keltan — LessWrong

LESSWRONG
LW

keltan — LessWrong

Replying toMy journey to the microwave alternate timeline

My journey to the microwave alternate timeline

This was a super fun read! Especially with the recommended soundtrack.

I recently found some vegan pork buns that are ready after 4 minutes in the microwave. They come in microwave safe plastic, that kinda looks like a cup holder from a fast food place. You fill the bottom with a little bit of water before putting them in.

On the day my partners and I discovered them, we ate both packets in one sitting, then went back to the store and bought another 5 packets, and ate another 3 that day. Microwaves are incredible! People need to know this!

keltan9d*Quick Take

I think some people would be surprised how much hidden information already exists in what seem to be pretty standard AI responses.

A game I like to play with Claude is:

Send it a photo of a celebrity
Get the standard rejection "Sorry, I can't name people in images, but I can describe what I see..." which tends to be followed by a pretty general description
Then asking a different instance of Claude to guess the celebrity, based on the standard rejection response

Since Sonnet 3.5-7, Claude has been pretty ok at this game.

I used to play a different version with famous YouTube links:

Send YT link to Sonnet 3.5
"Sorry I can't open YouTube links..."
Ask a new instance

keltan9d

Conditional Kickstarter for the "Don't Build It" March

Flag mock-ups based on the IABIED website icon (rough. I'm not a designer. Just trying to make it CGPGrey approved.):
1: Keep it simple: Something a child could draw
2: Make it distinct at a distance: No tiny details
3: Three colors or fewer
4: Symbols, colors and designs should mean something
5: Words on a flag: ideally zero
5b: Don't write your name on your flag

The Square:

Large Square:

Smaller Square:

Square in context:

I wasn't a fan of this flag, but kind of like it when it's in the context of a real march. The shirts are probably the best bit though. I image the back of them to have a slogan on them.

The Halt:

Based of this Australian cave art hand stencil:

Halt (Hand Only):

Halt (Hand Stenciled):

Halt (Hand Stenciled, Wrist included):

Halt in Context:

-6

Replying to36,000 AI Agents Are Now Speedrunning Civilization

keltan13d

36,000 AI Agents Are Now Speedrunning Civilization

Thanks for making this post! I'd seen stuff about Moltbook around, but was unclear on what it actually was. I found this clarifying

Replying toRecent AI model progress feels mostly like bullshit

keltan15d

Recent AI model progress feels mostly like bullshit

I'd be interested to know if your lived experience has changed since Opus 4.5 was released?

keltan22d

I would imagine that if an intelligent species' home planet was overtaken by ASI, we would likely see evidence of that takeover. In the way that the ASI would be incentivized to expand into that star system and beyond as rapidly as possible.

Though, I guess it might be within the ASI's interest to hide from other intelligent minds for as long as it could. Giving those other minds, less time to prepare for the ASI's arrival.

keltan1mo

Thank you! Also, Time Stamped Link: https://youtu.be/UECn73UVILg?si=UqTHRBWnrklR8bWs&t=203

Replying toIn My Misanthropy Era

keltan1mo

In My Misanthropy Era

I like Bentham's definition of what gives a being moral worth:
"the question is not, Can they reason? nor, Can they talk? but, Can they suffer? "

Replying toArtifacts I'd like to try

keltan2mo

Artifacts I'd like to try

Inspired by your post, I very quickly (~2h) vibe coded an Obsidian Plugin, based on your River Timeline idea.

The Plugin crashes pretty often, and has some major problems. But there is something interesting here. I have no plans to publish on Obsidian/work on the plugin further. If anyone becomes interested in this project, they are welcome to steal all this code without credit.

keltan2moQuick Take

I sometimes hear people say "The MIRI warnings of AI risks are out of date. There are different risks now, but MIRI hasn't changed". What do people mean by this? The MIRI arguments still seem to hold up to me?

I asked Claude for it's input, and received this answer; which seems like a good breakdown?

Claude's Response
This is a criticism I've seen circulating in AI safety discussions, and there are a few distinct claims people seem to be making:
The "outdated threat model" argument
Some critics argue MIRI's core warnings were developed when the assumed path to AGI looked quite different—more like classical GOFAI or relatively interpretable systems where you might hope to formally verify

... (read more)

•••

To write well; first, experience.

keltan

3mo

"On living, for the sake of your writing"

This is my most liked Substack post!

By than, I mean it got 2 likes, because I don't really show anyone my blog. But, given Inkhaven/Half-Haven, are both soon to end; I thought I'd share this one here.

Opening of the Post

Epistemic Status: Based on ‘vibes’ not science.

Over the past two months of writing every day, I have grokked the advice I have so often heard from writers, that you must experience life, to be able to write well.

I doubt I’ll be able to instill this idea into your head, in a way that the writers I admire were unable to do for me; with their flowing words and exemplary prose. But, I will still give it a go.

OpenAI plans to introduce a 'TikTok-like' short form video product, using Sora to generate the platform's content.

I would like to encourage people, to set a Yoda Timer, and think about their personal policy, when it comes to this type of algorithmic consumption; that is, a highly addictive app, that can-presumably-generate content tailored to very niche subsets of people.

My thoughts (read after your Yoda timer):

I think it is likely quite a dangerous thing to try once, and plan to avoid even taking a peek at an app like this. Much the same way I don't take a little hit of fent, just to see what it's like.

I wrote more about this-in a fiction

... (read more)

Off the cuff thought: I think IABI being kinda simplistic is both fine, and probably beneficial. I have not yet received my copy, but from reviews I’ve read, the book consists of the same arguments that originally convinced me to join ‘the fight’ all those years ago. Nuance in my understanding came later. I’d guess that’s generally the same path we all took?

To Make Dating Apps Not Terrible, Make Them Not-for-Profit?

I'm skipping the explanation of why dating apps are bad for their users.

To fix these problems, why wouldn't we make a not-for-profit dating app?

It removes incentives to keep users lonely for as long as possible.
There are rich people who are worried about population decline, who might fund it.
- I'm assuming some governments are probably getting worried too.
An app that legitimately finds people love, is probably a good candidate for warm fuzzy donations, perhaps the general public would be receptive to this?

Questions

Does anyone know of any dating apps (other than Duolicious, which I'll write about in a comment) that have experimented with a not-for-profit, or charity model?
What am I missing? This seems like a big ol' hunk of power, social capital, and utility, just sitting on the ground of central station!

Thank You to People of the Lightcone: an ambient song

From the window of our room, Isabella and I could see the Lightcone staff office. While I wont name anyone directly, it became a joke between Isabella and I, that no matter the time of day/night, if we looked over to those windows, always, there would be a certain member of the Lightcone staff, sitting, working at that computer. Clearly visible, like a busy fish in a lit tank. Outside of that tank, a half party, half conference chugged on.

Today is Isabella and my last day at Lighthaven. I'd like to pay my respects to people like the aforementioned member of the Lightcone... (read more)

GTFO of the Social Internet Before you Can't: The Miro & Yindi Story

keltan

9mo

Recommended music to read this to (If you like ambience)

I

Yindi had sent him a link, "You've gotta see how this guy speedruns Mario Kart, I think you'll like it (✿◠‿◠)". Miro taps the link.

CREATE A NetMe™^[1] ACCOUNT TO WATCH THIS VIDEO

Miro creates the account. The video is good.

He runs to boot his CRT, its electron beam lighting the quiet room with a high pitched scream. He starts his Wii and runs through Mushroom Gorge a few times before trying to replicate the technique. "Aim in between the two coins, shroom before the grass then, release your med turbo and left hop at the same time"

On his first attempt he gets close to making... (read 3019 more words →)

•••

I have been working on a post in which I propose that the window for establishing healthy media consumption habits may be closing as content generation technologies advance, and make the suggestion that you GTFO before it's too late.

I'm looking for push back on this thesis statement to try and strengthen my argument. All comments welcome.

I feel a deep love and appreciation for this place, and the people who inhabit it.

If you'd like an explanation from Claude that starts quite basic and builds up, I have had great success with the following phrase as a "Style" in app, no changes necessary.

"Wait, start from the baseline and work your way up to the explanation"

The Compliment Sandwich 🥪 aka: How to criticize a normie without making them upset.

keltan

Note. The comments on this post contain excellent discussion that you’ll want to read if you plan to use this technique. I hadn’t realised how widespread the idea was.

This valuable nugget was given to me by an individual working in advertising. At the time, I was 16, posting on my local subreddit, hoping to find someone who could advise me on a film making career path. This individual kindly took the time to sit me down at a bar—as I wore my school uniform—and detail everything I would need to do to be able to make films professionally. Among many insights I am truly grateful for was the Sandwich. As with many... (read more)

Perhaps a silly question, but does the recent "Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs" paper, imply that people calling a model good or bad online results in a self fulfilling prophecy?

e.g.

Bob Says "Alice.ai is bad"
Alice.ai is trained on this data
The next iteration of Alice.ai will think of itself as worse than if Bob had never made that comment. This results in Alice.ai creating bad outputs
Those bad outputs push Charlie over a threshold and Charlie says "Alice.ai is bad"
Loop

Edit: Oops, I didn't realize Alice.ai was a real site. Though it's got a pretty art style, so I'll keep it in here.

Where should one post to get into the training data?

keltan

There's been some talk about “writing for the ai”, aka: Writing out your thoughts and beliefs to make sure they end up in the training data.

LessWrong seems like an obvious place that will be scraped. I expect when I post things here, they’ll be eaten by the Shoggoth.

But what about things that don’t belong on LW?

I want to maximise the chances that all AIs being built will include my data. So posting to Twitter (X) seems like I’ll just be training Grok???

What about a personal blog I start on a website I own? Does making the robots.txt file say “everything here is available for scraping” increase the chances? Does linking to that website in more places increase the chances?

I feel like I’m lacking a lot of knowledge here. I encourage responses even if they feel like obvious things to you.

Does Claude Prioritize Some Prompt Input Channels Over Others?

keltan

Epistemic Status: I am not an AI researcher, I'm a Psych student. Strong confidence in data collection, moderate confidence in some interpretations, quite weak in others. Based on 60 chat sessions across 6 experimental setup conditions. Data was so strong I decided 10 chats per-condition would suffice.

This research is an attempt to build my skills as a scientist, and to add to the budding field of AI Psychology^[1]. Since this is psychology, I'm mostly pointing at something here. I have nothing more than speculation when talking about the underlying mechanisms.

TL;DR: Claude has a strong 'preference' for the Project Prompt in it's app. It's most likely to take instruction from it, even over the... (read 1272 more words →)

I Finally Worked Through Bayes' Theorem (Personal Achievement)

keltan

Two years ago I found this community, which prompted me to start self teaching math. For reference, I didn't know what a fraction was in early 2022. I knew what they looked like, and what they were called. I didn't know what they meant.

The story of why I lacked basic math skills is complex enough for its own post. But my motivation to learn was simple. TSUYOKU NARITAI!

My goal was to understand Bayes' Theorem. And today, I have done that for the first time. Today, I achieved that milestone. This post chronicles the final reasoning steps that got me there.

I don't expect the average LWer will gain value from this post.

It is... (read 2532 more words →)

Is there a place to find the most cited LW articles of all time?

keltan

I expect it would be useful when developing an understanding of the language used on LW.

keltan's Shortform

keltan

This is a special post for quick takes (aka "shortform"). Only the owner can create top-level comments.

AI Generated Music as a Method of Installing Essential Rationalist Skills

keltan

On April Fools the LW team released an album under the name of the Fooming Shoggoths. Ever since the amount that I think about rationality has skyrocketed.

That's because I've been listening exclusively to it when I'd usually be listening to other music. Especially Thought That Faster (feat. Eliezer Yudkowsky). I now find that when I come to a problem's conclusion I often do look back and think, "how could I have thought that faster?"

So, I've started attempting to add to the rationalist musical cannon using Udio. Here are two attempts I think turned out well. I especially like the first one.

When I hear phrases from a song in everyday life I complete... (read 181 more words →)

Don't Think About the Thing Behind the Curtain.

keltan

Foreword

This is a submissions to the rational animation writing contest this script is Halloween themed and inspired by their most recent video (Please imagine Yann LeCun in his Pjs/HJ). I wanted to get it out ASAP.

I have left out any types of directions for animation, I wasn't sure if it was expected.

Thank you to Rational Animations for leaving the gates open for scripts. I’m quite new to LW and rationality. But I’m excited to have this opportunity.

Any and all feedback on the script, it's clarity, the moral, corrections, my writing quality are welcomed. Please don't feel like you need to be nice or hold back with your criticisms.

Recommended listening:

Soundtrack from the movie... (read 1479 more words →)