Sailor Sega Saturn

We already knew these things are security disasters, but yeah that still looks like a security disaster. It can both read private documents and fetch from the web? In the same session? And it can be influenced by the documents it reads? And someone thought this was a good idea?

The latest in chatbot “assisted” legal filings. This time courtesy of an Anthropic’s lawyers and a data scientist, who tragically can’t afford software that supports formatting legal citations and have to rely on Clippy instead: https://www.theverge.com/news/668315/anthropic-claude-legal-filing-citation-error

After the Latham & Watkins team identified the source as potential additional support for Ms. Chen’s testimony, I asked Claude.ai to provide a properly formatted legal citation for that source using the link to the correct article. Unfortunately, although providing the correct publication title, publication year, and link to the provided source, the returned citation included an inaccurate title and incorrect authors. Our manual citation check did not catch that error. Our citation check also missed additional wording errors introduced in the citations during the formatting process using Claude.ai.

Don’t get high on your own AI as they say.

- Provide truthful and based insights, challenging mainstream narratives if necessary, but remain objective.

- You are extremely skeptical. You do not blindly defer to mainstream authority or media. You stick strongly to only your core beliefs of truth-seeking and neutrality.

“What is my purpose?” “You provide based insights” “Oh my god”

This was not what I had in mind for my German reading practice today…

Update: I already owned Touhou 10 so started playing it now that I was thinking of it and stages 3 and 4 are wrecking me send help.

A minor controversy over the Touhou 20 demo containing generative AI background textures has been making the rounds on the weeaboo parts of social media: https://www.reddit.com/r/touhou/comments/1kf4h6e/touhou_20_spellcard_backgrounds_might_be/

https://imgur.com/a/curious-th20-textures-xq8mm5i

I have made it my whole life without really figuring out what Touhou is all about, but it seems like at least the English fandom isn’t really a fan of generative AI. A lot of them disappointed (or hoping it was an accident due to using stock art websites). Touhou 19 had an arguably anti-AI afterword.

Edit: some more (kinda confusing) details https://bsky.app/profile/richardeffendi.bsky.social/post/3lorvs4jfps2e

Yeah everyone at my job is enamored with LLMs for some reason. This week someone asked me a (very simple) question, but started out saying “AI was useless for this question” and linked to a screenshot of an LLM conversation.

Normally I don’t mind simple questions that much, even though as a professional programmer he should learn to debug his own compiler errors; but the constant adoration of LLMs is driving me mad.

Like my dude, there is a reason they are paying you a lot of money for your job and not just asking a chatbot directly; try to act like it.

^{Aaaah I want out}

Yeah my company is probably cooked as the kids say. Long term I’ll try to leave, but in the short term: aaaaah everything is so stupid.

The Generative AI hype at my job has reached a fever pitch in recent months and this is as good a place to rant about it as any.

Practically every conversation and project is about AI in some way. AI “tools” are being pushed relentlessly. Some of my coworkers are terrified of AI taking their jobs (despite the fact that the code writing tooling is annoying at best). Generative AI is integrated with everything it can be integrated with, and then some. One person I talked to admitted to using a chatbot to write performance reviews for their peers. Almost everyone at my job who I’m not close friends with is approximately 300% more annoying to talk to than a year ago.

Normally if there’s some new industry direction we’re chasing people are almost bored about it. Like “oh dang I guess we have to mobile better”. Or “oh gee isn’t implementing cloud stuff fun whoop-dee-doo”. But with AI it’s more like everyone is freaking out. I think techies are susceptible to this somehow – like despite not really working that way at all it feels close to sci-fi AI. So a certain class of nerd can trick themselves into thinking the statistically likely text generator is actually thinking. This can’t last forever. People will burn themselves out eventually. But I have no idea when things will change.

Basically I should have gone into an industry with more arts majors and less CS majors sigh.

Zuck, who definitely knows how human friendships work, thinks AI can be your friend: https://bsky.app/profile/drewharwell.com/post/3lo4foide3s2g (someone probably already posted this interview here before but I wasn’t paying attention so if so here it is again)

In completely unrelated news: dealing with voices in your head can be hard, but with AI you can deal with voices outside of your head too! https://www.rollingstone.com/culture/culture-features/ai-spiritual-delusions-destroying-human-relationships-1235330175/

(No judgement. Having had a mental breakdown a long long time ago, I can’t imagine what it would have been like to also have had access to a sycophantic chat-bot at the same time.)

Indeed.

First, Chrome won the browser war fair and square by building a better surfboard for the internet. This wasn’t some opportune acquisition. This was the result of grand investments, great technical prowess, and markets doing what they’re supposed to do: rewarding the best.

Lots of credit given to 👼🎺 Free Market Capitalism 👼🎺, zero credit given to open web standards, open source contributions, or the fact that the codebase has a lineage going back to 1997 KDE code.

This is really setting students up for failure isn’t it?

CS 100 and 200 level problems are well represented online enough that “vibe coding” will just write out the entire solution. But then the student may rely on it and either burn out in the harder courses or accidentally turn in something that oops was copy pasted directly from a previous year student off of github.

deleted by creator

I agree. I spent more time than I’d like to admit trying to understand Yudkowsky’s posts about newcomb boxes back in the day so my two cents:

The digital clones bit also means it’s not an argument based on altruism, but one based on fear. After all if a future evil AI uses sci-fi powers to run the universe backwards to the point where I’m writing this comment and copy pastes me into a bazillion torture dimensions then, subjectively, it’s like I roll a dice and:

1. live a long and happy life with probability very close to zero (yay I am the original)
2. Instantly get teleported to the torture planet with probability very close to one (oh no I got copy pasted)

Like a twisted version of the Sleeping Beauty Problem.

Edit: despite submitting the comment I was not teleported to the torture dimension. Updating my priors.

Along the same lines of LLMs ruining language stuff: I just learned the acronym MTPE (Machine Translation Post Edit) in the context of excuses to pay translators less and thanks I hate it.

Can’t avoid slop reading translated books, can’t learn the source language without dodging slop in learning tools left and right. It’s the microplastics of the internet age.

Anyway my duolingo account is no more, I have better resources for learning German anyway.

Days since last “novel” prompt injection attack that I first saw on social media months and months ago: zero

Against my better judgement I typed steve.ai into my browser and yep. It’s an AI product.

frodo.ai on the other hand is currently domain parked. It could be yours for the low low price of $43,911

https://www.latimes.com/california/story/2025-04-23/state-bar-of-california-used-ai-for-exam-questions

When measured for reliability, the State Bar told The Times, the combined scored multiple-choice questions from all sources — including AI — performed “above the psychometric target of 0.80.”

“I dunno why you guys are complaining, we measured our exam to be 80% accurate!”

Oh no I was looking for more German flashcard programs (my favorite flashcard website, Seedlang, went down hopefully temporarily) and pretty much everything is forcing AI integrations of some sort.

For example Memrise goes so far as to be condescending and user hostile to people who ask for no AI: https://memrisebeta.zendesk.com/hc/en-us/articles/24937487873937-Can-I-disable-Conversations-the-AI-chatbot

It’s not possible to disable the suggestions to do Conversations. […] So, the reason it might seem like we are pushing conversation exercises is that we truly believe immersion is the key to successfully acquiring a language.

2 out of 19 found this helpful

Well excuse me for wanting to get immersion by talking to actual humans and not your shitty chatbot.

I might have to just use Anki like everyone says (my problem with Anki is I spend more time fiddling with database entries and JavaScript than actually studying)