Here’s last week’s BBC Future column. The original is here, where it was called “Why our brains love feedback”. I was inspired to write it by a meeting with artist Tim Lewis, which happened as part of a project I’m involved with : Furnace Park, which is seeing a piece of reclaimed land in an old industrial area of Sheffield transformed into a public space by the University.
A meeting with an artist gets Tom Stafford thinking about the essence of intelligence. Our ability to grasp, process and respond to information about the world allows us follow a purpose. In some ways, it’s what makes us, us.
In Tim Lewis’s world, bizarre kinetic sculptures move, flap wings, draw and even walk around. The British artist creates mechanical animals and animal machines – like Pony, a robotic ostrich with an arm for a neck and a poised hand for a head – that creak into life in a way that can seem unsettling, as if they have a strange, if awkward, life of their own. His latest creations are able to respond to the environment, and it makes me ponder the essence of intelligence – in some ways revealing what makes us, us.
I met Tim on a cold Friday afternoon to talk about his work, and while talking about the cogs and gears he uses to make his artwork move, he made a remark that made me stop in my tracks. The funny thing is, he said, all of the technology existed to make machines like this in the sixteenth century – the thing that stopped them wasn’t the technical know-how, it was because they lacked the right model of the mind.
Jetsam 2012, by Tim Lewis (Courtesy: Tim Lewis)
What model of the mind do you need to create a device like Tim’s Jetsam, a large wire mesh Kiwi-like creature that forages around its cage for pieces of a nest to build. The intelligence in this creation isn’t in the precision of the craftwork (although it is precise), or in the faithfulness to the kind of movements seen in nature (although it is faithful). The intelligence is in how it responds to the placing of the sticks. It isn’t programmed in advance, it identifies where each piece is and where it needs to go.
This gives Jetsam the hallmark of intelligence – flexibility. If the environment changes, say when the sticks are re-scattered at random, it can still adapt and find the materials to build its nest. Rather than a brain giving instructions such as “Do this”, feedback allows instructions such as “If this, do that; if that, do the other”. Crucially, feedback allows a machine to follow a purpose – if the goal changes, the machine can adapt.
It’s this quality that the sixteenth century clockwork models lacked, and one that we as humans almost take for granted. We grasp and process information about the world in many forms, including sights, smells or sounds. We may give these information sources different names, but in some sense, these are essentially the same stuff.
Cybernetics is the name given to the study of feedback, and systems that use feedback, in all their forms. The term comes from the Greek word for “to steer”, and inspiration for some of the early work on cybernetics sprang from automatic guiding systems developed during World War II for guns or radar antennae. Around the middle of the twentieth century cybernetics became an intellectual movement across many different disciplines. It created a common language that allowed engineers to talk with psychologists, or ecologists to talk to mathematicians, about living organisms from the viewpoint of information control systems.
A key message of cybernetics is that you can’t control something unless you have feedback – and that means measurement of the outcomes. You can’t hit a moving target unless you get feedback on changes to its movement, just as you can’t tell if a drug is a cure unless you get feedback on how many more people recover when they are given it. The flip side of this dictum is the promise that with feedback, you can control anything. The human brain seems to be the arch embodiment of this cybernetic principle. With the right feedback, individuals have been known to control things as unlikely as their own heart rate, or learn to shrink and expand their pupils at will. It even seems possible to control the firing of individual brain cells.
But enhanced feedback methods can accelerate learning about more mundane behaviours. For example, if you are learning to take basketball shots, augmented feedback in the form of “You were 3 inches off to the left” can help you learn faster and reach a higher skill level quicker. Perhaps the most powerful example of an augmented feedback loop is the development of writing, which allowed us to take language and experiences, and make them permanent, solidifying it against the ravages of time, space and memory.
Thanks to feedback we can become more than simple programs with simple reflexes, and develop more complex responses to the environment. Feedback allows animals like us to follow a purpose. Tim Lewis’s mechanical bird might seem simple, but in terms of intelligence it has more in common with us than with nearly all other machines that humans have built. Engines or clocks might be incredibly sophisticated, but until they are able to gather their own data about the environment they remain trapped in fixed patterns.
Feedback loops, on the other hand, beginning with the senses but extending out across time and many individuals, allow us to self-construct, letting us travel to places we don’t have the instructions for beforehand, and letting us build on the history of our actions. In this way humanity pulls itself up by its own bootstraps.
8 thoughts on “The essence of intelligence is feedback”
Very well written!
If this were written at the time of Skinner, however, it would have said ‘operant’ wherever you wrote ‘feedback’ and would’ve been quite indistinguishable from any behaviorist essay of that era 🙂
Good to see that the fundamental insights of behaviorists, if not remembered, are at least constantly re-discovered, due to their basic validity 🙂
Not behaviourism at all…
For what it is worth, I don’t find this essay (or Power’s take on control theory in general) very informative. I’m not saying he doesn’t have a valid critique of Skinner, just that from this I can’t work out what it is. Can you summarize?
What might be related is that Skinner’s paradigm of choice is actually designed to measure response rate, rather than response acquisition – it minimises/makes invisible behavioural variety before the point of responding. We’ve been working on a paradigm that can restore/highlight that early exploratory stuff. See http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0037749 and http://www.tomstafford.staff.shef.ac.uk/?p=131
Thanks, Tom. Glad I’m not the only one who wondered what the relevance was. I thought I was too dim to get what he tried to say with that reference.
Does it allow us to die from seeing something so cute? It’s not a real kiwi but just the shape is LOLCAT adorable.
Keep it away from MIT’s Big Dog though. He just learned to toss around cinder blocks (charming)!
Nice description of adaptations that in the news are generally hijacked by outdated concepts; most don’t consider how far out the adaptations extend, i.e. can plants and even geology / soils be said to change in response?
Cybernetics and behaviourism, while covering similar ground, do so in different ways. Operant and feedback are not synonymous, and differ in the areas of purpose and control. It’s a great blog post, what wonderful gadgets!
The idea that a feedback loop is the essence of consciousness is the centerpiece of Douglas Hofstader’s 2007 book “I Am A Strange Loop.” Here is a man who explores feedback in all media: mirrors, video, computing, tomatoes, mosquitoes, and the human mind. I cannot reproduce his wonderful prose here, but you must read it!
I think it’s time for a popular reappraisal of Skinner and behaviorism. Chomsky’s critique of it is fine for philosophers, but for practicing neuroscientists and psychologists, operant conditioning is quite informative.
I think behaviorism falls into a category we might call “useful circular thinking”.
What is a reinforcement? The thing that is the animal’s goal.
What is the animal’s goal? A reinforcement. 🙂
The same useful tautology operates in evolutionary thinking (with survival and fitness being the dialectic pair). Rather than deny the circularity, scientists should say that some apparently circular “arguments” are actually experimentally useful rules-of-thumb (rather than logical “propositions”).