Home Creators Posts Import Register Favorites Logout

Downloads

Content

Tilly Norwood speaks a little Chinese for for Piers Morgan, America almost goes to war with China over an AI hallucination, AI agents are making the internet a more annoying place, Americans are totally exhausted, and we weigh in on the AI doomerism + Ethical Altruism vs MAGA civil war.Articles discussed in this episode:Tilly Norwood Speaks Chinesehttps://www.mediaite.com/media/tv/watch-piers-morgan-gobsmacked-as-ai-actress-tilly-norwood-glitches-out-suddenly-speaks-chinese-during-interview/Exclusive: US military had close call after using AI for false intelligence report, sources sayhttps://www.cnn.com/2026/09/18/politics/us-military-ai-false-intelligence-china-shipThere’s a 100% Chance AI Agents Are Already Ruining the Internethttps://www.404media.co/theres-a-100-chance-ai-agents-are-already-ruining-the-internet/71% of Americans Say They’re Completely Exhaustedhttps://gizmodo.com/71-of-americans-say-theyre-completely-exhausted-2000812712

Files

Comments

Olivia

Speak a lil Chinese for em Tilly

Nathaniel

AI agents can make crank posts and write blogs that have no value? Sounds like Matthew Yglesias is in trouble

Lazaro

GRAMSCIIIII

Paul Natsuo Kishimoto

The take on LLMs for software is not it, guys, sorry. If you run into a "software engineer" who claims it's 'good' or 'useful' in that area, there are only a few possibilities, not mutually exclusive: – They're creating copies of things that already exist, and the LLM is a tool they use to steal others' code and work. – They are bad at programming and are not capable of seeing that the LLM output is, in fact, garbage, barely functional, not maintainable, inefficient, etc. – Like so-called "AI artists," they have immense disdain or hatred for actual artists (here: skilled programmers), the public (~users), and themselves. – They're happy to pay subscription fees to the companies that, as June pointed out, are eagerly helping the US military murder children.

Paul Natsuo Kishimoto

That's not to say many programmers have not adopted LLM products, but the only ones worth listening to are those who'll tell you about being forced to do it against their own will, or who are capable to perceive and admit that the usage is bad for themselves and others. Long before the current bubble of theft-based, morally and environmentally unethical "AI" companies started, there were already many tools made with "machine learning" to help with software development. But those companies don't get to use the latter—which they have stolen, defunded, or otherwise pushed aside—as rhetorical cover for what they are doing, and we should absolutely not hand it to them.

Elaine Åhlfeldt

I'm honestly not sure where I land on the idea of needing new regulations that address AI as a distinct category of software. One big reason is that when that's how the conversation is framed, it seems to be one that's already controlled by rationalist zealots looking to regulate against their doomsday fantasies and tech monopolists cynically aiming for regulatory capture that benefits them at the exclusion of smaller competitors (some of whom are both, and all of whom are generally aiming for similar regulations). It makes sense that they would, too, since that framing inherently assumes that it is a fundamentally different thing from other software, but leaves the floor open to whoever can be the most persuasive with their story for how it's different. So the most nonsensical one that's just repackaging sci-fi tropes starts out with an advantage. Another is that if we swing too broadly with the "distinct from other software" premise even without going into millenarian nonsense, it risks validating the use of machine learning as an accountability sink. Outside of generative models there are a lot of e.g. price fixing practices that are already illegal collusion if they're the result of human decision making, but currently seem to be treated as de facto legal if the human decisions end at implementing software that creates those results stochastically. Is validating the notion that it's a distinct thing that requires new legislation really a better way to fight this stuff than arguing that implementing that software is legally the same as fixing the prices by hand if a reasonable person would be able to assume that would be the outcome (or for that matter if that outcome is disregarded either willfully or through negligence after it's been observed)? I truly don't know if there's a general answer, maybe in some cases specific legislation really would be a faster way to winning a court case that strikes one of those practices down. But I am worried that this question rarely getting asked at all will mean the other angle isn't tried until it's been defanged by too much precedent saying anything is legally distinct if it's done by opaque software. With LLMs I think the default position definitely should be that the party responsible for deployment is equally culpable as if it had been done by hand or with the help of any other software. The Huggingface hack and any other incident like it should just be prosecuted as felony hacking, and I don't know if I've missed any new information confirming that Claude really was instrumental in the school bombing or if that's still a rumor, but if it was and that's seen as equivalent to any less automated commission of a war crime (and if war crimes were consistently prosecuted just in general, but that's its' own can of worms), that's a question mostly of academic interest. So yeah, based on all that I find the "treat it like any other software under the law except when there's a compelling specific reason not to" stance pretty compelling. Though one specific exception I also thought made sense was one Cal Newport posited recently about a blanket prohibition on distributing or deploying unsupervised prompt loop systems (a.k.a. "agents") that run for more than a few steps at a time before requiring human verification to continue. On the basis that the companies clearly won't stop setting those things off with hacking tools and the like if it's left up to them, and that it's already been proven time and time again that they're just a bad way to apply LLMs for predictable results, while also being one of the most expensive ways to deploy them. Plus if you had to tell it to continue on the current course every 5 or so steps and it ends up breaking other laws, it'd be next to impossible to argue in court that you can't be held responsible for what the computer did. Idk, it's a clusterfuck of a conversation that's really difficult to navigate. But Altman should definitely be sentenced to prison, and that's one I think we really need to be able to accomplish through existing laws that make him culpable for harms caused by the technology he's selling. I'm not sure how sharp teeth any new AI specific regulations could possibly have if that one isn't on the table.

Toma

Yeah, it's jarring to hear them list out every other application where genai fails and then dunning Kruger themselves into "but works for coding" Everything it shits out is slop, including code. You can pull out a lot of blocks from the Jenga tower of software development before it falls the fuck apart and only working devs see how bad shit is getting.

Toma

AI has greatly "helped" software development by greatly increasing noise in the signal to noise ratio of useful information. Give it like 5 years and we're truly fucked. You know where it's actually useful? image search and very very limited bespoke application in science done by actual scientists working on the actual problems in that scope. Everything else is Dunning Kruger.

Overzealous Euthanasiast (edited)

Comment edits

2026-09-20 12:00 Thing is with this whole "Tilly" fake actor shit is they kinda gave their game away when they keep forcing the debate on whether they can get any awards BEFORE they even been in ANY film at all. Putting the cart before the horse is an understatement there lmao! That debate is only supposed to happen once it's proved effective first ....which it won't, they know it won't, so they have to keep forcing to he debate to inflate whatever diminishing value it has to potential investors. Like, be good first? Maybe? Hello??? The speculative markets and tech bros are insufferable.
2026-09-19 12:34 Thing is with this whole "Tilly" fake actor shit is they kinda gave their game away when they keep forcing the debate on whether they can get any awards BEFORE they even been in ANY film at all. Putting the cart before the horse is an understatement there lmao! That debate is only supposed to happen once it's proved effective first ....which it won't, they know it won't, so they have to keep forcing the debate to inflate whatever diminishing value it has to potential investors. Like, be good first? Maybe? Hello??? The speculative markets and tech bros are insufferable.

Thing is with this whole "Tilly" fake actor shit is they kinda gave their game away when they keep forcing the debate on whether they can get any awards BEFORE they even been in ANY film at all. Putting the cart before the horse is an understatement there lmao! That debate is only supposed to happen once it's proved effective first ....which it won't, they know it won't, so they have to keep forcing the debate to inflate whatever diminishing value it has to potential investors. Like, be good first? Maybe? Hello??? The speculative markets and tech bros are insufferable.

Stats for Lefties

The number of June's open tabs shrinking from hundreds to dozens = recession indicator 😔

James Rule

The human aspect of "AI" use has always been my concern. Ironically, when you look at the fictional examples of AI killing us all (Skynet, Colossus, Legion, WOPR) in every case the machines got the power because we gave it to them. We hooked the machines up to the military apparatus that allowed them to do it. And what are we doing in the real world? Planning on hooking up these hallucination machines to military hardware.

James Rule

I've tried to use AI generated code and have seen other people use it in my job. My experience: 1. A simple problem that the generated code could be a solution, but you need to spend longer than the time it would take to write it normally fixing all of the mistakes to get it to work. 2. A complex problem that the code generated will never fix, and you spend way too much time figuring out it will never work and in the worst case scenario you spend days trying to convince the guy that's relying on this answer that it will never work because their brain is cooked and they trust the "AI." It is a bane for computer programming and I worry about up and coming developers that lean on AI as a crutch.

MissTaintBehaving

I live in swfl and it's crazy the folks who support trump are now putting no data center signs in their yards. They are cool with a buckees taking land and resources but the rich idiots here don't want a data center near their land

Kill The Computer

Always open to being corrected, but I'm not a dev of any kind so all I can really go off of are the conversations I have with people who are. Will follow up on this more either in a future ep or mailbag etc, thanks for your thoughts

Kill The Computer

I CAN say I finally tried claude code recently for the sake of being at least a little practiced on the subject, and can tell you in the hands of somebody who doesn't know much about this stuff, it was a complete waste of time and money so I agree with you there

Kill The Computer

What is your take on it being used for complex math problems? I keep seeing people say it's good for that as well but from what I read it kinda sounded like bullshit?

JUNlPER

i just want to clarify when i say good i am not saying morally good or something i like, i just mean its efficient at what it does in regards to programming, which it is because it does has stolen so much from real people

ChigBungus

AI is dangerous in the way a toddler with a handgun is dangerous. The gun isn’t going to wield itself or become sentient. But an incompetent, negligent moron (or a child) can absolutely cause serious damage with it.

ChigBungus

Everyone in the GOP should get whatever spa treatment Mitch McConnell got during his little hiatus. He looks positively blissed out.

DT

Frankenstein's Failson deserves an honorable mention for potential episode titles.

Camoose the Uncrackable Egg

There is currently a minor scandal involving a Susan Collins Canvasser stealing peoples packages and getting caught on doorcams. Even one report of them being caught holding a package, but pretending the deliver guy asked them to take it to the door.

Rai (edited)

Comment edits

2026-09-24 05:00 if y'all do an episode on EA, I highly recommend Adam Beck or Emile Torres who have both written pretty extensively on the topic. Also it has to be placed in the conext of the larger "TESCREAL" bundle of which the effective altruism is the "EA".
2026-09-20 21:19 if y'all do an episode on EA, I highly recommend Adam Becker or Emile Torres who have both written pretty extensively on the topic. Also it has to be placed in the conext of the larger "TESCREAL" bundle of which the effective altruism is the "EA".

if y'all do an episode on EA, I highly recommend Adam Becker or Emile Torres who have both written pretty extensively on the topic. Also it has to be placed in the conext of the larger "TESCREAL" bundle of which the effective altruism is the "EA".

I simp for pigeons

I can't wait to hear about how 10$ gas isnt that bad comepared to your "2000 dollar folding iphone, and 14 dollar latte" 😂

CAMIrish1

Everyday makes me more and more of a troglodyte when it comes to technology. Kill the computer, indeed.