Reflections on an A.I. Mirror

The Sorcerer's Apprentice…

I find it absolutely fascinating that flows of humanity can be described by mathematics of fluid dynamics. Not that I know anything about fluid dynamics, I don’t. But then again, Squirrel is often distracted and skitters around looking for nibbles.

In a crowd of people ambling along a sidewalk, or in cars bumper-to-bumper on a turnpike, our movements can be described by the same math that describes the movement of water molecules in a pipe. Friction, boundary layers, viscosity, all can be quantified and related. None of it is impacted by intention.

What does this say about our free will? Do we have free will if our next act can be predicted by formula? Do water or oil molecules have more intention than we give credit for? We all know that oil and water don’t mix … Squirrel, again. Sorry.

But the question stands. If the action of a water molecule and a driver can be described by similar mathematics, in advance, do we have free will?

And what about those times when we know we’re being manipulated? Screens in my life are addictive. Maybe because I have an addictive personality, a bad case of ADHD, a history of risky behavior, probable dopamine deficiency, a plethora of bad habits, a notebook full of worse choices — you know the list.

At least I no longer suffer from feelings of terminal uniqueness. Fellow travelers know what I mean about that, too. But that’s the point, here.

Musings on fluid dynamics may raise only trivial questions, on the surface (tension?). But, how much of our “willed” behavior is governed by mathematical principles? Can we break free? Or would that just be finding another place on a probability curve, where we’d sit as an unstable statistical outlier, occupying a place predicted by a percentage?

For squirrels who feel at times like we’re not really making the decisions, where do we fall into social, or corporatist programming? Social media and Alternative Intelligence are prime candidates, of course, but there’s also phone calls answered by AI, forms fillable only online, in-car cameras reporting directly to our boss or insurance companies, and spies in our living room (looking at you, Alexa!)… oops. Sorry.

Squirrel spent too much time today looking at feedback loops between people and A.I. This was sparked by OpenAI’s “Code Red” directive from founder Sam Altman to his company: To maintain its competitive edge against Google’s “Gemini,” Anthropic’s “Claude” and others, OpenAI is going to focus more on ChatGPT’s interactions with customers.

Specifically, ChatGPT will depend more on “one click” affirmation of interest, more on metrics describing how “sticky” a response, tone of voice, or possibly a type of answer may be for a viewer, as opposed to the “best” possible response.

Sticky, as in, how long will the eyeballs stick around. And we all know, eyeballs are money-balls. As if I asked about hormonal triggers of hunger response, and the results included flavors of ice cream.

If, as ChatGPT responses drift toward “give them what they want” which have a different objective for Altman’s ChatGPT than they do for those asking questions, will this alter our predictable behavior?

If our unconscious desires factor more directly into ChatGPT responses, which in turn reflect our own thoughts, which again impact what we ask, which ChatGPT constantly monitors and uses to modify its responses to keep eyeballs around for its own objectives… etc.

That’s not just Squirrel, though probably reads like it.

I wanted to know if this was a “recursive loop” or an “iterative loop.” Recursive loops refer to themselves (“until you find glasses look for glasses”), iterative loops do things in order (look for glasses in bathroom, look for glasses in bedroom, look for glasses in kitchen) over and over until something changes (found glasses).

The difference may be trivial, but recursive loops can lead to chaos. They also have self similarity in scale, but that’s another Squirrel move and will be left until later.

The conclusion was that the loop was probably iterative, though I’m not so sure. In any case, without a terminating condition, the loop could theoretically spiral toward infinity. Now, outside conditions (popcorn, Christmas, “Black Sails”) can certainly interrupt this process for me, personally.

But for society, which is a larger organism than me with similar but different dynamics, possibly predictable by fluid dynamics or some other formulas tucked away in labs (at Google, OpenAI, Meta, Microsoft and Amazon), this feedback loop might lead to a deadening of exploration. As if we were on a turnpike, seemingly each of us alone in our car, but moving along in traffic flows that are predictable by clever, impersonal, and irresistible mathematical rules.

Because, what ChatGPT intends to provide will be more of a mirror of what we already are than what we could be. Back and forth, between its mind and ours, until something disappears.

As always the case with self-absorption, images between parallel mirrors get smaller with each iteration.