Apple Intelligence: These Bots Can't Reason and "Possibly" Never Will: Part 1

"The future is dark, with a darkness as much of the womb as the grave."-Rebecca Solnit

"The future is so bright, I gotta wear shades." -Timbuk 3

Gentle Readers,

Long time, no newsletter. I had a planned short leave that "unexpectedly" turned into a much longer one. While I was gone, two things happened.

One, I was (and still am) experiencing a bizarre health event. While always a possibility, theoretically, not predictable, not at all probable, and as far as we know entirely unique.

Two, a once intimate friend went to bed on Friday the 13th of September and did not wake up on Saturday the 14th.

Everyone gets sick, and people die, I don't mean to be melodramatic. However, either through brute force or design, these events are proving convenient ways to talk about larger more challenging ideas that are looming this historic week.

The title refers to the Apple Intelligence demo--and that's is what these posts are about. Cross-functional data management is the most important research initiative at Singular XQ, and the one I do the most work on. It's what I'm always thinking about. It should surprise no one that I'm using my personal life and challenges to think the problem through from other perspectives.

So the tone will be a little different this week. But not, really different than before. Despite the irresistible charms of Abe Vigoda, a dead-man-walking, cooly arguing for his life for "old time's sake," (as I'm sure many a politician and AI data broker may be doing in the next few months), Puzo's ultimate moral is true: the distinction between business and personal is a false narrative used by the inhumane. All of this is personal.

I won't belabor how my personal stories connect to the AI delusions, frauds, and deceptions that have been circling us. I trust your intelligence to understand how shocks, odds, predictions, probabilities, randomness, and chaos play into fantasies, delusions, and illusions LLMs are able to create, and draw your own conclusions.

The majority of you have a master's degree or higher, most of you in competitive STEM fields, so a note on specialist language: it is understood the two faces technical and folk psychology terms like coincidence, probability, randomness, chance, likelihood, chaos, and stochastic have. Language emanates from emotional, locomotive bodies, that are embedded in organic and synthetic environments, networked with other bodies real and imagined. When we speak or write, what's in our head and our true intents bleed into the words. Even the most precise, unambiguous speaker will occasionally use a synonym--"on purpose" or "accidentally"--that will reveal an area of anxiety, ambivalence, or confusion they are trying to hide, sometimes from themselves most of all.

In other words, imagine the next female Fields Medal winner really does understand that probability is the superset and randomness is a phenomenon described within that superset, even if at the cocktail party she uses the words interchangeably to non-mathematicians to describe the events that ended her father's life; she might literally have written the book on why "haphazard" is not an accurate description of "randomness" and still slip or find it effective to use one word for the other, especially when she describes the mass shooting that widowed her mother and left her fatherless as "random" in her eulogy. Here is what she understands that a lot of people who have been working in deep data, particularly social data, since the 90s understand:

Because the shooting took place at a place of worship after a series of public speeches endlessly clipped and played on loop on social media concerning her father's religious identity...

Because there were FB DMs, Reddit threads, TikTok videos, text messages, email, and X posts suggesting that something like it should be done...

Because a deep fake of a politician running for local office suggested that this religious community might be causing problems in their town council ...

...and because from longitudinal studies since Little Rock we know that social media can be used effectively for psychological operations to encourage radicalization and foment violence in people who wouldn't normally have acted on their own...

...it isn't a possibility that the shooting would take place...

...it was a predictable certainty.

And while we might think of these chaos actors as a "lunatic fringe," if campaigns achieve even a mere .003% conversion rate in the United States, as it did in the UK in August, at scale here that's still 10,200 human souls acting out.

Because our future Field Medalist knows all of this, she knows that her father's death was not random, it was stochastic. It was part of a growing pattern that needs to be called out and named. Personally, was theoretically possible but unpredictable and improbable that her father specifically would be killed in a mass shooting. Socially, it was certain by a significant degree of certainty, highly probable, and utterly predictable that someone's father would.

Similarly, a lot of us in this community know it isn't possible that these LLMs won't ever be able to reason. It's certain that they won't. To the degree we are able to achieve certainty about anything. As certain as it is someone else will lose a father in a mass shooting.

I think it is time to consider that these epic failures aren't an accident; that the purpose of LLMs was never intended to emulate human reasoning but to dismantle it. This model is the direct outgrowth of the Cambridge Analytica scandal, Steve Bannon's wildly successful gamer strategies, and the 2016 election. "Reasoning" is Janet Leigh with a suitcase of cash naked in the shower.

In the next post, I will tell you a little more about my friend Jason (not his real name) and the coincidences, synchronicities, and statistics about his death not to be self-indulgent (hopefully) but to illustrate that what big data promise has always been about (not the chicanery) is to make better predictions beyond our five senses, and to develop community memory and insight.

The promise is:

what if I can put all the best minds in history together with all the best mind s living today on a noded network and make that network dynamic? What if everyone has access to this like a public library regardless of race, class, orientation, gender, religion, or location? What if we can learn from minds that formerly would have been undiscovered because we leave each other digital traces and living archives, the way we have left stone sculptures, and architecture, and cave paintings, and aqueducts, and the Sofia Hagia? What if learning was continuous, sustained, open, and ongoing and we all decided it mattered and should be conserved and protected?

While I'm talking about difficult realities I remain a techno optimist. I still listen to 80's music. I still like to wear shades. Like the peeping Toms with xray eyes.