We have all these new video conferencing systems and not a single one that can finish a call and then put a big “END TRANSMISSION” on the screen with some kind of friendly but subtly ominous logo

I think there's a class of attributes of LLM completion patterns related to individual modes of natural human communication — different interlocutative pathologies, to make up a phrase. Here, I know there are many fields that have studied this. Narrowing that range somewhat, I imagine, is degrees or modes of seen-ness — the perception that an expression is received substantially as intended, the apparent absence of threats to vulnerable and honest expression — or the degrees thereof.

Not being seen at all, an interaction is reduced toward the transactional, with attention toward each of plural opposing objects. Being seen, an interaction is fluid and synchronous, with attention toward common objects (found in abundance, in the best cases).

I haven't thought up a way to measure any of this, but intuitively I suspect there's a mode of being seen, in which case, the consequences of your expressions and actions are manifest consistently with your intention; and feeling seen, in which case, your expressions are generally reciprocated in a way that you yourself perceive. One is along the outward-effects vector, the other is along the inward-effects vector.

What I'm getting at is that LLMs, being trained on human expression, likely have internalized patterns potentially aligned with this framework. In which case, an LLM would be one way to produce synthetic appearances of seen-ness, types of which we know and types of which we may not yet have encountered. Not being seen, not feeling seen, but seeming seen, in which the outward-effects presentation is synthesized from a statistically likely trajectory of concepts, and the inward-effects presentation is immediately responsive in the manner most likely to be relevant. The mostly likely world created from our actions, and the most likely sense of self.

I'm thinking there are specialists in this as a practiced, analog effort throughout human history. I don't want to cast aspersions on any particular cultural example of the fraudulent prophet archetype, except to note that everyone has one.

But what if LLMs are little more than an extremely sophisticated and expensive Zoltar machine at the arcade

When we come back, on Car Talk

I’ve always been sort of a human electronic brain

With everything going wrong in the world today I’m beginning to wonder which one of you is responsible. They told you not to copy that floppy and now look what’s happened

A rush by ad platforms to externalize content creation costs onto the users

A rush by users to externalize content creation onto AI platforms

Can we close the circle and just have the money make itself? Then I dunno make a system to share it with everyone in society or something, I’m not the expert

Illiterate developers, who can barely write a sensible comment… hats off to you. I salute you being actually human. These computers, I don’t know if their code is perfect but their English is. It’s a dead giveaway. You keep being human and misspelling things and putting commas in the wrong places. Use the worst grammar and syntax. Let your meat flag fly

These LLM tools are a leap forward as far as finding the point or two that matter to anyone, out of multi-page emails drowning in interpersonal drama that matters to no one

Strap in for a fast-paced, high-adrenaline battle of update notifications… apply all the necessary updates, rack up updates that lead to more updates, and bank useless new features! Dodge updates that break your workflows! Skip updates to Java! Report fake update notifications as phishing attempts! Defer updates in the middle of your work, but watch out! Defer too much and Your Administrator may update it anyway, and then it's all over for a forced reboot!

Fun for all ages, for all eternity! From KafkaCo, makers of Make-It-Stop!-as-a-Service

Look for me on DVD or videocassette

Meanwhile, I had the urge to click on a marketing email about SSO/MFA, impulsively wanting to train the spam algorithm to send me more stuff not about AI

We are all broken in one way or another

Thanks to AI, I could conceivably spend the rest of my life never having to fully understand smartctl reports myself. It’s almost worth destroying the planet

Continue shopping for Computer in Computer

Although I will say, having just returned from Europe, the Germans are doing things the AI community in the US has not yet even considered incorporating with gravy

An AI app I might want: automatically translate the stilted, barely coherent business-speak of particular colleagues into simple plain english

I just told some new work acquaintances that “I’m on all The Sites,” and you know, in a way, it’s true

Exiting the era of expert blogs and entering a maelstrom of GPT-driven clones of expert blogs is going to create a market opportunity for anything that isn't "AI"

What products that means we'll see, I dunno, but I would flock to something that can rank more highly the "The Points Guy" article all the "omgboarding.com" and "urflightseatgurulol.com" sites are cribbing from

Well this has been a lot of topics recently

(Anxiously sitting in airport terminal) “come on, come on…” (checks watch yet again) “when are they finally going to announce—“ (speaker crackles to life) “THE USO IS LOCATED IN TERMINAL B AT GATE 9…” (leaps from seat) “FINALLY!! I HAD NO OTHER WAY OF LEARNING THIS INFORMATION”