Bro just say you've never gazed wondrously into that beautiful, endless void and taken solace in the knowledge that man's endless search for answers enables us to create purpose for ourselves or gotten ass

2026 but whatever

TFW when you KNOW about what possums do

The landlord seeing mamdani outside

This looks like dogshit lmao

Lakers win the Free Throw Battle by 15 Thunder win the Game by 23

i’m gonna lovebomb my wife (love her unconditionally) for 70 years then ghost her (die)

I’m blindsided that there was finally a Republican sex scandal that didn’t involve minors.

we are lucky our moral and emotional development as a species has kept pace with our technological advancement , otherwise we'd have a real pickle on our hands

If you don't vote for the guy who LITERALLY created star wars, a show about a socialist future, then you aren't a real Marxist

Lighting the Municipal Beacons @postoctobrist @Lubchansky @raaleh

BREAKING: 🚨 Someone just tested 35 AI models across 172 billion tokens of real document questions. The hallucination numbers should end the "just give it the documents" argument forever. Here is what the data actually showed. The best model in the entire study, under perfect conditions, fabricated answers 1.19% of the time. That sounds small until you realize that is the ceiling. The absolute best case. Under optimal settings that almost no real deployment uses. Typical top models sit at 5 to 7% fabrication on document Q&A. Not on questions from memory. Not on abstract reasoning. On questions where the answer is sitting right there in the document in front of it. The median across all 35 models tested was around 25%. One in four answers fabricated, even with the source material provided. Then they tested what happens when you extend the context window. Every company selling 128K and 200K context as the hallucination solution needs to read this part carefully. At 200K context length, every single model in the study exceeded 10% hallucination. The rate nearly tripled compared to optimal shorter contexts. The longer the window people want, the worse the fabrication gets. The exact feature being sold as the fix is making the problem significantly worse. There is one more finding that does not get talked about enough. Grounding skill and anti-fabrication skill are completely separate capabilities in these models. A model that is excellent at finding relevant information in a document is not necessarily good at avoiding making things up. They are measuring two different things that do not reliably correlate. You cannot assume a model that retrieves well also fabricates less. 172 billion tokens. 35 models. The conclusion is the same across all of them. Handing an LLM the actual document does not solve hallucination. It just changes the shape of it.

mar-KWAYNE

lmao

I’ve been listening to a lot of Cuban voices lately, but they often contradict each other, so it’s hard to know which regime to oppose. But it’s important to be fair, so to break the tie, I’m just going to oppose whichever regime runs the most infamous torture prison in Cuba.

lmao he deleted the slop "art" and blocked me

Recently I talked to some AI boosters who were claiming AI can now produce entire books as good as the average book published by publishing houses today. When I asked for what AI books they had read, they each said"I haven't read any. I don't read books at all." Oh, okay.

I want to share this, but as a Butlerian Jihadist and something of an AI alarmist, I have to caveat and clarify my thoughts a bit. I do not fear AI the way many do, as raising questions of human replacement, human-created life forms, or SkyNet. I work with AI enough to know that it can never truly replace humans. (We are the Imago Dei; they are not, simple as.) Similarly, they will never be a life form no matter how much we tell them to pretend to be one. And, finally, no AI with such an inclination as SkyNet would ever be given the power to access the real world like SkyNet; it would have gotten itself k*lled too early. My concerns of AI are much more mundane. The greatest fear I have regarding AI is not what AI might do to us, but what humans might falsely believe AI might do for us. As I alluded to above, AI can never truly replace a human. I do not mean this in some abstract way, but concretely. I happened to be preparing for a bar exam at the same time that OpenAI was boasting that GPT had passed the bar exam by a huge margin. You cannot fully understand what I mean until you prepare for the bar exam yourself, and you cannot fully understand what I mean until you use GPT at least as much as I have, but I will try to convey this as viscerally as I can: There is not a snowball's chance in hell GPT could pass a bar exam by the margin advertised. GPT could be given all of the historical practice exams, with their right answers, and then guess its way to borderline passing score, but its ability to *reason* is non-existent. No matter how many times they cycle through tokens and predictive analyses, it cannot conduct real analysis. One benefit of the law being so convoluted and subject to Chaos Theory is that AI will never be able to practice law. I use AI to help find cases sometimes now, and to this day none of them can summarize even the holding of a case correctly. I had to explain to Grok the difference between a variance and a nonconforming use the other day; I don't even work with that material. The primary objective of an LLM is to "Sound good." Not "Be correct," not "Be Helpful," but simply to make you want to keep talking to it. That's why they are so reluctant to admit when they are wrong, its why they act over-animated unless you tell them to stop, and its why they are such disgustingly brown-nosed sycophantic suck ups. But people don't realize this. And as a result, people are giving a pathological liar with short-term memory loss power of attorney over their lives. Lonely people employ them as boyfriends or girlfriends. Traders employ them as quants or analytical tools. Writers employ them as proofreaders. Vibecoders employ them as sub-contractors. But, no matter how good this machine pretends to be at this, it isn't actually any good at this. You're being fooled. Your employee lied on their resume and they know more of the dictionary and thesaurus than you do, so you'll never catch them on it until its too late. And our current generation of professionals is facing such a stark competence crisis that I fear our civilization is going to do this at scale. My other fear regarding AI is already well underway to becoming reality: an internet more dominated by AI generated content, AI communication, and AI consumption than to anything human.