The Runaway Species. David Eagleman

a lifeless pile of gravel.

      Chitra Ganesh and Simone Leigh’s My dreams, my works must wait till after hell

      At first blush, this combination of the living and non-living may only seem useful to generate art projects, but it also forms a solution to the problem of the world’s cracking buildings and roads. Half of the world’s constructions – everything from roads and bridges to high-rises – are made of concrete, a building material that is notoriously vulnerable to the elements and is difficult to mend. To fix this problem, chemists turned to the natural world. They added a particular strain of bacteria to the concrete, along with the bacteria’s favorite food. As long as the concrete is intact, the bacteria lie dormant. But if it cracks, the bacteria become active. Consuming the meal that has been waiting for them, they spawn and spread, excreting calcite that seals up the damage. Thanks to this unique blend of microorganisms and building materials, the concrete heals itself.⁴

      In a similar vein, our neural networks are skilled at mixing our digital world with our analog one. Computers may surpass us in processing power, but some skills that are trivial for humans have proved arduous for our silicon avatars. One of these has traditionally been image recognition. Picking out a face is easy for a child, but it’s difficult for a computer.

      Why? To a computer, a digital photograph is nothing more than a collection of pixels of different colors and intensities. The computer needs to learn higher-order patterns in order to identify the content of the photo, and it takes millions of examples to get there. This problem came to the fore in the early 2000s, as people around the world began to upload billions of images to the web. Google wanted a way to automatically label them, but, try as they might, they couldn’t come up with algorithms that worked.