An Interface Neuroid, or IN, exists in the space between humans and computers, bridging the gap between pure biological neurons and pure digital neurons with a two-way information channel. INs enable a person to gradually digitize their mind.
Pure vs hybrid
At first, you may have to squint a little to see the neural connection.
What we currently think of as a neuron is either purely biological or purely digital and forms networks with other neurons of the same type.
Yet the Interface Neuroid spans the biological and the digital, forming an interface between neurons of these different types.
A neuroid is a not-quite-neuron, or in this case a neuron on steroids. Neurons send and receive signals to other neurons to form a neural network, whether biogical or digital. Since that's exactly what the Interface Neuroid does, it's neuron-like.
But it is also a higher-level ("bigger") construct than what we normally consider to be neurons. Furthermore, we have not yet embraced the type of neuron that acts as an interface between biological and digital neurons.
So it is a neuroid, or a not-yet-neuron. Assuming you're squinting a little.
Despite the Interface Neuroid (or IN) being a higher-level construct than what we normally think of as a neuron, it still accepts and emits the same type of fundamentally simple signals that neurons use.
Input as decisions
Like with most neurons, input is a stream of signals or "spikes" each separated by an interval of time. In this case, each signal represents the fact that a specific person made a specific type of decision.
The specific type of decision could be "I ate an apple", or "I wrote a book", or "I spoke at a conference", or anything else. It doesn't matter how small the decision was or how big and difficult the decision proved to be as long as it was made and it's good to keep doing on a regular basis.
Each Interface Neuroid maps directly to a single person and handles one type of decision by that person. This means that every distinct type of decision requires its own Interface Neuroid. It also means that each Interface Neuroid deals with pure signals that need no labels or other metadata attached. The label is attached to the IN, not the signals.
Output as prediction
Output, like with most neurons, is a periodic spike (or action potential) triggered by the pattern of input. With Interface Neuroids, the output is a probabilistic prediction of when the next input is likely to occur, but for now we'll treat the output as a definitive spike.
At its most basic, the output is generated using naïve time series forecasting. This bit of jargon means that if you decided to eat an apple for lunch every day for the past two weeks, there's a really good chance you'll eat an apple for lunch tomorrow too.
More specifically, if the average time between recent input signals is 24 hours, there's a good chance the next input spike will occur roughly 24 hours after the last input spike.
It's this straightforward because the input is all the same type of decision and originates from a single person. Forecasting would become immensely complicated if we mixed in bananas and sandwiches, or if we mixed in the apple-eating habits of other people. Forecasting a pure time series like this is surprisingly accurate because it fits how individual minds work and how we as individuals tend to go about our lives.
Still, there's no practical limit to the complexity of this predictive capability. As a modular component, it can expand from naïve forecasting into a complex neural network that accepts input from many Interface Neuroids and other sources to produce nuanced predictions. With a carefully designed API, it would be possible to plug in advanced cloud-hosted engines, perhaps for a fee.
Feedback loop via reward
So, where exactly does this output go? Back to the person who generated the input, back to the biological realm.
But the output signal is not typically delivered directly. Before delivery, the digital prediction is enriched by any available means to provide just enough psychological (or endocrinological) reward that the person repeats their decision and provides the next input signal. In simpler terms, the prediction is used to motivate the person to make the same type of decision again and provide it as input. This completes the feedback loop.
Like with the predictions, the capability to generate psychological reward can be of any degree of complexity and have a variety of designs. It too is modular and can expose an API to allow experimentation by ML enthusiasts.
When the output and input are connected this way, it becomes possible to assess the quality of each prediction by comparing it to the actual next input signal. In this way, both prediction and reward capabilities can improve over time in a reinforcement scheme.
Closing the digital loop
With an effective implementation, an Interface Neuroid generates a "pull" because it transforms past inputs into motivation for future inputs. Since an IN operates on a single type of decision, imagine it as pulling a thread from your mind and spinning it into digital gold.
The more INs a person has, the more threads that are pulled at the same time. Then, the more nuanced and accurate the digital predictions become, the more the collection of digital threads begins to resemble the originating mind.
The number of INs a person can have has no upper limit. There's no limit to how they can be connected together into networks, including with other people. No upper limit on the prediction and reward capabilities. And no prescription about how exactly to implement the pattern.