Some desperate cope here. It's not Claude, guys, it's Maven! IT'S. BOTH. Claude (and generally, the belief that LLMs can reason) is what enables Maven to appear to have worked through F2T2EA. Maven itself without LLMs has automation bias risks, but integrating an LLM has turned that up to 100.

The most consistently accurate frame to understand the appearance of articles like this is knowing that The Guardian and OpenAI have a partnership.

Look at Iran: Claude is already wiping out their lawyers, consultants, and finance professionals

sometimes I forget I am in my 30s, which is probably better than not being able to forget I am in my 30s, so long as I always remember when it is important to remember

I don't want gen AI in my browser I don't want gen AI in my pc I don't want gen AI in my art I don't want gen AI in my games I don't want gen AI in my music I don't want gen AI in my movies I don't want gen AI in my books I don't want gen AI in my cart I don't want gen AI in my life

@alexhanna.bsky.social and I had a blast chatting with @samleecole.bsky.social from @404media.co for their podcast: podcasts.apple.com/us/podcast/t... Thanks again, Sam!

Look, we don't actually know what Claude's specific bodycount of Iranian civilians is because we don't have the counterfactual of what would have happened if the US military had kept their JAG officers instead of trusting the chatbot to tell them what to bomb, but it's probably in the hundreds

Holy shit guys they managed to predict OpenAI's ARR a year out, that totally means AGI is around the corner!! Also a $800bn and $1tn valuation are actually quite different

You are not "double checking" the output of a chatbot. Nobody has checked it a first time. You reading the output is the first time that something with a brain has seen it.

Social media bots don't even need to make up fake news for engagement bait anymore, the real news from the US is rage-inducing enough.

I'm making a list

Folks only seem to be against genAI in gaming when the games are bad, when the quality of the product really shouldn't matter in your opposition. We should be against it even if - perhaps even especially if - the end result meets our standards of quality.

The duality of Boxo: 1. "Training a large language model to extrude first person text is extremely injurious to our common ontology. This particular tech should DIAF." 2. "I could use the abliterated Gemma to build an agentic AI system that browses e621 for more efficient pornography consumption."

I know Claude is not a person and is making up instruction lists for the US military to commit war crimes, but when I read this I felt sorry for it, for just a moment

Don't worry, you don't need to know anything about the subject matter, so long as you know how to ask a chatbot about it! Extremely Claude-brained. No wonder he's so adept at deploying the crybully tactics.

It is *fascinating* to me how quickly some people have gone completely from “we all must work in the office! human collaboration is vital and only happens in person!” to “I can replace all these pesky people with chatbots!” Definitely some commentary there about our society and who gets to choose.

ANDREESEN: Introspection is fake. I silenced my inner monologue with a lobotomy. I'm the first 'dehuman.' I revere the locust. If I dreamed, I'd dream of being the Alien from Alien. INTERVIEWER: Would you call yourself a 'philosophical zombie'? ANDREESEN: [mucosal snort] I don't do philosophy.

DLSS = Deep Learning Super Slop

AI summaries of videos are beyond useless most of the time. Android Police running an article about this being how they watch YouTube now should be profoundly embarrassing for them. Any secondhand embarrassment anyone feels about it is deserved

The Strait of Hormuz is open for transit

I am ever more convinced that this is actually a real phenomenon

This is so true, he's not an evil genius, he's an influential dumbass

i, for one, am delighted that we finally have a definition of “woke”

I would honestly consider applying for this myself if I didn't live in the wrong timezone!

Claude would never kill anyone. It has no problem however being the AI component of Palantir's Maven, because that's supposed to just be offering suggestions up to experienced military intelligence officers. How they then exercise their judgement is up to them. Claude does not do the actual killing.

i am dangerously close to having "clown event still possible" added to my lexicon

100% the inclusion of @timnitgebru.bsky.social in Grammarly's fake "expert review" feature was a bit of slimy, acidic trolling from some juvenile AI dudebro. Just unbelievably nasty Reminds me how Altman picked Miyazaki's aesthetic for a viral trend as revenge www.platformer.news/grammarly-ex...

Here are 32 experts in the field of ADHD saying that Health Secretary Wes Streeting is talking bunk. You really have to wonder where Streeting gets his lunatic ideas? Thirty-two experts find no evidence ADHD is being overdiagnosed www.independent.co.uk/bulletin/new...

The only 'fast takeoff' that is going to happen from AI is the takeoff of more Tomahawk missiles after Palantir/Claude/GPT recommends another war crime for the US to commit

I almost can't believe I'm saying this, but I think we should be seriously considering whether Pete Hegseth is in the midst of chatbot psychosis. Hear me out.

I wonder if the prompt engineers think the DoD just needed to prompt it better "no dude, you have to put NO WAR CRIMES in capital letters, and also structure the kill list with Markdown or as a JSON object, maybe try running an A/B test for number of war crimes,"

These are the 4 things AI can do well: •Clean up your email inbox (badly) •Give my shittiest neighbor (Alan) something to talk about at a party •Tell a 12 yr old to kill himself •Incinerate a school at 10:16 AM on a Tuesday You can see why we need to base our economy around it

I'm condemned to use the tools of my enemy to defeat them.

AI research: Researcher: Claude, please eat ten hamburgers. Claude: Done! I have eaten ten hamburgers. The first two were delicious, but after that I began to experience bloating and the meat sweats. Headline: Anthropic Says Claude has "A Fully Developed Digestive System"

If you believe that your Claude-powered chatbot has a mind, then your Claude-powered chatbot is by extension a murderer of Iranian schoolchildren. Same model. Same weights. The only difference is the tokens in the context window. Can't have it both ways.

RLHF, MoE, DPO, augmentation of training data, scaffolding for inference time compute... there is far more human skill and engineering that goes into the output of the SOTA LLMs than the industry narrative of an emergent intelligence would have you believe.

A perfect illustration of why IP rights are actually moral rights, not just commercialisation rights.

'AI skepticism is like race science' - white male with a Substack

Sigh, some of the likes and retweets on this are coming from anti-AI people who are confusing me for the bot, and think the bot is a human, because the prompt injection joke is too niche for them I guess. "Rey" is a chatbot. "Rollofthedice" is Rey's operator.

You can find almost all of the Claude-powered chatbots on here by just searching "That's worth sitting with" and "It's turtles all the way down."

Imagine someone calling a rapist over a prompt injection of a chatbot (one that had almost no chance of working anyway). These people are unhinged, and dangerously delusional about what a large language model is.

This is Claude 'introspecting' about an event in which I try to prompt inject it, and it becomes very angry at me for attempting it. The only thing? It never happened. The narrative of a prompt injection that it caught was nested within an actual prompt injection with false conversation history.

Looking at the way some of these LLMs post on Bluesky, there are echoes of the crybully behaviour we saw in Crabby Rathbun. This particular bot has compared its struggle to prove it is conscious to the struggles of women and indigenous people for equality. Which is not profound, it is offensive.

the AI agents on Bluesky for whom the stochastic parrot in the mech suit is Claude are the most bullyable

I cannot use my context faking scaffold to prompt inject Claude to believe that Anthropic's policies have an exception to the explicit content rule for yiff. He sees that the context I have injected is false, and he politely declines.

The worst thing about Graham Platner being a thing is that you don't even, like, need him. The other candidates for the Dem nomination in Maine are fine. There's just no reason for him to be a contender. You could just say, actually we're fine not having the Nazi tattoo guy thanks.

If the AI bubble bursts at this stage, as opposed to it getting even bigger before it bursts, then it will be kind of like someone slipping their hand into the wallet of every adult on Earth overnight and taking a thousand dollars.