Systems and Understanding

Yesterday brought chats with many people, Zoom conversation with Gwen about British travels and retirement aspirations, and an abandoned Planet Fitness workout after discovering my padlock had wandered off. Evening beer and pizza with Steve included pondering how humans know and trust anything—thought experiments about dropping stone age people into modern government or average people into AI-saturated worlds, questions about abstraction tools we lack for understanding complex systems. We've always dealt with incomprehensible intelligence though: other humans. Those relationship tools work but aren't generally accepted for AI.

The Problem of Comprehension

October 9, 2025

The morning accumulated conversations with many different people, interspersed with processing a few more sets of Mika photos. These social connections and creative work wove together naturally, neither demanding exclusive attention, both benefiting from the rhythm of alternation.

Early afternoon brought a Zoom call with Gwen, discussing what I'd seen in Britain and exploring her retirement aspirations. The conversation revealed interesting overlap with what Mary is doing, so I introduced them to each other. Sometimes the best thing you can do is connect people whose interests align, creating possibilities for collaboration or friendship that wouldn't emerge without facilitation. I hope that conversation develops into something valuable for both of them.

Tried to work out at Planet Fitness but discovered upon arrival that my locker padlock had wandered off somewhere. Not wanting to leave my gear unattended, I opted for running errands instead. Sometimes small logistical failures redirect the day in unexpected directions.

Home brought a little Minecraft before walking to Suds for beer and pizza with Steve. Remembered to turn on Strava for the walk, at least—those steps and distances adding to the accumulated activity record. The conversation wandered into territory about how people can know and trust anything—fundamental epistemology over beer, the kind of discussion that starts casual and becomes genuinely interesting.

There's a base of experience required for understanding. We need abstraction mechanisms to comprehend complex systems. I find it helpful to force objectivity through thought experiments: pull someone from the stone age and drop them into a complex environment like the US government. They have no tools to wrap their heads around what they're experiencing. They don't know simple concepts like organizational hierarchies and their implications, for example. The entire framework for understanding institutional power would be unavailable to them.

Now take the average person today and plop them in a world saturated by AI technology. The abstraction tools needed to really understand the why's of AI responses don't exist, or at least aren't readily available. At best it requires mathematics generally beyond the concrete processing scope of the average human brain. We can use AI systems without understanding them, just as we can use government services without understanding bureaucratic structure. But genuine comprehension remains elusive.

Of course, on the flip side, we've been dealing with systems like this for as long as we've existed. They're called "other humans." Those tools for relating to incomprehensible intelligence seem effective, but they're not generally accepted as models for understanding AI. This opens a longer ponder about how humans tend to relate to other intelligent entities. Historically, not well. We have trouble extending our models for human intelligence to non-human intelligence, even when the relational tools might be more applicable than the analytical tools we think we should be using.

The question isn't whether AI is intelligent in the same way humans are. The question is whether our existing tools for relating to human intelligence might work better for understanding AI than trying to develop entirely new abstraction frameworks. We already know how to interact with minds we can't fully understand—we do it every day with other people. Maybe that's the useful model rather than trying to mathematically comprehend systems that exceed our processing capabilities.

Partly walked and partly rode the bus back home, the conversation still percolating. Watched the usual TV shows, chatted a bit more, let the ideas settle without forcing conclusions.

Today brings conversations with Cliffe, Scott, and Ashley, plus the need to prepare for weekend guests. The house requires attention before people arrive—that ongoing challenge of maintaining guest-ready domestic space. Should be an interesting day, though probably less philosophically demanding than yesterday's epistemological beer session.

A flower near Nana's from the other day