Bathing in a cognitive flow

From AI #19: Hofstadter, Sutskever, Leike by Zvi Mowshowitz.  

Rather than using humans as a metaphor to help us understand AI, often it is more useful to use AI as a metaphor for understanding humans. What data are you (or someone else) being trained on? What is determining the human feedback driving a person’s reinforcement learning? What principles are they using to evaluate results and update heuristics? In what ways are they ‘vibing’ in various contexts, how do they respond to prompt engineering? Which of our systems are configured the way they are due to our severe compute and data limitations, or the original evolutionary training conditions?

I like this because it ties to two of my refrains - Negative space and cognitive pollution.  

One of our hardest tasks when considering an issue is to see the negative space.  We tend to see the duck, not the rabbit (or vice versa).  The old crone, not the young girl (or vice versa.)  This is not a matter of bias or malevolence, its just how we are programmed.  We can become better disciplined about looking for the negative space but I don't know that it ever becomes second nature.

  

Right now there is a lot of panicked conversation about AI.  I like the reversal of the metaphor though and think about AI to better understand humans.  It is possible that AI might end up being an exogenous shock to the human evolutionary process but I suspect not.  It will be a shock of greater or lesser impact to which we adapt more or less easily but I do not know of anything I have seen argued yet that creates an extinction level event.

The one caveat is that technology evolve differently and faster than biology.  That is neither good nor bad.  It just is and we don't have a lot of case studies with well defined edges.  

I like Mowshowitz's observation as well as it prompts another metaphor which ties to another focus of mine.  Forget being trained on data sets.

AI and people are both shaped by flows of information which we can call being trained on if we want.  But we are all and always passing through a flow of data anchored in time.  A flow of experience, a flow of incomprehensible data, of information, of knowledge, a flow of cognition.  We are all washed by different rivers of experience, knowledge, cognition.

As Heraclitus observed, no man can ever step in the same river twice.  We are all necessarily unique not just because we are unique biologically but because we all experience different flows.  We don't fully comprehend the degree to which we are shaped by the river flows of experience, knowledge, cognition, nor do we understand how those flows determine future values, goals and behaviors.  And all this is analogous to AI and Large Language Models but our comprehension of AI and LLMs is a tiny slice of our greater experience with, but steady incomprehension of experience, knowledge, cognition flows on individuals.  

For ourselves as people, I have for years been talking about the challenges of cognitive pollution.  Keeping with the flowing river metaphor, we are shaped differently whether we are bathed in a refreshing and clean mountain stream or whether we are bathed in a slow moving river of cognitive pollution, filled with error, untested assumption, presumptuous mind-reading, false beliefs, and malevolent contrariness.  

We will always make mistakes but we can make a difference in how clean is the cognitive raiver in which we bathe.