Content Tagged “ai”

• Palantir and the US Strike on the Iranian Girls School

  This is the best analysis I’ve seen of the US military strike that killed over a hundred young girls in a school in southern Iran.

  By the time the war began, “AI safety” and “alignment” and “hallucination” and “stochastic parrots” had become the terms of every argument about artificial intelligence, structuring and limiting what we could even say. Worse, “artificial intelligence” itself had come to be synonymous with LLMs. When the school was bombed, those were the terms people reached for, despite the fact that this critical apparatus offered a poor fit for the older, more mature stack of technologies involved in targeting. The real question, the question almost nobody was asking, is not about Claude or any language model. It is a bureaucratic question about what happened to the kill chain, and the answer is Palantir.

  → The Guardian: AI got the blame for the Iran school bombing. The truth is far more worrying

  ---

• Israel’s AI Howitzer

  War is horrible. The only thing that seems more horrible than war, to me, is war that isn’t even waged by humans, but by computers.

  Israel’s military has taken a step toward a more automated battlefield, deploying a new artillery system that uses artificial intelligence to compress the time between identifying a target and opening fire. The system, known as the Ro’em, was deployed by the IDF’s 282nd Artillery Regiment during operations in southern Lebanon, where it was used to strike Hezbollah rocket and anti-tank launch sites in support of ground forces.

  Where its predecessor, the US-made Doher, relied on manual processes and large crews, the new system operates with a level of automation that brings it closer to a semi-autonomous battlefield machine. At the center of that transformation is artificial intelligence. Unlike traditional artillery, where targeting, loading and firing are largely manual, the Ro’em compresses these steps into a largely automated sequence. Once a target is designated by an operator or commander, the system can independently load ammunition, calculate firing solutions, aim and fire with minimal human intervention. It is also integrated into broader military command-and-control networks, allowing it to receive targets directly from intelligence systems or operational headquarters.

  → CTech: Israel’s new AI-powered artillery makes combat debut in Lebanon

  What could possibly go wrong?

  ---

• Meta Buys Moltbook

  This is crazy.

  → Business Insider: Meta just bought Moltbook, the viral social network for AI agents

  On Tuesday, Meta confirmed it had acquired Moltbook, a Reddit-style forum for AI agents. As part of the deal, first reported by Axios, Moltbook’s creators, Matt Schlicht and Ben Parr, will join Meta’s Superintelligence Labs. Meta did not disclose a price for its Moltbook deal.

  Essentially an acqui-hire. Which is insane, because those guys bragged that they didn’t write a single line of code for Moltbook. Not only is Moltbook in itself a delusional idea, it is also a security nightmare, as it has been hacked more times than I can count. And it is a huge scam. Through the hacks, it was revealed that most “agent” accounts on there were actually run directly by humans. At one point, one guy created over 60% of all accounts on the platform through a script. I’ve discussed all of this on this podcast episode. But what came out after that, is that now people use it to promote shitcoins. Because you can just talk agents into spending their user’s money on those and then rugpull their asses.

  The most interesting theory I heard on this acquisition was advanced by Buzzkill Jr. on the latest episode of the DH Unplugged podcast: Meta bought Moltbook because they see it as the future for advertising. Why advertise to humans when you can advertise to their AI agents instead? Since these lack all critical thinking skills, they’re much easier to convince to spend your money than you are. 🦀🦀🦀

  ---

• You Need Saving, Not That Game

  This is the dumbest tweet I’ve read all week:

  

  These AI bros just continuously pump out the lowest IQ bullshit on the internet. No wonder they need artificial intelligence, they certainly don't have any of their own.

  This game is abandonware. I you search for its name, using a search engine, this is the first hit. Not only can you simply download the game and it plays perfectly well in DOSbox, you can play it in your browser right on that site!

  Why does this dude think that this game needs to be “revived”? And at the horrendous waste of energy inherent in using an LLM for six hours, no less? Kids, stop using AI. Angela Collier is right, it turns your brain to sludge!

  ---

• Clippy for LLMs

  Oh god. Why do I do these things to myself? While researching some stuff I remembered about Clippy — after being reminded by the stupid Wikipedia mascot — I discovered this abomination:

  Some dude made a Clippy skin for LLMs. I mean, the way in which those LLMs talk is pretty reminiscent of Clippy and about as annoying, I will give him that. But this is just pure insanity!

  

  Reading through the project's README file, the dude indeed seems pretty delusional:

  I am so grateful to Microsoft - not only for everything they've done for Electron, but also for giving us one of the most iconic characters and designs of computing history.

  Looking at his website, I think I know why.

  I work on Claude at Anthropic. Electron Co-Maintainer. Maintainer of tons of JavaScript modules.

  Turns out he studied here in Düsseldorf and used to be editor-in-chief of the Apple fanboy site Apfeltalk back in the day, before heading to Silicon Valley and touring all the cool new start-ups and tech trends.

  This is your brain: 🧠 This is your brain on Silicon Valley: 🧀

  ---

• Wading Through AI

  Casey Muratori has started a new video series called Wading Through AI with his friend (and AI researcher) Demetri Spanos. They have just released the first episode and it’s excellent. Like anything from Casey, it is well worth a watch!

  → Should You Be A Carpenter? – Wading Through AI, episode 1

  ---

• Casey Muratori is Right on Everything

  I’ve been watching stuff from Casey Muratori for a while now. He seems to be a great guy. I started with Handmade Hero and I’ve now followed all of his stuff everywhere, I think.

  Here he is on a video podcast, explaining why a) software is so shit these days and b) being completely correct about AI on several levels at once. The dude is great. I love him.

  ---

• Richard Sutton on LLMs

  So, I’m coming to the realisation that I don’t really have a problem with AI. I have a problem with LLMs specifically. And the problem is that LLMs don’t deal with reality. They get their input from us. Essentially, we describe the world with language — in a very imperfect way — and then we upload that to the internet. The LLM takes that data and forms a picture of the world from that. But it’s a picture from a picture. The LLM doesn’t act on the real world. It can’t, because it does not understand the real world, it only understands language — which, anyone who’s studied linguistics for a bit will tell you, is a very imperfect representation of the world in the first place. But there turns out to be another way to build AI. It is called reinforcement learning, or RL, and it could probably be used to build something that is actually useful. Because it operates on a picture it has built of the actual world.

  I came to this conclusion by watching the following interview with Richard Sutton, a legendary AI researcher and major proponent of RL. It is well worth a watch, the whole hour of it.

  → Dwarkesh Patel: Richard Sutton – Father of RL thinks LLMs are a dead end

  The two publications referenced in this video:

  • The Bitter Lesson, Richard S. Sutton, 13 March 2019
  • The Big World Hypothesis and its Ramifications for Artificial Intelligence, Khurram Javed & Richard S. Sutton, 2024

  ---