e Connectivity: Human neurons typically support several hundred connections

e Connectivity: Human neurons typically support several hundred connections (synapses). Moreover, the complex pattern of these connections is very meaningful. (See our next point.) Computer units typically make only a handful of connections, in regular, fixed patterns. e Development (self-assembly with interactive sculpting): The human brain grows its units by cell divisions and orchestrates them into coherent structures by movement and folding. It also proliferates an abundance of connections among the cells. An important part of its sculpting occurs through active processes during infancy and childhood, as the individual interacts with his or her environment. In this process, many connections are winnowed away, while others are strengthened, depending on their effectiveness in use. Thus, the fine structure of the brain is tuned through interaction with the external world—a rich source of information and feedback! e /ntegration (sensors and actuators): The human brain comes equipped with a variety of sensory organs, notably including its outgrowth eyes, and with versatile actuators, including hands that build, legs that walk, and mouths that speak. Those sensors and actuators are seamlessy integrated into the brain’s information- processing centers, having been honed over millions of years of natural selection. We interpret their raw signals and control their large-scale actions with minimal conscious attention. The flip side is that we don’t know how we do it, and the implementation is opaque. It’s proving surprisingly difficult to reach human standards on these “routine” input-output functions. These advantages of human brains over currently engineered artifacts are profound. Human brains supply an inspiring existence proof, showing us several ways we can get more out of matter. When, if ever, will our engineering catch up? I don’t know for sure, but let me offer some informed opinions. The challenges of three-dimensionality and, to a lesser extent, self-repair don’t look overwhelming. They present some tough engineering problems, but many incremental improvements are easy to imagine, and there are clear paths forward. And while the powers of human eyes, hands, and other sensory organs and actuators are wonderfully effective, their abilities are far from exhausting any physical limits. Optical systems can take pictures with higher resolution in space, time, and color, and in more regions of the electromagnetic spectrum; robots can move faster and be stronger; and so forth. In these domains, the components necessary for superhuman performance, along many axes, are already available. The bottleneck is getting information into and out of them, rapidly, in the language of the information-processing units. And this brings us to the remaining, and I think most profound, advantages of brains over artificial devices, which stem from their connectivity and interactive development. Those two advantages are synergistic, since it is interactive development that sculpts the massively wired but sprawling structure of the infant brain, enabled by exponential growth of neurons and synapses, to get tuned into the extraordinary instrument it becomes. Computer scientists are beginning to discover the power of the brain’s architecture: Neural nets, whose basic design, as their name suggests, was directly inspired by the brain’s, have scored some spectacular successes in game playing and pattern recognition, as noted. But present-day engineering has nothing comparable—in 60 HOUSE_OVERSIGHT_016280