Creative Intelligence
Albert Camus once said that “true art is characterized by an irresistible urge in the creative artist.” Henry Ward Beecher similarly wrote that “Every artist dips his brush in his own soul, and paints his own nature into his pictures.”
You don’t have to look far to find quotes like these, because art is something we consider intensely human. Art and the artist are so thoroughly intertwined that we can’t bear to think of one without the other.
For better or worse, we’re going to have to rethink this comfortable little notion. Machine intelligence is advancing to the point where algorithms have begun to invade the world of culture and the aesthetic. From recommendations to evaluation to the production of art itself, computers are becoming a force to be reckoned with in the creative realm.
The Search for Creative Assets
When you make a TV ad in Ukraine (as I have), you generally do it on a tight budget. You certainly don’t have the money to buy the rights to the latest hit by a big pop star or a vintage Beatles classic. There are some local musicians who can create something for you, but thats a pretty involved effort and, to be frank, the quality isn’t worth it.
I found a good solution with DeWolfe Music, which is an online database that gives you access to thousands of songs from unknown composers and performers. You can search by music genre, keyword (e.g. an artist that you’re trying to emulate) and tempo, quickly find what you need and license the music for a small fee.
Newer services, such as Pandora and Spotify, deploy a similar idea in order to build custom radio stations. Rather than a human programming director choosing your music, you can just give the software some clues about what you might want to listen to and it designs a selection from a nearly infinite database to cater to your mood and preference.
This is all done through the use of complex mathematical techniques, such as Bayesian classifiers and Gaussian copulas, that recognize similarities between data sets. So just like a sommelier might ask you what wine you typically like and offer you something similar, recommendation algorithms can do the same with music, films and even art.
Cultivating Creativity
Being able to search and find elements of art and culture is one thing, but can computers appreciate quality?
Mike McCready has shown that they can. His company, Music X-Ray, offers a service where composers can upload their music to evaluate its hit potential and it has been shown, in many cases, to outperform professional music executives (reportedly predicting the success of Norah Jones when many industry experts were skeptical).
As crazy as the idea sounds, you’ve probably recently listened to many songs identified by the service. Every major label now uses some version of it and they’re not alone. Movie studios employ a similar service, called Epagogix to tell them which scripts are likely to become box office hits.
And it doesn’t stop there. Music scholar and composer David Cope has built algorithms which create music that has drawn critical acclaim. In fact, even music experts can’t tell the difference. When Cope’s computer generated music was played along with a Bach piece and another original composition, they couldn’t correctly identify which was which.
What is Creativity?
As impressive as all of this is, it creates a particularly thorny, visceral problem: If creativity can be reduced to an algorithm, doesn’t it lose its soul?
While I admit, I find something troubling in all of this as well, after thinking it through I have come to believe that artificial intelligence actually has the potential to help us appreciate creativity even more, in much the same way as Richard Feynman explains how understanding the inner workings of a flower help him acknowledge its beauty.
Our brains, in many ways, are inferior to computers. They transmit information at the relatively feeble rate of 200 mph, vs the speed of light for computer chips. They get tired, need nourishment, age, forget things and don’t interface with other databases of information effectively. Objectively speaking, they are slow, inefficient and prone to error.
Their saving grace is that they are a massively parallel complex network. They are made up of 100 billion neurons and each one can connect to any other. These interfaces, called synapses, optimize themselves as they strengthen and decay with use and link to the outside world through our five senses of sight, touch, taste, hearing and smell.
While computers have relatively few active pathways at work at any given time, we have millions, giving rise to infinite permutations of thoughts, feelings, bodily functions and sensory inputs. Perhaps not surprisingly, these hierarchies get tangled, resulting in strange loops that manifest themselves in what we have come to know as original creativity.
As Douglas Hofstadter said, “In the end, we self-perceiving, self-inventing, locked-in mirages are little miracles of self-reference.”
Rethinking the Creative Process
The creative process has always been cloaked in mystery and artistic types tend to be resistant to formality. Nevertheless, professional individuals and organizations strive to develop an effective framework to enhance the productivity and quality of their work and creativity researchers have been able to identify some basic principles of creativity.
However, in light of the breakthroughs in machine creativity, I think that we need to revisit past thinking about creativity and identify three major processes:
Forming Intent: Every creative endeavor has its purpose. Designers work on a brief that someone else creates while true artists form their own purpose, but in either case, the final result is, in essence, a solution to a particular problem.
For example, my purpose in creating ads was to sell a product, while Picasso’s purpose in creating cubism was to establish a fundamentally different way at looking at the world. In the final analysis, both are judged by the significance and the degree to which the original intent was fulfilled.
Searching the Domain: As Howard Gardner explained in his highly regarded study, Creating Minds
, great creativity requires a thorough knowledge of the domain. Picasso’s cubism, for example, was inspired by his encounter with obscure African art. The larger your database of experience, the greater your ability to create.
Computers obviously far outperform humans in this regard. They have practically limitless memory and their vast computational power enables them to search incredibly quickly and accurately.
Tangling Hierarchies: As I’ve written before, great breakthroughs come from synthesizing across domains, whether that be Picasso’s blending of European and African art or Rock and Roll’s unique fusion of various american music styles. It is when two or more ideas collide in a meaningful way that people find inspiration and creative flow.
It is this last trick, that of emulating the strange loops in our mind, which computers have recently learned how to do, that has given rise to machine creativity. David Cope, for example, found that his computer generated music was dull and lifeless until he injected an element of randomness into his algorithms.
Flying By Wire
Pilots don’t really fly planes anymore as much as they direct them. These days, their controls and instruments don’t actually connect to the plane’s mechanism, but to computers which translate their intent into meaningful action. In doing so, they make flying far safer and more efficient.
This is known as flying by wire and we don’t see anything threatening or strange about it. While at first it may seem to be a bit more disconcerting when computers start navigating the realm of abstract thought rather than the mechanics of aviation, it shouldn’t be, any more than driving a car should affect our feelings about walking.
So what makes us creative? Our ability to form our own intent. It is only through creating a purpose that is uniquely our own that we can fully embody the human spirit.
– Greg
Great and stimulating article Greg. As a professional creative I know some of the methodology and strategies of the creative process can be replicated and passed on.
And variations of their elements could create novel content that essentially ‘works’.
We see it when work is copied or has “influence” upon others.
That is really the power of our collective mind as we take portions of what others do and reconfigure in our own way, however, for me the real driver for creative expression is our emotional response to the world and our existence – inclusive of our philosophical point of view on life and death and what it all means. Mozarts Requiem mass could not have been created by computers. A facsimile of might be replicated with the right parameters set but this is just humans playing with their toys. The intent, the visceral requirement to express is, I believe, part of the human soul. We can be fooled by a picture of reality – only in the sense that it is a clever picture. Real music is a real expression of the spirit. A reflection of the landscape of the emotions at the time. Computer generated creative output has no real landscape of emotion to reflect. Eventually we will be able to tell the difference as easily as we can tell the difference between CGI and reality.
Excellent post Greg, I particularly like the definition of forming intent.
For a creative artist, that purpose may arrive after the product and possibly enlighten future creative direction.
For a Designer such an approach will probably end in commercial mayhem.
The intersection of commerce and art is intriguing.
Sometimes it is only when viewing the collective output of an individual artist that we can hope to get some understanding of their purpose, the development of work over time helping to inform our judgement.
I’m also reminded of a Picasso quote ( my Google algorithm has failed to deliver me) stating – Aesthetics means as much to me as entomology does to an ant….
Regards
Rohan
Thanks Jimi, but I think you’ve missed my point. Computers ARE beginning to produce original works like Mozart’s Requiem and at such a level that critics can’t tell the difference. In fact, many praised his music’s humanity and feeling before they found out a computer composed it.
On the less artsy side, next time you read a box score or a financial report, don’t be so sure that it wasn’t written by an algorithm. Computers are beginning to take over the function of writing basic articles for major publications.
All this has already happened. It is fact, not speculation. What happens when computers are 1000 times more powerful? That’s only about 15 years away, but then you have to remember that the algorithms are improving very quickly as well, so a thousand fold increase in machine intelligence will happen much faster than that. Ray Kurzweil predicts that by 2030, computers will display intelligence indistinguishable from a human.
This is all pretty scary stuff and it will require us to rethink much of what we thought we knew.
– Greg
Oh crap!
Rohan,
I think that’s very true. For me, sometimes the best thing about writing a blog post is finding out how it ends:-)
– Greg
It is a bit disconcerting. We’re all going to have to learn to make peace with our machines over the next decade or so.
– Greg
That’s a scary thought – and for the time being I bow to your knowledge of and exposure to this technology, Greg. But I hadn’t actually missed the point – I was simply countering by offering that we would get used to the differences and be able to spot the machine music after a while – as those in the know can spot the difference in CGI offerings – no matter how incredible convincing they are – in the visual field. Why should audio be any different – even from a composition point of view. What your suggesting is the Monkey’s and typewriters theory is now true and I just don’t believe it. You will fool some of the people some of the time – but ~I do not believe you’ll fool those experts and artists once they are more familiar with the technology you describe. Then, like any form of legerdemain – we will simply enjoy the illusion – but ultimately not be fooled that it is real. I bloody hope so anyway.
I think that to some extent the infinite monkey theorem is true (and have an upcoming post about that coincidentally), but I think it also goes beyond that. Computers are beginning to simulate human brains.
That’s why I made the point about intent. There are some things that computers will probably never do, like strike out at a little league game or have a failed marriage or be trilled by the birth of their child. What makes our creations special is that they are ours, that we intended to create them and not something else.
In the end, the infinite monkey theorem is about curation, not creation. So why should it matter if computers can perform many tasks that we thought were uniquely human as long as the end result is meaningful?
– Greg
Algorithms even write books now. It takes them about 20 minutes to write it. http://singularityhub.com/2012/12/13/patented-book-writing-system-lets-one-professor-create-hundreds-of-thousands-of-amazon-books-and-counting/
Cool! Thanks for the link.
– Greg
Hi Greg – fascinating stuff, and really stimulating. In one sense we can look at this as just the next step in machines facilitating what humans want and need to do. It’s just that we have come to accept creativity and original thought as “ours”, and feel a sense of threat when soulless machines tread on our turf. I think that’s the wrong approach, certainly when faced with something inevitable. Why not embrace it? Why not reach creative options sooner? It doesn’t prevent human creativity, nor the true advantages that we will retain – discernment and decision. People with diverse opinions and original thought will decide what works for them and what doesn’t, both in what they like and what they buy. There’s no way that I see people morphing into automatons, just a hell of a lot more creativity and interest.
Thanks Kevin,
To be honest, I have mixed feelings about the whole thing, much like I do about Big Data and privacy, but one way or another, we’re going to have to deal with it. We’re all going to be increasingly “flying by wire.”
– Greg
Hi Greg,
A great provocative post.
For me, with an early back ground in computer/electronic music, I think I have been dealing with this concern/topic for 25 + years. I use to give lectures to local music clubs in the mid 80s to attempt to relieve their fear of electronic music and let them know that they will not being going to classical concerts that had no musicians. Now I suppose could add no composer. 😉
Even in the late 80s with our Amiga computers, electronic composers like myself were eager to use algorithms to generate midi data including rhythmic, pitch and even tonal information to include in our recorded or interactive compositions. At the time the critique most often levied against us was that we were not the creators, the computer was. My defense was – Intent. “Being a composer” – ie creator, “does not necessarily mean that I need to deliberate over every sonority, every pulse, every little sonic event. It is my intent to illicit an intellectual and emotional response through my application of craft that makes me a composer”
I now work in a much wider media and communication spectrum (Marketing,Comm and learning development) and as people more and more have been identified as data points to move one way or another, I find that now I, like you, am more uncomfortable about the shift to algorithmic creation and big data. (The two I think are related). That is not to say that I deny it or even deny the potential business benefits. But I do think there is a human downside – just because this stuff run amok will sap a little humanness out of us. To make this even more conflicting, as a composer, I would not hesitate to turn to algorithms to help me create an electronic piece. 🙂 it really comes down to your last statement:
“Our ability to form our own intent. It is only through creating a purpose that is uniquely our own that we can fully embody the human spirit.”
The sense of discomfort I think comes from knowing that algorithmic creation (and big data) will be utilized without that intent.
Stephen,
Thanks for adding some context that I, as an utterly unartistic person, am unable to do.
My concerns go a bit deeper than what you’ve described. In a very real sense, we are creating a new life form, without giving it much thought. Most of the discussions you have about machine intelligence go something like:
“But computers can’t…”
“Actually they already are..”
And then kind of a stunned silence.
It just doesn’t seem like we’ve had anything like the level of discussion that we had about earlier technologies, like nuclear power or even TV. I’ll be writing a lot more about this topic next year.
Thanks for all your support. Have a great New Year!
– Greg
We seem to be struggling to find enough jobs to give humans a living wage already, even just in the industrialised world. What are we going to do if we end up with a planet of nigh on 9 billion broke, unemployable people, overseen by a tiny wealthy elite that own all the robots? What will that do for democratic, peace loving communities? Suddenly The Matrix isn’t seeming quite so preposterous after all!
Fran,
That’s a very important point one that I’ve written about before (https://digitaltonto.com/2012/are-computers-the-problem/) and one that I plan to spend a considerable amount of time on over the next year.
In short, I think two things are key:
1. There are some things that computers will never do. They will never strike out at a little league game, have their heart broken or see their child born. It is through seeking fulfillment that we form intent and human’s role will increasingly be enhancing the lives of other humans.
2. We need to reorganize our institutions to adapt to this new reality, by leveraging the respective skill sets of both machines and humans. In reality, we’ve been doing this for a century, but as the skill sets of machines are exponentially increasing, we’re going to have to completely rethink how we make decisions.
As I said, I’ll be spending quite a bit of time on this in the coming year. Expect to see two posts out in the next month or so. One about the “Simulation Economy” and one on revisiting the “Productivity Paradox.”
Until then, have a safe and happy New Year!
– Greg
Hi Greg,
First time I’ve had the good fortune to run across your blog – good stuff. I don’t have any specific knowledge about Picasso’s goals, but wonder about the accuracy of the statement that every creative endeavor has its purpose. Maybe there is something about great creative endeavors that cannot be replicated by the computer overlords.
I think that great artists like Dali, Hemingway, Michelangelo, Mozart, etc. didn’t have a purpose as much as they had inspiration. We’ve all heard about “starving artists”, and there are many who don’t achieve commercial success in their lifetimes. Maybe these folks grew an illogical ‘sense of purpose’ where they just had to get their art out, regardless of it being a bad career choice. They saw a flower and had to paint it, they heard the music in their heads and had to play it. (Rodriguez?) So, if following their human inspiration makes no logical sense, a computer would have to be programmed to be inspired and driven to do something without a rational purpose. I suppose given infinite processing power that the infinite monkey theorem could be applied, and in some cases succeed, but whose human sensibilities would curate all of the infinite machine nonsense? I’m not sure using that theorem is realistic any more than saying an alien could land and improve on / relieve us of all human effort. Possible, but how probable?
There’s a post on my EdRodPOV . com blog with my thoughts on how breakthrough innovation usually requires inspiration, and how the best way to “get” inspiration is to train your brain through the practice and sharing of art. Feel free to comment if you have a chance to read it. I’m following you on Twitter now so, “I’ll be back”. Best, E
Thanks Ed.
You may be right. Purpose might be too strong a term in that not every creative endeavor has a clear teleology, but I do think that every creative endeavor has an intent, even if it they might not be able to articulate it outside of the creative work itself. In fact, that was the basic point I was trying to make, that what is essentially human about creative works is the intent.
However, I disagree that a computer can’t be programmed to do something without a rational purpose (after all, what is a random variable). What they don’t have is motivation that comes from life experience. Computers don’t strike out in little league, fall in love, have children and so on.
This is part of the reason why algorithms work very well within specific genres, but fail when it comes to creating new ones.
Hope to see you again.
– Greg
This is probably my favorite article of the last four months. That’s because its subject, its subtopic ranging, and its style all appeal to me so much.
All the more interesting, then, that I have a huge objection to it as well. The idea of “creativity” is perhaps accidentally confused in a way that needs clarification.
The function of DESIGN is to create a direct correspondence between the way something has been formed and the way that it will be used. While many designers are imaginative, clever, innovative, etc. etc. in their solutions to that problem, design is not at all “the same thing” as art. Instead, many artists include design in their work, and many designers include an art-making mode in their work. There is an enormous universe of art, however, that is completely unconcerned with design and that still fully embodies the core impulses and experiences of the inventive modalities of artmaking. The point here is that not all rectangles are squares: creativity can spawn both art and design. But art is not synonymous to design. It is design that is being “fingerprinted” and computerized, not art.
Because of the many excellent comments already made, I will cut short here and add the one other thing that I think must be said. There is an enormous world of art, including great and terrible work, that has produced artifacts which are NOT a solution to a problem but instead are evidence of a Solving Process. In this work, taken case by case, it is possible that neither the process nor its progress was anything known to the artist before the work was being done. Motivators to do this kind of work are not necessarily “requirements” nor “goals”. And doing those speculative explorations may or may not ultimately “result” in output that intends to address anything but itself. Normally, what happens is that, each for their own purposes, analysts, critics, teachers, collectors, etc. are concerned with “recognizing” such artifacts. Computerizing their cognition is an entirely separate matter from doing the work done by the artist — but also, many working artists are themselves able to assess the states of their work from both the intrinsic and extrinsic perspectives, and there is no universal rule about how, as an artist, one should respond to extrinsic feedback that you’ve generated yourself!
All that said, thank goodness for great design. Without it, a lot of art might simply be inaccessible to too many people.
Great points Malcolm. I agree that computerizing aspects of artists work does not hinder art. The key to great art, as well as design, is the solving process that you noted above. Whatever the capabilities of computers, we still get to choose our own problems. Whether we use technology as part of the process or not isn’t really what’s important.
– Greg
> Being able to search and find elements of art and culture is one thing, but
> can computers appreciate quality?
> Mike McCready has shown that they can. His company, Music X-Ray, offers
> a service where composers can upload their music to evaluate its hit potential…
So the writer’s mixing two points. (What else is new?) Quality is not hit-potential. I have no doubt computer algorithms can be created to predict, with some accuracy, hits, because, like the weather, they follow certain predictable patterns (to a point–which is why weather forecasting, though getting better, still has inaccuracies). However quality is another thing, and is actually antithetical to “hit potential.” As Steve Jobs once commented (paraphrasing here) something that’s so fashionable today will seem repulsive next week. So much of the “hit” music if the past is awful. Fifty years from today, the same will be true of today’s music, and something of barely modest popularity and appreciation today will be regarded as the finest content of our time. I defy computers to predict *that.*
I really wish these article writers would think a little more and a little deeper before posting their drivel.
Please try to be a bit more civil or I won’t be able to approve future comments.
As to your point, if you read a bit more carefully you would have noticed the reference to David Cope’s work, whose algorithms generated music that drew acclaim from top critics. When played along with music from human composers, professional music critics weren’t even able to tell which was composed by a human and which was composed by a robot.
– Greg
I guess the question we have to ask is does generating music that draws acclaim equal creativity. Just because something sells does not mean it truly has value, it definitely has perceived value that is clear but will that value stand the test of time? Or is creatively really in the end about the artist and their struggle to create it? Does the fact others enjoy it and find value in it just a bonus? (I say this realizing to a starving artist many times they would prefer other find lots of value in there work.)
Ok back to reality…
Your response “It just doesn’t seem like we’ve had anything like the level of discussion that we had about earlier technologies, like nuclear power or even TV.”
I completely agree with this and find this to be a very disturbing thought. But I question our ability as a nation to have a thoughtful discussion in our mass media. And even if we could that would not stop the market from developing and delivering technology that replaces humans if it delivers more efficiency.
It feels like we are on a down hill slide where efficiency is king and if computers can produce works of art that sell… then get rid of the artist because they cost more. But the great things about being human is another “artist” will come along and make art capturing how computers killed all the artist;)
I guess what I am saying is our natural urge to create needs to be tempered a bit by community and thinking about how the things we create will benefit or hurt society. Right now the only rule we use is, does it sell? If that continues to be out only measurement we may just create ourselves out of existence before too long.
To be honest, I don’t think its a useful exercise to try to define whether algorithms like David Cope’s are creative are not. There are various definitions of creativity and Cope’s compositions would meet some and not others. The Infinite Monkey Theorem has been around for at least a century and nobody has really answered the question it poses definitively.
So I think it’s just best to say that algorithms can be creative “in some way” and leave it at that.
– Greg
Haha, wow you grabbed one line. Please do me the favor of reading my response before responding.
I believe you completely missed my point, creation is not about whether the work is creative it is about the person who made the art. Art is first and foremost very personal, it is the process of making it that is central. If that were not so local art stores would not survive. We have turned art into a consumption actively and in that turned our urge to create into something that is driving us off a cliff by measuring the “creativety” of creation by whether it makes loads of cash or paid pundits think it is “good” assigning it value.
That is why we are staring AI in the face because our urge to create is now driven by the market so we no longer ask is what we are creating good for society, we only ask will it make money.
Jeff,
I’m sorry that you found my response insufficient, but I receive dozens of comments per week and do my best to respond to each one. While I cannot address each and every point, I do try to respond to the ones that I feel are pertinent.
– Greg
One thing I think is interesting is that computers seem to be trapped within the parameters we set them, variability included. I’m not talking about originality necessarily (can there really be originality?) but the ability to foresee, or feel in tune. A lot of what makes human creations timeless is that they resonate with a location, a time, an ideal shared by people. And from there, there are a number of variables which only come into play in those moments, in those conditions. I guess I think the particular human brand of creativity is that our experience, which is always catapulting forward into [some] degrees of unknowing while being conscious and reflective of where it been, and where it is now. I’ve read of articles critiquing such software as used by these record labels because they gradually homogenize genres. I can’t think of particular names now, but a Google search should render results for some research indicating that music genres (barring rock and hip-hop) are becoming less complex and more homogenized.