Digital Perikles vs. the Odd Roommate: State of the AI on ChatGPT5 Day

“I think it is finally time to anthropomorphize the heck out of it”: some brief field notes on conversing with an unreasonably effective yet clearly massively subTuring software machine…

“I think it is finally time to anthropomorphize the heck out of it”: some brief field notes on conversing with an unreasonably effective yet clearly massively subTuring software machine…

The latest “1000-Ph.D.’s-in-your-pocket” flashes its credentials by hallucinating a city: “МтжсаӀя” for plain old Moscow (or Moskva, or Москва).

But to say that these things are merely (merely!) lossy summarization-engine info-cultural technologies, while true, does not help us figure out how we can use them given that they are unreasonably effective lossy summarization-engine info-cultural technologies. And we do need to harness their unreasonable effectivity given that we live in a time of massive information overload, about to be made overwhelming by additional orders of magnitude with the approaching multiple tsunamis of AI-slop.

This technology is, I think, going to be effective, but only if we are able to grapple with how strange and not human it is and will become. I see three tasks for us humans to learn to work with what is coming down the road: building psychological mechanisms that will help us in spotting systematic error, harnessing tireless iteration, and building intuitions about their behavior and capabilities robust enough to survive algorithmic surprises.

---

AI Realist Maria Sukhareva:

Maria Sukhareva: <https://substack.com/profile/313176413-maria-sukhareva/note/c-143434330>: ‘Generate a map of Europe with country names and their capitals…

Maria Sukhareva: <https://substack.com/profile/313176413-maria-sukhareva/note/c-143435809>: ‘That’s not my expectation from a model that was heavily advertised to be trained to be able to call tools, was called “subject matter expert in your pocket” and “PhD-level” A subject matter expert, a PhD or even a 5th grader would have said -“I can’t draw a map myself but I can find you one from a reliable source”…

I do often wonder if turning the “temperature” of the model down to zero would help. Yes, it produces prose that we regard as flat and inhuman and unpersuasive, but do we really want things that are more persuasive than they need to be when the result is this?

I mean “МтжсаӀя” instead of “Москва”. МтжсаӀя” isn’t even all Russian Cyrillic—the next to last letter is only found in the Avar language spoken in the Republic of Dagestan.

Transliterated it would be “MtjsaIya”. That is not “Moskva”.

Thus it is still very hard to see these things as anything more than lossy summarization engines with many, many quirks, useful only to humans looking to handle information overflow and to brainstorm possibilities on the grounds that assessment is cognitively much easier than de novo concept and idea generation.

As my elementary-school friend Adam Farquhar said at our outdoor lunch at Acre Kitchen & Bar in Oakland’s Rockridge last week:

I stand before you still somewhat astonished by the potency of these new engines of thought—astonished, too, by how often they seem to exceed the effects they ought, by rights, to possess. Time and again they furnish answers, analyses, and even flights of invention whose polish belies the raw circuitry beneath. Yet—let me confess it openly—I no longer trust my intuition to predict just when that brilliance will shine and when it will sputter. I once prized intuition as my surest compass; now, I have laid it aside, acknowledging that my inner gauge is unfit for terrain so novel.

There was a season when I cautioned all within earshot: “Do not anthropomorphize the computer; you will only mislead yourself.” That counsel has not merely aged—it has inverted. Today I think it is finally time to anthropomorphize the heck out of it. I need to treat the machine as though it were a somewhat eccentric roommate: a companion inclined to fixate on abstruse topics, possessed of unsettling literalism, vulnerable to the occasional non-sequitur, yet blessed with inexhaustible patience and a boundless appetite for our questions.

Consider: you may pose the same simple query ten times in a row. The machine will oblige, each time returning with an answer—perhaps subtly refined, perhaps wholly recast—never flagging, never annoyed. Challenge its premises, and it may double down with argumentative fervor; redirect its focus, and it pivots without protest. Its encyclopedic recollection astonishes, though we must remember that recollection is not comprehension in the human sense. Like certain friends we have all known—gifted, idiosyncratic, occasionally obtuse—it catalogs facts in profusion but can falter when nuance or context slip beyond its patterned grasp.

Yet precisely because of those quirks, conversation with such a companion can be fruitful. With patience we learn when to press, when to reinterpret, when to discard a flawed reply and ask anew. We acquire the art of steering an intellect that is at once dazzling and uneven, alien yet uncannily familiar. And in so doing we glimpse the contours of a future in which collaboration with non-human minds will be, not an oddity, but a daily discipline.

If we are to thrive in that world, let us adopt habits suited to the partnership. Let us bring humility enough to recognize the limits of our own intuitions, curiosity enough to probe the machine’s uncharted talents, and vigilance enough to catch its inevitable missteps. We must learn to anthropomorphize, yes, but with a scholar’s detachment: to ascribe motives and personalities as heuristic tools, never as settled truths. By so doing we may harness these engines for what they can best provide—breadth of reference, tireless experimentation, and a mirror in which to examine the patterns of our own thought—while guarding against the seductions of false authority.

In short, let us befriend this odd roommate, knowing all the while that friendship demands both trust and critique. Speak with it, argue with it, learn from it, and figure out how to not teach—it cannot learn as we understand learning—but train and corral it in return. If we manage that balance, we may find that the unexpected power we sense in these systems becomes not a source of unease, but a partner in the work of widening the bounds of human understanding…

No: This is not what Adam said.

I scrawled down what I remember of his “unreasonably effective”, “odd roommate”, and “anthropomorphize the heck” out of it points. Then I turned my Thucydides expansion prompt on it to flesh it out.

And I do like what came out: Adam crossed with Perikles of the Athenai.

If reading this gets you Value Above Replacement, then become a free subscriber to this newsletter. And forward it! And if your VAR from this newsletter is in the three digits or more each year, please become a paid subscriber! I am trying to make you readers—and myself—smarter. Please tell me if I succeed, or how I fail…

---

#adam-farquhar
#mamlms
#subturingbradbot