*An exploratory blog-post draft. A guess at where our interactions with technology might be heading, rather than a prediction of exactly how or when we will get there.* I think we may eventually stop thinking about using technology as operating a collection of systems. Instead of learning which buttons to press and which application to open, we will communicate what we want. Sometimes we will not even need to do that. A system that knows us well enough might recognize what we are trying to accomplish and take care of it. That sounds like a dramatic change, but the most useful examples might be ordinary things. ## Running out of toilet paper Right now, ordering toilet paper involves a surprisingly long chain of interactions. First I notice that I am running low. Then I pick up my phone or open my laptop, find a service, search for the product, compare options, click buttons, enter whatever information the service needs, and place the order. My goal was never to operate a shopping application. My goal was to have toilet paper when I needed it. A future system might know how much I ordered last time, how long that quantity usually lasts, and how much time I have actually spent at home since then. It could estimate when I am likely to need more and arrange for it to arrive. The first few times, it might ask whether its estimate is right and whether I want it to place the order. Once we establish a pattern, I might give it permission to keep doing that within limits I choose. If the cost is inconsequential to me and it consistently gets the decision right, why would I want to approve the same purchase over and over? The point would not be to make shopping faster. It would be to remove an interaction that did not need to happen in the first place. ## The knowledge we currently need just to use our tools We have become so accustomed to the way computers work that we rarely notice how much we have to learn before they become useful. I need to know how to use the device. Then I need to know how to use the particular application. I need to know what that application can do, where its features are, and which sequence of actions will produce the result I want. Its developer also needs to have anticipated what someone like me might want to accomplish and provided a way to do it. There is a whole library of knowledge involved in navigating our devices and services. Because we accumulate it gradually, it starts to feel natural. But knowing where a setting lives is not the same thing as knowing what we want that setting to accomplish. Keyboards, mice, and touchscreens are not necessarily limiting what we can achieve. They are limiting how seamless the process can be. We keep translating our intentions into the language of each individual tool. I do not think all of that translation will remain necessary. ## Tell the system once Another part of this change would be continuity. Today, I give the same information to different devices and services repeatedly. My address. My payment details. My preferences. We often accept this as obvious: I entered that information on my laptop, so naturally I need to enter it again on my phone. But why should the device define the boundary of what the system knows? I imagine a system where I provide information once and it is available wherever I authorize its use. The phone, laptop, television, and other devices would be different ways of interacting with the same underlying system, rather than separate islands that each need to be taught about me. That does not have to mean every service receives everything. The useful idea is that I should not have to keep supplying the same information merely because I am using a different screen. ## Preferences that become more useful over time The most valuable information might not be my address or credit card. It might be an increasingly useful understanding of my preferences. How do I like to be spoken to? How do I like my food prepared? What kinds of movies do I enjoy, and what specifically do I dislike about the ones I do not? I might watch a movie and give some feedback. The system could use that feedback next time. If I have a bad experience with a product, it could remember what was wrong with it, rather than merely recording that I disliked that particular brand. It could avoid other products with the same qualities. Preferences could also follow me into new environments. If I like to sleep late and arrive at a hotel with automated blinds, the system could communicate that preference, with my permission, so the room does not open the blinds early in the morning. Instead of starting from scratch with every service, each interaction could contribute to a more useful understanding of me. ## Goals instead of a collection of tools With less sophisticated technology, the basic pattern is that I have an intention, identify the tools that might help, and operate those tools individually. What more sophisticated computing might do is abstract away much of that tooling. I describe the goal, and the system works out which tools are necessary. It might even create a tool or a temporary control surface for the particular task. That does not mean screens or controls disappear. It means they stop being the starting point. Suppose I want to create a model of something. Naturally, I will probably want to see the model while it is being created. A screen is useful for that. But I would not necessarily need to sit at a computer, download a program, learn its interface, and navigate its menus before getting started. I could describe what I want, and the model could appear on a nearby screen. If rotating it is easier by touch, I might use my phone to turn it around. If I want to change a dimension, I might say so. The television is an output. The phone is an input. The actual model and computation might be somewhere else entirely. These devices are participating in one interaction, not each running a separate version of my task. ## Interfaces chosen for the task Voice seems likely to be a major part of this because speaking is such a natural way for humans to express intentions. We acquire speech through ordinary human interaction, while reading, writing, typing, and navigating software require more deliberate instruction. But I do not think voice needs to become the only interface. Sometimes looking is better. Sometimes touching is better. Sometimes a keyboard is exactly what we want. The change is that the interface could be chosen according to the task and the person. Buttons, keyboards, mice, and visual controls would be available where they help, rather than required because that is how a particular application was built. I might begin by speaking, inspect the result on a large screen, adjust it with my hand, and continue the conversation from another device. I would not need to think about switching applications or transferring the task between machines. ## Sometimes the best interaction is no interaction There are also things I may not want to operate at all. If I do not want to think about food, perhaps I say, “I’m hungry,” and the system chooses something I tend to enjoy and orders it. If I prefer even less involvement, it might learn my routine and suggest or arrange food when I am likely to need it, within permissions and a budget I have already established. That will not suit everyone, or every decision. Sometimes choosing the meal is part of the pleasure. Sometimes the cost or consequence makes confirmation important. The level of involvement should itself be a preference, not something the system decides for everyone. Still, it is worth asking how many of our current interactions exist because we genuinely want to make a decision, and how many exist because the tools cannot proceed without us operating them. ## A different meaning of using technology My guess is that the long-term change will be less about better applications and more about no longer needing to think in terms of applications for many ordinary tasks. We will still have goals. We will still want to inspect, create, change our minds, and take control. But we may spend much less time learning the machinery that sits between an intention and its fulfillment. The difficult part will be making the system good at distinguishing a guess from a decision it has permission to make. It will need to learn without becoming presumptuous, remember without making every detail available to everyone, and remain easy to correct. Those are conditions for the idea to be useful, not reasons the current arrangement must be permanent. Perhaps using technology will eventually feel less like operating a machine and more like expressing what we want, with the machinery quietly arranging itself around that intention.