A frontier in machine learning seeks to mimic human learning by assembling computers into teams to solve problems. This type of deep learning, called “reinforcement learning,” allows machines to acquire knowledge, take collective action with that knowledge, and observe the consequences on their world and on their teammates. 

This machine deep learning environment unsurprisingly mimics the sort educators try to create for students. I described the characteristics of these environments in my last post about deeper learning in schools.

The infrastructure for machine deep learning is available for all and priced to move by Amazon, greatly improving the access for researchers and scientists to create machine teams to tackle the sorts of complex problems–how to achieve the UN’s sustainable development goals, how to create meaningful art in our internet-fueled, content-drenched landscape–that we hope they can solve.

Machines learning in a social and authentic context not only holds more promise, it’s perhaps a moral imperative. Looking back on more rudimentary machine learning will make this evident.

Early Machine Learning and Its Perils

A computer can examine a stack of photos and identify which contain, for example, dogs, much faster than a human can. The process to teach the machine involves feeding it a stack of photos known to contain dogs, a stack of photos known to not contain dogs, and telling it which stack is which. The machine has learned, and can now efficiently sort for us.

Of course this “learn’d” machine could be useful, but it will have limitations. For example, it might not be able to recognize a cartoon dog as a dog, which isn’t a huge deal. But it might categorize a wolf as a dog, which could have consequences, if a human relies on that machine to sort animals according to threat level.

This sort of machine learning is discrete–the learning of facts or patterns is separated from the messiness of a contingent universe as well as from the emotional and experiential luggage that, for humans, make facts matter. For much of human history, the feelings and physiological responses of an encounter with a growling dog vs. with a dog that wags its tail reveal themselves to a child somewhat simultaneous with the knowledge about which species the four legged animal they are looking at actually is.

The discreteness of machine learning is a key aspect of the fear of machines taking over the world. Without the co-development of identity and social and emotional intelligence, machines can’t comprehend the consequences of learned facts in the profound way humans can. Machines designed to optimize without the construction of a conscience will be psychopaths and will construct The Matrix.

Infrastructure for Human Deeper Learning

Of course, the same is probably true for humans: someone a comprehensive grasp of a wide set of facts, without context or a sense of consequence, would certainly possess some sort of pathology. Or if not that, at least a dangerous sense of emptiness.

So, as noted in last week’s post, fact acquisition accompanied by purpose, connections, and intellectual depth is what can lead to deeper learning and, perhaps also, a sense of fulfilment or the feeling of flourishing, as Aristotle put it.

But I can’t help but think that the “classroom” full of computers working together to solve an authentic problem, no matter how much they look like a team, would be missing a key element: wonder. 

The recovery of the ancient poet Lucretius’ work On the Nature of Things, described by Stephen Greenblatt in The Swerve, provided early Renaissance scholars with a key insight into how scientific progress and humanism could co-exist. Scientific knowledge does not eradicate the wonder typically inscribed in superstitions, oracles, and tea leaves. Rather, Lucretius’ poem suggests that “knowing the way things are awakens the deepest wonder.”

How is it that Walt Whitman seems always to be responding to today’s most worrisome fears and highest hopes? And doesn’t it seem like a machine will never write such a poem?

WHEN I HEARD THE LEARN’D ASTRONOMER.

(Check out this Portuguese translation and excellent comic rendering of the poem.)