What if computers are now being creative?

Sarvasv Kulpati writes about a fascinating robotics project, where researchers were exploring the different ways a 6-legged robot could walk – the idea being to to find the least possible foot-ground contact needed, so that the robot could keep walking even if damaged.  The robot was programmed to try all sorts of gaits (walking styles), and to look for the lowest amount of contact.  

You might wonder if the robot could manage using only two or three feet; a few people might even consider one foot as viable (more a drag than a walk, perhaps).  But the robot managed to find a way to walk to with no feet touching the ground at all! It turned out that the robot had flipped itself over, and managed to walk on its ‘knees’ – a solution that had not been foreseen by the human researchers. I want to argue that this is creativity – of a limited sort perhaps, but creativity nevertheless – and that it has been executed by a robot. 

This may seem heresy to many, and the standard argument is that robots and computers can never be creative; they can merely do as they are programmed.  But the evidence is mounting against that view, which is itself just the latest iteration in a mistaken view of what’s really important in being human.  This is latest chapter in the millenia-long story of us realising that what once seemed mysterious or magical can be brought into the domain of human understanding.  The story has many spin-offs and implications. Freud famously wrote that humanity has in the course of time had to endure… two great outrages.  The first, he claimed, was our realisation that we are not at the centre of the universe, despite the appearance of the sky, and the rising and setting sun.  The second was the relegation of man from a special creation to a descendent from the animal world,  as shown by Darwin and Wallace and so many others.  Freud claimed that his own insights into the way our minds work are a third and most bitter blow in that each of us is not even master in his own house, but that he must remain content with the veriest scraps of information about what is going on unconsciously in his own mind.  And Freud’s notion of the unconscious, if not his other ideas, is now widely accepted.   Over recent decades we’ve largely come to terms with these three blows, but AI brings a fourth one here, now that our own generative, creative, artistic powers – the things that we often hear machines will never do – are clearly within the scope of technological progress.  I want to discuss two examples and briefly consider what it might mean for schools in years to come.

The first example is well known and widely reported on – games such as Chess and Go.  It’s a fact that the best programs now play better than the best humans; and that computers can search massive databases and explore branching possibility spaces in ways that are utterly different to human game-playing, which seems to reply on intuition and pattern spotting.   Given the different styles, perhaps we can say that humans still have something different.  Or at least, we could  say that until recently.  Because what’s absolutely fascinating is that AlphaGo Zero, the strongest programme, does not taking the traditional brute-force approach, but instead just starts from the rules, and works out, by itself, the best strategies.  That’s right – no-one codes the best strategies, but the machines nevertheless learn them.  Ian Sample notes that when AlphaGo Zero played against the previous best Go program  (which itself beat the strongest ever human 4 – 1) it won 100 – 0.  When generalised to Chess, the programme took just four hours of self-play and learning before beating the then-world champion chess program, Stockfish 8, over a 100 game series.  So in both Chess and Go, AlphaGo went from nothing to world-beating with no human input on strategies.  This is, I think, the very essence of creativity.  Historian Yuval Harari has even argued that when chess players come up with creative new ideas, such is the current state of the art that there now is immediate suspicion that the ideas are machine-generated.  So creativity in games has certainly left the humans-only realm.

My second example here extends to the more usual creative domain – the Arts.  There are some striking cases – such as, for example, the fact that an artwork created using AI has been sold at auction for $432,500.  It’s reported on DesignBloom that the artwork was produced using an algorithm and a data set of 15,000 portraits painted between the 14th and 20th centuries – which makes it sound much like the earlier iteration of chess-playing machines, relying on brute-force and speed.  If so, the high price may say more about novelty value than artistic value, and it may be that there is no possibility for machine learning here – after all, there are no rules of Art like there are rules of Chess, and no clear ways to ‘play’ Art to see who wins.  But even so, there are some remarkable things happening.  Here are some images from Bloomberg Business week.
 

While these pictures are hardly great, if a student painted them, I doubt we would say ‘robotic’ or ‘unimaginative’ – we would likely say things like inspired by impressionists, or homage to Monet or something.  Perhaps the question should not be is this Art?  but the far more interesting is this good Art?  or to put it another way, not is it creative?  but how creative is it?  It’s not just painting, either – writing music is falling into the digital domain.  Below is a short clip of music generated from the DeepBach computer project.  Not brilliant, but better than the vast majority of humans could do.  

(As an interesting aside, journalist John Vincent has argued jazz will be the first genre where there will be real breakthroughs.  We tend to think of the improvised, unstructured nature of jazz as protecting it from machine learning, but it turns out that what’s really hard to learn is the overarching structure of a piece — the way a symphony might return to a central
theme, for example, repeating the melody with added tweaks and flourishes so that it moves into new territory. When it comes to jazz, this is less of a problem. People don’t expect structure from jazz.)  

It’s still not clear if computers will ever compose great music of any genre, and even harder to imagine that computers will ever create great plays – but it now seems to be an empirical question, not a theoretical one, so we’ll have to wait and see.  The notion that computers can be creative may be unsettling, in the same way that Freud described the effect of other advances, but maybe that’s just the shock of the new. 

Does this then render the Arts somehow in danger of obsolesce, like we so often hear so many current professions are?  My own view is that this one one worry we really do not need to indulge.  Even if (that a big, big if) computers end up being genuinely creative across the Arts, that won’t matter, because the arts were never about ‘getting stuff done’.  An engineer might be displaced if machines can build bridges, but humans will still enjoy composing, painting, dancing even if machines are also composing, painting, or dancing.  It’s about the doing – and perhaps the sports analogy is the best one.  Usain Bolt’s time over 100m is not diminished by the much faster motorbikes – it’s just a different sphere of endeavour.   So human creativity will remain undiminished; indeed if the economy is radically re-shaped by artificial intelligence and machine learning as some are suggesting, perhaps there’ll be far more time for us to pursue the Arts that were once seen as the highest calling. 

I must study politics and war that my sons may have liberty to study mathematics and philosophy, geography, natural history, naval architecture, navigation, commerce, and agriculture, in order to give their children a right to study painting, poetry, music, architecture, statuary, tapestry, and porcelain.  John Adams, 2nd President of USA, around 1800.

To my mind, the coming AI revolution will only augment the role of Arts in our society.  Computer programmers may be out of a job, but artists will be in greater demand than ever.