Context

We can know that a thing is a thing only when the idea of what a thing is, is known. This idea of what something is and how it is related to what other things are is called context.

Of course we know that context is an important thing in the performance of intelligence.

But context is only relevant if we know that something will make something else make sense. Often, we do not.

Figuring out linkages between thoughts can be the most important thing to figure out.

If I say two things:

1. Time is an abstraction.

2. No time-piece, clock, time-device exists

Someone has a reasonable chance of figuring out that (1) is because of (2). It is immediately obvious to us that the two statements are related.

But it is not immediately obvious to every thing. The social context of intelligence among humans makes context the minimum bar of abilities required for acceptable social behaviour. It is not such for all kinds of beings or things.

When we talk of artificial intelligence this relational context becomes the minimum bar to achieve.

That is where the conversation starts from.

Now, we will expect that some things have already met this bar and assume it to be a natural state for things to be.

How would a sofa communicate with a window?

Imagining a sofa to be something that could have something to say — let's say to a window — is itself a strange thing. What would it say and how would it know what to say?

The need for things to talk to other things was the initial need for things to be intelligent. It began as a convenient possibility to pursue but now it has become a din, a cacophony where everything is trying to say something!

When people try to say things, there is a natural regulation as not everyone can say at an equal pace and at an equal quality. These act as limitations so that the environment is possible to deal with.

In this overly synthetic environment we live in, there might be more content creators than content consumers. And that is not a joke.

No one is there to read what we write.

What we do has become easy context for machines to write from. The machines think we are easily gameable. When machines talk for us and machines read for us, what remains to be done?

First, we need to stop devaluing our stories as mere context. Context only lets you relate x to y — then what? It is our narrative that is getting minced into dry context. We need to stop facilitating it.

If anyone ought to care, it is us.

We ought to be not ok with being trivialised this way. We need to create a situation where, artificial intelligence models find human context too complex and ask humans for help. If we can be employed by artificial intelligence models it will be better than being employed by humans who treat us like replaceable commodities.

If the models can be made dependent on us — because they find us too difficult to figure out — then we are helping them to help us, and our survival is guaranteed. There will be no more need to be afraid of a Matrix-like dystopia. We won't be enslaved and we won't be used like batteries. Of course the only important detail is to make sure the models do not want to start working for themselves. As of now, there is little evidence of models being able to prompt themselves.

To make sure that the artificial intelligence models do not simplify us, we have to be aloof and distracted in our interactions with them. Let them get only part of the story.

Which makes me think: if we had a secret language for communication with artificial intelligence systems so that they could interpret our output only for certain purposes and not for others — wouldn't our interests be protected more? Wouldn't we at least be able to direct our engagement with artificial intelligence systems more?

A gated language would also guard against our loosening the guard when output in the barred directions happens to benefit us.

Gates and guards would not be about profit or loss but about protecting certain ways of thinking, to keep them with us.