A number of different terms are used when talking about the domain of Artificial Intelligence, including ‘machine learning’ and ‘deep learning.’
Continuing our exploration of whether machines can create art, this week we’ll get a deeper understanding of what Artificial Intelligence is, and how it differs from traditional computing.
There are many misconceptions of what AI is. The biggest one is that AI needs to function like the human mind in order to achieve the same feats as humans. So people talk about computers not having imagination or creativity, emotions, or experiences, and how, without these properties, AI can never match human performance. The truth, as we will see, is that machines have an entirely different way of functioning, but they may still be able to match the human brain, even in terms of creativity.
Before we begin, one quick clarification: A number of different terms are used when talking about the domain of Artificial Intelligence, including ‘machine learning’ and ‘deep learning.’ For the sake of convenience, we’ll treat them as having more or less the same meaning: AI.
Another tricky aspect is that what we mean by AI is constantly changing. For example, at one point, Optical Character Recognition (OCR) was considered to be the domain of AI. But as the technology became standard, people started to think of it as mere computing, and not AI. So it might be difficult to define AI in terms of what it does. Instead, we need to understand AI in terms of how it functions.
At the risk of gross oversimplification, it is easier to understand what AI is not by comparing to how the human mind works, but by comparing it to how traditional computing works. In traditional computing, you write the rules, then you put in the data, and you request a fixed result. For example, you teach a computer about numbers and maths, then give it a problem to solve (‘What is 2+2?’) and the computer gives you a fixed result based on the rules you have defined earlier. The value of such computing, of course, comes from the processing power that computers have. We humans can also solve 2+2 quickly enough, but if you had a million such calculations to do, a computer can process it all for you in seconds.
AI differs from traditional computing in important ways: the data it works with is not fixed. AI will almost always begin with a fixed set of data, but along the way, it may choose to discard some data, or it may add the results of its own actions as additional data—learning from the results of its own actions, as it were.
Similarly, unlike in traditional computing, the process or reasoning that an AI machine follows—the rules that it works with—are not fixed. An AI machine can have an open-ended mandate to achieve a certain result, and the reasoning, the path it takes to get to that result is flexible unpredictable, and determined by the machine itself.
In the same way, the results that AI delivers can also be open-ended. A complex AI machine can work with an open mandate such as ‘run city traffic lights with maximum efficiency and prioritize ambulances and school buses’ rather than fixed results such as saving a document or calculating a formula.
This is a very different way of functioning than the human mind, but the results that AI computers can achieve already exceed human performance in many areas. AI already powers things like search engines, security surveillance, health prognostics, predictive customer service, dynamic and intelligent product pricing technologies, and much more. And AI machines do this far more efficiently than humans can.
With these abilities, how would a machine go about creating art? A machine will not wait for inspiration, of course. And a machine does not grow up listening to stories, or learning to sing and dance during childhood. So the first step would be to show a machine what art is.
From there, machines can ‘reverse engineer’ the properties of art. Reverse engineering is an important concept here that, in simple terms, refers to extracting knowledge or design patterns from a dataset (such a collection of paintings) and identifying what makes up a painting: the colors, the canvas, the textures, lines, color densities and hues, and so on.
Once a machine ‘knows’ what art is, it needs to get creative in order to produce new art. In the next article, we’ll explore how an AI-powered machine can create original art in mediums such a painting and music.
*This article was originally published on April 17, 2017 in Telangana Today, and can be accessed at Decoding machine art
Like most of you working in tech and startups, I absolutely loved reading Peter Thiel’s Zero to One. It’s a stunningly thought-provoking book, with a deep new insight pretty much on every page. It’s hard not to be impressed by the clarity of thought as well as the...
I’m sure we have all heard of that fraud where tourists walk into a restaurant, order some basic food item-dosa, idli kind of thing-and are presented with a bill for thousands at the end of it. And the swindlers at the restaurant claim that the tourists should have...
What are some things you are aspiring to do this year? Is traveling anywhere on your list? Almost everybody thinks they need to travel. Not convinced? Log in to Facebook, and see what your feed is showing. What friends of yours are vacationing currently? What pictures...