Artificial Intelligence, or something else?

Talk about Artificial Intelligence (AI) is everywhere these days. It appears to have weaseled its way into almost every aspect of our lives because AI is a catch-all term. In recent times there has always been a handful of researchers attempting to create true artificial intelligence but, in most cases, with what the public generally sees its simply a rapidly developed statistical model predicting the most likely response to a prompt. The true usefulness of these models comes from training them for specific purposes, acting as an extension of our abilities.

AI as a cognitive prosthesis

I was recently reading an interview of one of my favorite authors, Kim Stanely Robinson, who writes some fantastic sci-fi which naturally includes AI. They share the same thought that I have, that AI is a poor name for what these models really are. Peel off the shiny outer facades of anthropomorphizing, public relations, and user interfaces from these applications and what you get is a bunch of statistics and equations to represent relationships. Kim puts forward terms like ‘extremely rapid computation’ and ‘assisted data analysis’ to deemphasize the human qualities that these AI ultimately parrot, without truly enacting.

My personal favorite term brought up though was ‘cognitive prosthesis’. While prothesis typically represents an artificial element meant to replace a missing biological one, like a leg or finger. I like to think he meant prothesis as an expansion of our ability, like if we strap into an exoskeleton to enhance our bodies natural abilities. As a user of AI tools such as ChatGPT, YOLO , and SAM, this is at least how these AI tools feel to me. I could go through my gigabytes of lettuce photos identifying the plants, counting the pixels, and determining the canopy size myself, or I could have AI do those bulk processes much faster than I could possibly hope to ever achieve. I use these tools as a method to expand my abilities, not completely replacing my efforts.

Debunkbot

I want to bring up an excellent example of how AI applications can lend itself to this idea of cognitive prosthesis. Have you heard of the term gish gallop? It’s basically a debate technique that involves asking questions or bringing up points in such a rapid, carefree manner that the opponent cannot answer or disprove fast enough (Dr. Steven Novella explains this well). It is a popular tactic with pseudoscience and conspiracy theory proponents because it is hard to combat and remain an effect communicator. If only, there was some sort of cognitive prosthesis that could provide a way for people to express all their doubts and reasons for thinking a certain way and was able to keep up with genuine or accidental gish gallop attempts…

Debunkbot is just that. It’s a chatbot designed to listen to the user, understand their logic and ultimately refute falsehoods. And one thing it can do better than humans is recall all the facts and convincingly present them without getting flustered or frustrated with endless questions or misunderstandings. What would take a single or multiple subject matter experts’ minutes to explain can occur in seconds. If you want to hear the creators talk about this ‘cognitive prosthesis’ and how it works I highly recommend this podcast episode.

Conclusion

One of the reasons I like this term ‘cognitive prosthesis’ so much, is that like a real-life prosthesis you can’t just slap one on and magically work it like an extension of your body. Doing so results in clumsy actions that may be inefficient or even fail to achieve the desired results. It requires some level of understanding about the task, processes involved, and ultimately some practice to truly wield in a beneficial manner.

The same goes with AI tools, anyone can use them, but if you don’t understand what you are trying to do in the first place it’s much harder to properly use them. That’s why I think everyone should make a concentrated effort to understand the task and the tools limitations before attempting to use AI. Perhaps then, we might move past the term AI to something that more accurately describes what it is and how we interact with it.

Proudly written without large language models.

This work is licensed under CC BY 4.0