Cabinet of Wonders

Share this post

☞ Emergence, Program Synthesis, and Evolution

arbesman.substack.com

☞ Emergence, Program Synthesis, and Evolution

These are a few of my favorite things...all combined into a single piece of research.

Samuel Arbesman
Aug 24, 2021
Share this post

☞ Emergence, Program Synthesis, and Evolution

arbesman.substack.com
File:People watching flock of birds from bird viewing deck.jpg
Birds flocking.

The ALIFE 2021 conference—a conference all about artificial life—happened last month. And I was struck by one talk in particular: “Generating Agent Based Models From Scratch With Genetic Programming.” And that research explores a fascinating line of thinking.

In complex systems you find emergent phenomena, where the rules of individual agents combine in novel ways to result in complex and sometimes seemingly unexpected behavior. But what if you could automatically generate these rules of the agents? Specifically, what if you could use evolutionary computation to evolve the small ruleset that each agent operates by, such that the resulting behavior adheres to the emergent one seen in nature? For example, could you evolve the flocking rules for how each bird flies, then run a whole flock of these agents and get flocking behavior?

Using evolutionary computation along with program synthesis—a process of generating novel computer code according to certain specifications—the researchers Rory Greig and Jordi Arranz succeeded. And they even did it for the flocking example!

Figure from “Generating Agent Based Models From Scratch With Genetic Programming”

As per the paper:

This technique enables us to generate models which are relatively free from existing domain priors and human preconceptions, and may shed light on completely new dynamics which have been overlooked because they are unintuitive or non-obvious. The output is an interpretable symbolic model which can be understood and extended by a human modeller, so this could be used for automatically building quick prototype models before a modeller refines them.

Fascinating work and I’m excited to see more in this space.

I also delivered a keynote at an ALIFE 2021 special session on artificial perception on “How to Think About Failure,” which was a lot of fun.

Share


The Last Policeman is the first book in a trilogy of novels revolving around a police detective who insists on doing his job of solving mysteries, even though there’s a complication: there are only a few months to go before an asteroid is predicted to collide with Earth and destroy all human civilization. While definitely imbued with an outer-space premise, the books of the trilogy are not really science fiction; they’re very much novels about simply living in the shadow of this impending doom, even as society crumbles around you. As per Wikipedia:

Winters describes the work as an “existential detective novel”, turning on the question of why people do things in spite of their long-term unimportance.

And as depressing as this sounds, it’s also fantastic (or perhaps that’s just the pandemic talking). Because these are the kinds of books that help the reader to crystallize what truly matters: what should each of us care about, if we can pull back the veneer of normalcy? The successive books of the trilogy explore the changes in society as the asteroid gets closer and closer, as well as how the main character continues to just solve different cases.

With that, I will leave the reader with this excerpt from the Talmud wherein a debate is recorded about whether humanity should have even been created:

For two and a half years, Beit Shammai and Beit Hillel disagreed. These say: It would have been preferable had man not been created than to have been created. And those said: It is preferable for man to have been created than had he not been created. Ultimately, they were counted and concluded: It would have been preferable had man not been created than to have been created. However, now that he has been created, he should examine his actions that he has performed and seek to correct them.


Finally, here a few things I’ve recently enjoyed:

  • Why I’m a proud solutionist: My friend Jason Crawford’s take on optimism and technological progress.

  • How To Terraform Venus (Quickly) [video]

  • A Deep Dive into Space Elevators

  • On Geofoam

  • “I wanted a working LEGO space computer but couldn’t find anybody making them, so I 3D printed one (well actually 2)” This is unbelievable.

  • And “The Feel-Good Recliner That Cures What Ails You,” on the Adirondack chair:

    It wouldn’t be the first time that furniture designed for the sick or infirm crossed into the mainstream, says Patricia Kane, a curator of American Decorative Arts at Yale University Art Gallery, who obtained a Westport chair for the collection in 2002. A classic example is the wing chair, whose dual protuberances were likely designed to support the head of elderly or infirm occupants and which—like Bunnell’s patented chair—were sometimes fitted with commodes. “Nowadays they’ve migrated into our living rooms, and we think of them as living room furniture,” Kane says.

Until next month.

Share this post

☞ Emergence, Program Synthesis, and Evolution

arbesman.substack.com
Comments
TopNewCommunity

No posts

Ready for more?

© 2023 Samuel Arbesman
Privacy ∙ Terms ∙ Collection notice
Start WritingGet the app
Substack is the home for great writing