Learning to Learn, aka Prediction Error

They say failing is the key to learning, and a lot of cliché sounding stuff that feels too feel-goody to be true, more like self-help hype to sell books or make kids feel better about themselves without real accomplishments. The thing is, like many things, there is actually a nugget of truth in this otherwise cliché sounding message.

If we take a look at both information theory, ala Claude Shannon, and some basic neuro-biological functions, as well as a hint from quantum computing, it all makes a lot more sense - seriously, indulge me a moment and I'll explain. In information theory, "information" is, pretty specifically, that which informs us of something. Which is to say, something we didn't know already. The "surprise" quality of some stimuli or data is the key to how informative it is.

Similarly, our neurological functions are primed to use the minimal necessary resources to function, which is the result of evolving on a very limited supply of calories, and follows the "path of least resistance" in a more fundamental way that physics tends to do. Our full perceptual, cognitive, and learning faculties don't really kick in when we see the same ol' stuff we already know, since we can more or less autopilot through that.

However, when we run into a prediction error - "oh crap, that's not what I expected!" - our brain lights up and really pays attention, including perceiving and ingesting far more raw information, and creating new neural pathways to etch the lesson into our wetware so we can make use of this information in similar future situations. This is why trauma is so damn memorable but that school lecture puts you to sleep.

It's not exactly "failure" as we might interpret it, but prediction error which kicks on the whole learning and remembering apparatus. This is also why methods like the memory palace work - as they tend to combine high levels of novelty with the more boring thing we want to recall, like a clown standing on a pink unicorn holding a sign with an otherwise dull sequence of numbers. This goes directly against our usual stimuli and prediction, so helps etch in the pattern far more effectively.

The quantum computing aspect is a bit more subtle, as it's more a means than mechanism. In the most notable algorithm for using quantum computers to factor large numbers, it is first necessary to make a random guess. How accurate this guess is has no relevance, we just need to make some guess to start the process. Why? Because then we have something to measure against, to predict, measure the accuracy of the prediction, and use as feedback to adjust our possible answer.

Life in general works similarly. Thinking about something, leaving it in "quantum superposition" is one thing, but there's an old saying that if you want to choose between two apparently equal outcomes, you flip a coin. Once the coin is in the air, you will become aware of a bias of desire towards which way it lands, and thus know your preference - despite them seeming equal prior to tossing the coin. Once we step out onto a path in a direction, it becomes pretty apparent, typically, whether that path is leading us a direction we want to go or not, even if all paths seemed relatively similar or equal prior to setting out.

This is due to the feedback we get from the action and direct experience of the thing, which is not present in our mental models of it - particularly if it's novel and our predictions are wrong. Once we set out, we immediately start getting feedback, which can inform our prediction system of alignment or error towards a certain goal. Thus, a random choice, just to get moving, is the way to trigger prediction success or error - and thus find (real) information, and begin to learn and move towards the real goal. The phrase "just do it", though co-opted by a certain well-known brand, is an expression of this (the brand name in question being the name of an old deity which represented this principle).

Fear of failure is our worst foil here, as it prevents this initial movement, and so we can not learn or advance forward in any way, remaining in the quantum superposition of outcomes, where we receive no new or useful information towards any of them, and assuring failure by lack of action itself - a self-fulfilling prophecy. Another old saying is "the journey of a thousand miles begins with a single step", which speaks to the same principle of taking action to move forward at all.

Taking (informed) risks, trying new things, and "failing" (but not really failing as we might think of it, just failing to know what to expect beforehand, and with a strong possibility for error and corrections along the way), pushing ourselves past our known prediction ability and into new territory where it will likely falter, seeking past our comfort zones, and reaching beyond the known, embracing novelty, is the key to learning, growth, and advancement of all sorts - the meta or ur-principle of learning itself.

Hopefully this helps make more and more useful sense of a very valid, if cliché sounding, bit of sage advice. Be bold, be daring, try things you don't already know or have confidence of success at - and learn! :-)