 We humans spend a third of our lifetime of sleep. Not sleeping enough affects our ability to work, to learn, or even to drive safely. Scientists discover two distinct types of sleep. Rapid eye movement, during which we have vivid dreams, and non-rem sleep, which is a deep slumber. We now know that both rem and non-rem sleep are controlled by dedicated, precisely wired circuits in the brain, and when these circuits are damaged, we develop all kinds of sleep-related problems. So the goal of my research team is to figure out the sleep circuits so that we can improve and perhaps even control sleep. The good news is that we already have a rough idea of where to look. But the bad news is that these brain regions are not just involved in sleep control. The brain is more like a bowl of spaghetti, very difficult to entangle. Fortunately, the neurons serving different functions often turn on different genes and produce different proteins in order to do their jobs. In other words, they contain distinct cell markers. And once we identify the sleep neurons, we can label them with distinct colors, so we can actually introduce light-sensitive proteins so that we can use light to either activate them or shut them down. So if we can use light to control spiking of the sleep neurons, we can control whether the animal goes into sleep or wakes up. Our current work is mostly done in the mouse. What we found is that when we turn on light to activate these neurons, we can reliably convert non-rum sleep to rum sleep within a few seconds after we flip the light switch. And once we identify some of the sleep neurons, we also have the tools to map out their inputs and outputs of this very complex network. In other words, we can use the identified sleep neurons as a starting point to trace out this big network. This would allow us to figure out what goes wrong when someone develops a sleep disorder so we can come up with effective strategies to repair the sleep circuit. But when safer techniques are developed for humans, we could potentially turn sleep on and off at will. So my question for the audience is, if that becomes possible, do we want sleep on demand?