parse data, learn from it, and then make a decision or prediction about something in the world. AI takes data points based on our usage to improve itself. That data is not exclusively useful to the AI. AI translates the data it collects into digestible information in the form of graphs, charts, and spreadsheets for use in research studies and advertising.
Graphs and charts are a very basic yet interesting example of visuals created by AI. None of the AI mentioned in the uses above could pass the Turing Test though. They are examples of Narrow AI designed to perform a single task and do it as well as, or better than, we humans can.
Photo & Video: Image of the generative concept art process.
We are looking for AI that has the potential to think like a human mind, This type of AI is called a "Generative Adversarial Network." These systems can do a lot of things and have the most potential at dreaming. A key component of creativity is dreaming, thinking differently. We, humans, are all just dreaming our reality. Furthermore, We humans interpret our reality with our minds. That's why, when we look at a cloud, we can see things other than a cloud. It may look like a cow... or a horse. Or you may see a face in the tiles of a bathroom floor. We can discern the truth most of the time, but our minds play this game all day long. "What is that?" our mind says. "Is it a dog?" Our minds ask the question and we go about answering it, forming the image in our minds. Generative Adversarial Networks do the same thing.
Photo: An AI decision making process.
In humans, this ability to see patterns and imagine things is called Pareidolia. It's our ability to perceive recognizable images or meaningful patterns where none may exist. Generative Adversarial Networks do something similar using an approach to AI called "Deep Learning". Deep Learning is used to search for meaning in something, not just memorize what it is.
A great example of this is Google's Deep Dream Project from 2016. EscapeLab a research laboratory based at the Department of Architecture, University of Thessaly describes Deep Dream in a way that helps us wrap our brains around a complex concept. "Deep dream is designed to detect faces and other patterns in images, to automatically classify those images. However, once trained, the network can also be run in reverse, being asked to adjust an image slightly so that a given output neuron yields a higher confidence score. This can be used for visualizations to understand the emergent structure of the neural network better." The examples below were generated by the Deep Dream Project. You can create your own using the same algorithms below:
Michael Copeland, in his article: What’s the Difference Between Artificial Intelligence, Machine Learning, and Deep Learning? shared his thoughts on AI. "Today, image recognition by machines trained via deep learning in some scenarios is better than humans, and that ranges from cat photos to identifying indicators for cancer in blood and tumors in MRI scans. Google’s AlphaGo learned the game Go, and trained for its match — it tuned its neural network — by playing against itself over and over and over."
The best example of these come from Alphabet-owned company "Deepmind. "Deepmind's Neural Turing machine possesses neural networks that can access external memory, resulting in a computer that mimics the short-term memory of the human brain."
This all sounds like some terrifying sci-fi film, to be honest. It all seemed harmless enough when we were discussing AI in the context of our smartphones but now it's beginning to feel eerily Alien. Let's pause to regain our composure and play with a friendly example of a neural network in the form of the game Quick Draw below.
The Art of Self-Teaching
Many of the most powerful AI examples are focused on Image and speech recognition because we are creating systems built to understand us. What is amazing about these systems is that they are able to take that information and present something new based on what they are given. This translation of source material is key to determining if AI is creative. So, can, we answer our question? Can we teach a computer to be creative? I think the examples provided above make it clear computers don't need to be taught to think differently. AI inherently looks at the world differently and there is no denying that it is able to create. In a way, we are teaching AI but the key to these neural networks is that it is capable of teaching themselves.
These systems will become more advanced, more intelligent. Let's look back to our first seemingly unanswerable question "What is art?" Many would argue that intent is the key to creating art. AI systems currently don't have their own motivations. Sure, you can teach a computer to play a game and run up a high score or paint a picture based on some other pictures. But these systems have no motivations of their own. Art remains a human endeavor but AI serves as a tool that allows anyone to mimic a master painter, to manipulate video in a generative way to create photo-real special effects. AI, as it stands now, is the new brush that artists will use to create the art of the future.
(Click on above image to open video) Video (Ai was used to scan Jim carrey's face based on his many cinematic rolls. That AI was them able to project Jim Carry's face frame by frame onto Jack Nickolson's face in what has become known as a DeepFake.)
Link to video: https://youtu.be/-ZRUZzZPGto
Deep Learning: https://en.wikipedia.org/wiki/Deep_learning
Student Approaches to Learning: https://en.wikipedia.org/wiki/Student_approaches_to_learning
Journey on the Deep Dream Video: https://www.youtube.com/watch?v=SCE-QeDfXtA
Google Research Blog on Inceptionism: https://ai.googleblog.com/2015/06/inceptionism-going-deeper-into-neural.html
Slate Article on the Future of AI: https://slate.com/technology/2015/03/elon-musk-stephen-hawking-artificial-intelligence-the-state-of-a-i-research.html