Ammar Mian

Creating High Standards for Yourself

The Myth of the Objective

Computer scientists Kenneth Stanley and Joel Lehman wrote an odd little book a few years ago called Why Greatness Cannot Be Planned: The Myth of the Objective. Pretty provocative title, right?

Stanley & Lehman wrote the book after an experiment they ran called Picbreeder. This website presents visitors with a handful of images and two options: Evolve or Build from Scratch. Either pick an existing picture and "evolve" it, or create a new picture.

You need zero artistic skills to evolve or create, because a computer algorithm does the work for you. If building from scratch, the program will randomly generate a sequence of colored strokes. If evolving, the program will randomly alter the existing image.

Over the course of thousands of generations of people evolving and creating on top of each other's previous evolutions and creations, this is the kind of work that resulted:

Keep in mind that all of these images started off as resembling nothing recognizable at all. Over time, with enough iterations and rounds of evolution, something "great" began to emerge.

Stanley & Lehman theoretically expected this to happen, but they really did not expect the degree to which permutations of small changes would produce really cool artwork.

Thus, their overarching claim: in complex domains, highly-specific long-term objectives can be misleading, since we don't know the precise steps required to achieve these distant goals. It's better to seek novelty and produce many iterations of whatever you're working on.

I. Seek Novelty: Super (Curious) Mario

Artificial researchers observed the same lesson take form when they set a program loose in Super Mario World's Mushroom Kingdom that possessed zero prior knowledge of the rules of the game. The intelligent agent's only goal? Stay curious.

Before every step it took, the program would predict the state of the world if it took that step. Afterwards, it would compare what actually happened to what it predicted would happen. It optimized its movements for maximum "surprise" -- the difference between what it thought would happen and what actually happened.

Across many generations, as every new "failure" taught it something new about the structure of its universe, this program eventually figured out how to beat the level.

The surprise principle isn't without its warts. Sometimes, Mario would get stuck just bouncing into walls. But the researchers pointed out that "We can’t just go for what’s novel without worrying about what’s relevant, and we can’t learn what’s relevant without this curiosity bonus to drive our exploration."

II. Produce Many Iterations

Curiosity doesn't really work if you're playing one-shot games. Imagine if free solo climber Alex Honnold couldn't use ropes while training for dangerous climbs. He wouldn't be able to learn from his mistakes because he'd be dead. Because his tools allowed him to more safely experiment over and over again, he became the first person to climb El Capitan's 3,000 foot wall without ropes.

Your tools and your environment must permit you to iterate repeatedly, at relatively low cost.

This principle appears in the art world, too! In his awesome book Old Masters and Young Geniuses, David Gelertner claimed that there are two distinct life cycles for groundbreaking artists: the Experimental and the Conceptual.

"Experimental artists build their skills gradually over the course of their careers, improving their work slowly over long periods. These artists are perfectionists and are typically plagued by frustration at their inability to achieve their goals."

The primary example of the Experimental Artist is painter Paul Cezanne, who spent decades producing thousands of paintings, many of which depicted the same scene. He grew obsessed with perfecting the representation of the foreshortened contours of three-dimensional objects. He didn't achieve fame until his 50s and 60s.

In contrast,

"artists who have made conceptual innovations have been motivated by the desire to communicate specific ideas or emotions. Their goals for a particular work can usually be stated precisely, before its production, either as a desired image or as a desired process for the work's execution."

The primary example of the Conceptual Artist is Pablo Picasso. Picasso rose to fame in his 20s, when his Cubist paintings shook the art world. He had a clear idea of what he was trying to do -- to paint objects that people understood without the constraint of "realism" or visual accuracy.

But the two models, and their archetypal artists, were not so different after all. Cezanne produced countless paintings before he struck gold. But so, too, did Picasso. His notebooks and journals are riddled with hundreds of preparatory and exploratory sketches for nearly every single work he ever produced.

So what's the difference between the Conceptual and the Experimental? Unclear.

I find the similarity more revealing -- Cezanne and Picasso both produced an incredibly high volume of work, even if only a small percentage of it ever saw the public eye. They benefitted from iteration-friendly technologies (paints, paper, graphite, brushes) that were relatively inexpensive and easy to acquire.

Conclusion: Extreme Programming for Non-Programmers

If you work with software, none of the above should be too surprising. The entire profession of software development revolves around building programs and interfaces when you aren't certain upfront about the What and the How.

Much of modern industry is still stuck in the old Waterfall approach: create specific goals, set a distant deadline, and then ship the final product at that distant deadline.

This approach has proven to not work in most domains. It has been the source of way too many burnt-out and defensive engineers, shitty products, abandoned high-budget projects, and dead businesses.

Programmers realized there's a better way. This collective wisdom was compiled in 1999 by Kent Beck in his book Extreme Programming Explained: Embrace Change.

Beck's main motivation here:

"Prepare for success. Don't protect yourself from success by holding back. Do your best and then deal with the consequences. That's extreme. You leave yourself exposed."

And how do we do this? By creating high standards in the small.

Creating high standards in the small looks like taking pride and paying curious attention to every little thing you do.

  1. If you notice something intriguing, run with it (in the small).
  2. If you notice something wrong, poke at it (in the small).
  3. If you need help, seek it out (in the small).
  4. If you see someone struggling, reach out (in the small).

What's the smallest way you can test, attempt, validate, achieve something? Design your system around a progression of small iterations. And then go be creative, curious, and open-minded.

By reducing the grandiosity of our aims, we can end up accomplishing very grandiose things. We can focus on working iteratively, trying our best on each iteration, and making each iteration short enough so we can iterate again and again and again.

May 2019