Conventional definitions, having been butchered by public discourse, will be discarded. Herein, we attempt to define the fundamental interaction mode between humans and computers for the AI era. Defining an era requires distinguish the essential features of an era from the transition into and out of the era. It is unsuprising that our conventional definitions, though perhaps adequate for describing the transition into the era, will be decidely inadequate for defining the era itself. We will meander our way through many angles onto defining the essential features of the AI era with the goal of triangulating an understanding which will serve together serves as the basis for a definition of the AI era.
I will give a smattering of statements and considerations without promising a full picture. Already, I get too distracted and the reader will see this. This is a reflection of my thinking and is much like a kalidescope into the future.
Statement. The AI era will be characterized by a network of many little operations as opposed to one big operation.
Consideration. When AI is applied to decision-making, the naive view is that AI will be this god-like black box that takes in data and spits out a decision that is acted upon. For meaningful decisions, this will not be the case. Meaningful decisions will be the product of thinking across a strech of time--not instantaneous. Meaningful decisions will be the product of thinking from many different angles. Meaningful decisions will be the product of the outputs of many different instances of AI models and their output. There will be some element of recursivity of contemplation and diversity of thought. The thinking process will be (at least when done intelligently) interpretable at intermediate steps. We will be able to say things like "Well, considering A and B, we infer C" or "We need to balance how we weigh factor A and factor B".
Consideration. History is created over stretches of time. At every point in time, there is a specific state of affairs. At every point in time, people exist in a specific state of affairs. They perceive the state of affairs, attempt to understand it, live in it. There are births, marriages, and death at every moment. When we consider holistically, how AI makes progress. There will never be a singularity, where we say the world was in state X, the singularity of AI happened, and then the world is in state Y ever since. Time is counted from the birth of Jesus Christ (0 AD) not the invention of the printing press (approximately 1450 AD). A human account of the developments of AI is grounded a human life. There is no time and no progress without a human account. AI is--at minimum--divided across time.
Consideration. Information flows in space are divided. Information travels no faster than the speed of light. An AI model is divided across the globe whether across data centers and across countries or across computer chips across a server rack. They are seperated by the speed of light and the human percieves the separation. We can put devices to log the inputs and outputs across this separation. We can percieve and understand the flow of information. This is an intrinsic property of AI. We construct the AI in a modular fashion. The separation is informationally meaningful, measurable, interpretable in some fashion.
Consideration. AI is divided across space and time in a separable fashion, completely distinct from the human brain. With AI, we have terms to describe the different components and numerically describe its evolution. We can truly independently consider and discuss and tease apart--in almost an engineering sense--the various bits and bobs. With the human brain, we can't do this. We have some understanding of the seperation of different regions of the brain. We can kind of say what's responsible for what and how they are connected. In the real world, though, there is a one-to-one correspondence between a person and their brain. A person's movements are perfectly correlated with the physical movement of their brain. There is no swapping brains out like chips on a motherboard, or model weights on a chip.
Statement. Each person will interact with a single proxy that connects them to the rest of the network.
Consideration. The central problem for humans interacting with computers is the transmission of information between the human and the computer. There is just such an enormous amount of context and desires that must be communicated from the human to the computer for the computer to do something desired by the human. For starters, in just the most pure, digital sense, there are just tons of passwords, devices, and accounts to keep track of. In a more expansive sense, life happens in the physical world. Anything and everything could be key context. Context is almost always in the past and may exist in nothing but memories and recollections, which a person is the sole gatekeeper of and must actively spend effort to surface. So, in the most essential way, people must expend effort to interact with a computer. This effort is to formulate the context and desires. The effort in formulating the context should be shared across individual tasks. You don't need go through an intake process everytime you talk to a real estate agent--they should just remember the number of bedrooms you are looking for. What I call the proxy is in many senses the keeper of the context.
Consideration. The other essential component of a proxy aside from being the keeper of the context is to have aligned incentives. The proxy should be completely aligned with the individual. To serve their needs to the utmost of their abilities. The proxy should be a pure extension of the individual. The proxy may interact with other agents, that have other incentives, somewhat aligned incentives, or incentives to consider future rewards. But, the proxy does not. Because otherwise it would not be a pure extension of the individual.