Learning paths of least resistance

Why does falling water form a droplet? Why does a paper plane always nosedive? Why does light travel in straight lines?

One of the most surprising things I learned during my physics degree was the Principle of Least Action. My favourite formulation is from 18th Century French mathematician Pierre Louis Maupertuis:

Nature is thrifty in all its actions.

The Principle of Least Action is quite broad: it applies to light, which takes the shortest path to its destination: hence travelling in straight lines. It applies to paper planes and water droplets, which form the most aerodynamic shapes they can when falling. It applies to beaches, and muscles, and stars.

In my mind, it establishes a kind of natural minimalism to all systems of control: light touches work better with complex systems.

I recently started reading more from Astro Teller, director of X. Astro applies the principle of least action, in a sense, to building cultures which not only embody creativity but successfully systematise it: baking it in to a company. You can listen to an Exponential View podcast featuring Astro Teller here.

The high-level principle is that people in organisations adapt to succeed within the organisational culture. Adapting the culture is the best lever to adapt individual behaviour within it. There are strong parallels with learning management here, where it is far more effective (and scalable) to establish and adapt cultural practice than to negotiate a social contract with each individual learner.

If you do it right, they'll start developing more and more extreme versions of that same culture, and guard it more fiercely than you ever could.

Least action and learning

Learning is the optimal state

Human brains are bundles of neurons, Hebbian dynamics fire-wire, but what is being learned?

(cf learning IS innovation and innovation IS learning. The feeling of learning is the same as the feeling of discovery; they are the same thing)

Direct attention

What is to be learned is about making things salient in the context: train people to look for salience

Frame discovered abstractions

Don't teach abstractions but, once they emerge, name them. Names are powerful and give flexibility.

We think people are inherently ethical if you give them half an excuse to be good people. But training them on the basics of ethics gives us a shared vocabulary and platform for discussions. (cf concretions to abstractions)

Set audacious goals

Motivate with either something to be made, or something to be discovered. People either like to go down or up the concretion chain. Some like to do both: have both strands. Let them go where they want to go. Nurture outlandish ideas: nothing is magic, everything is practicable.

Invert the system of control: maximise empowerment

Traditional education systems are gatekeepers to qualifications. Students are forced to direct their attention to learning in order to sit an examination and get the qualification. Invert this.

Be there for them

Remove extraneous complexity and pressure. Self-reliance comes out of self-love. Model it and give it.

Be the system you want your learners to operate in

Astro Teller "I feel that X is at a metal level itself a moonshot". Be the system you want your learners to operate in. Be the people you want your learners to be.

Businesses are cultures. People respond to the unconscious and conscious level signals that happen inside a culture that tell people: here's how we really want you to behave. cf cultures of learning & 'what gets learned'

"(why leave grafitti on the walls?) Don't polish, this place is a work in progress." cf team and learning incentives: "one of ten thousand signals as to what the culture is and how to behave. We decide what's important and then have to pick a thousand signals that intimate this"

Focus on the process, not the outcome

Edison "I've not failed 10k times, I found 10k ways to not make something work"

If you're serious about innovation, you would be lucky to succeed 1% of the time. That's just a rule. The real question is, how efficiently are you going to throw out the other 99%?

Outcome -> protect the outcome. Process -> forget the outcome. If we're intellectually honest, we can find Achilles' heels in our projects. You can only invert the system in this way if learners are in control or else they're working to your outcome.

5 principles: iterate; perspective shift; smart risks; dispassionately assess; reinforce these principles.

Tight learning loops with dispassionate assessment. Audacity also requires perspective shifting and smart risks.

X celebrate things that were great to have tried; not whether they worked out or not. (cf learn to try, not to succeed)

As soon as you reward people on the outcome rather than the quality of the experiment, the innovation is dead.

Focus not on learning things but learning faster

Tune the global system of learning through metacognition. Then tune the self-regulation system that trains metacognition.

Shape the chaos

("How do you turn it into an organisation/direction that isn't chaos?")

Well, it's a kind of managed chaos. One analogy is Charlie and Chocolate Factory: the magic doesn't come from Wonka, it comes from the Oompa-Loompas. If you organised them into straight lines and not misbehave, the magic would leave the building. On the other hand, they're sufficiently crazy and disorganised that they need to be a little protected and lightly organised. Appreciate the subtlety of being responsibly irresponsible: give people the space to think differently from the rest of the world and from each other. It's celebrated but shaped.

Apply structure to meaningful entities: , e.g. "a 'moonshot' has to have three elements: 1/ a huge problem with the world we can name, describe, want to solve. 2/ a radical solution - a science-fiction based thing. 3/ some kind of technology core that gives us some belief we could make some progress."

—- notes —–

Nurture outlandish ideas until they become everyday

X is a decade old

I feel that X is at a metal level itself a moonshot

"It sounds totally crazy" to "that's gonna be a thing" - Waymo (self driving cars, $2b raise recently) and Wing (drone delivery) as examples.

X has gone from "would it be possible to make a place with a strong positive feedback loop with employees and parent company, and the world?" in 10 years, to "yes"

Came out of a fear from Larry Page/Sergey Brin that Google would slow down (despite 20%)

X was created to go make new things. It hands things to Alphabet which can be good for the world, have a hard tech core, and can help Alphabet grow financially too.

"We've looked at around 2000 ideas with some seriousness (weeks to months of work). Out the other side, maybe 10.

Google Brain was the first major thing that left X."

It's worth taking a real run at things - Malta is gridscale energy

Edison "I've not failed 10k times, I found 10k ways to not make something work"

If you're serious about innovation, you would be lucky to succeed 1% of the time. That's just a rule. The real question is, how efficiently are you going to throw out the other 99%?

If we're intellectually honest, we can find Achilles' heels in our projects.

(This inverts the traditional logic of business - cf learners in control)

"Either I'm not going to do this or I'm going to get really good at it." - on becoming a CEO.

Why when you tell people to do the stuff in business books, put the posters up etc, why doesn't it happen?

("How do you turn it into an organisation/direction that isn't chaos?")

(How do execs that come from more trad backgrounds handle it?)

Different ones have to learn different things. E.g. Wendy Tan White: technologist, entrepreneur, investor. I was telling her "You've got the responsible part. I know you've got the irresponsible part in you but you need to let it out." This was a signal that she could unlearn the self-stifling she'd learned her entire life. "I want to stop self-stifling, I'm ready to stop self-stifling" (cf education approaches that explicitly permit authenticity, and "you can")

5 principles: iterate; perspective shift; smart risks; dispassionately assess; reinforce these principles.

Tight learning loops with dispassionate assessment. Audacity also requires perspective shifting and smart risks.

X celebrate things that were great to have tried; not whether they worked out or not. (cf learn to try, not to succeed)

As soon as you reward people on the outcome rather than the quality of the experiment, the innovation is dead.

(cf learning IS innovation and innovation IS learning. The feeling of learning is the same as the feeling of discovery; they are the same thing)

Larry and Sergey created a willingness to do X. It's not that other businesses couldn't do it; it's a willingness to have 99% of experiments turn up as zero.

If the computational agriculture group were only able to pick one plant, one thing, rather than "let's look at 20 simultaneously and scrappy and iterate with farmers"…

A moonshot has to have three elements: 1/ a huge problem with the world we can name, describe, want to solve. 2/ a radical solution - a science-fiction based thing. 3/ some kind of technology core that gives us some belief we could make some progress.

Diversity of thought, gender, race, veterans - perspectival, see the world differently (cf community)