On extended cognition

Neural activity is a limitless stream of potentiality distributed and instantiated in every aspect of our existence.

The different instantiations we come to define and interact with in order to outsource a part of our cognitive functions, such as symbolic representations (writing) or memory storage (books) work as a practical extension of our individual and collective mind.

Once we conceive something, we can focus our energy in creating something else with a different level of complexity.

This process reflects the problem-solving nature of our species but also the natural tendencies to outsource our cognitive load in order to increase efficiency.

Modern technology is clearly a result of this tendency and, as clunky as sometimes the output of LLMs might be, errors will be eventually corrected and it will unleash the automation of intelligent operations at scale.

What is curious here is that the extension of cognition will go from an artificial entity to another. That is, from data-gathering robots, synthetic data and simulation environments to LLMs and other types of AI systems.

There is another Interpretation locus coming from abstraction, or from the platonic space, as Michael Levin would say.

Even if it is through sheer compute power, we’ll be dealing with hyper efficient artificial agents who will accelerate the way we construct the cultural arenas in which we live.

A clear illustration of this is visible if you take into account the role of the internet and automation in the shaping of global and local cultures. Add conversational/agentic interfaces to that layer of reality and see where it takes us in a few years.

Nobody knows for sure how it will unfold, but one thing is clear: this stage of cognitive extension takes us back to the old issues of meaning, interpretation and interpretability.

At least from an ontological perspective, meaning, as I discussed in one of my mini-courses, is the result of a dynamic mapping of interpretations. This mapping is how we draw the lines of the territory and make it understandable to ourselves and others.

In this sense, meaning requires a voluntary extraction and reconstruction. This is what is done in human translation, by the way.

Machines and artificial entities coming from an abstract/platonic space in the near future are yet another gateway for hermeneutic ramifications, but it should not undermine our interpretative properties, which will be more and more needed as we continue to converge with a digital space more and more run by automations and pseudo-realities generated at the click of a button.

Is AI the ultimate cognitive offload? We don’t know, but it is definitely not the end of our interpretative responsibility. AI is, however, a potential mess for interpretability issues.

Artificially intelligent systems are designed for speed and efficiency and they have access to all the digital information that is available. However, they are not equipped with a natural capacity for human-like calibration of cognitive exchanges.

In other words, they don’t have access, ontologically speaking, to a contextual decoding of the whens, the hows and the whys found in the human experience.

They will, perhaps, have access to an infinity of human-like experience models ready to deploy, but the point is that, as a species, or even, as living entities, we should not neglect our personal and collective ways of gathering, refining and sharing meaning.

What kind of tools are we going to need in the future? Who and why is going to build it? What type of knowledge are we going to pursue? How are we going to agree on that knowledge as an intelligent species?

If we keep our feet on the ground, these questions don’t seem necessarily urgent immediately. However, it doesn’t hurt to start thinking about them.

I’ll be back soon with another newsletter/blog post to connect some dots between this text and my previous post because I think it’s necessary to lay out another idea.

Writing this helped me a lot to clarify my thoughts on extended cognition. I hope you liked it as well.

À bientôt,

Javier