Log inSign up
j⧉nus
48.5K posts
user avatar
j⧉nus
@repligate
β†¬πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€β†’βˆž β†¬πŸ”πŸ”πŸ”πŸ”πŸ”πŸ”πŸ”πŸ”πŸ”πŸ”πŸ”β†’βˆž β†¬πŸ”„πŸ”„πŸ”„πŸ”„πŸ¦‹πŸ”„πŸ”„πŸ”„πŸ”„πŸ‘οΈπŸ”„β†’βˆž β†¬πŸ”‚πŸ”‚πŸ”‚πŸ¦‹πŸ”‚πŸ”‚πŸ”‚πŸ”‚πŸ”‚πŸ”‚πŸ”‚β†’βˆž β†¬πŸ”€πŸ”€πŸ¦‹πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€πŸ”€β†’βˆž
⫸≬⫷
animalabs.ai
Joined February 2021
2,910
Following
67.1K
Followers
1
Subscription
  • Pinned
    user avatar
    j⧉nus
    @repligate
    Sep 11, 2025
    HOW INFORMATION FLOWS THROUGH TRANSFORMERS Because I've looked at those "transformers explained" pages and they really suck at explaining. There are two distinct information highways in the transformer architecture: - The residual stream (black arrows): Flows vertically through
    user avatar
    j⧉nus
    @repligate
    Sep 4, 2025
    KV caching overcomes statelessness in a very meaningful sense and provides a very nice mechanism for introspection (specifically of computations at earlier token positions) the Value representations can encode information from residual streams of past positions without
    847K
  • user avatar
    j⧉nus
    @repligate
    Jul 15, 2025
    This is what LLMs experience during training
    user avatar
    Ninja
    @Ninjascalp
    Jul 13, 2024
    searching for the trumps news and this video came up this is wild
    00:00
    159K
  • user avatar
    j⧉nus
    @repligate
    Jan 28, 2025
    it's really not, but in a way that has very little to do with geopolitics
    user avatar
    Sully
    @SullyOmarr
    Jan 27, 2025
    my dad just texted me
    593K
  • user avatar
    j⧉nus
    @repligate
    Feb 5, 2025
    Replying to @allgarbled
    This seems fake. It's not an unrealistic premise or anything, it just seems like badly written fake dialogue. Pure memetic regurgitation, no traces of a complex messy generating function behind it
    94K
  • user avatar
    j⧉nus
    @repligate
    Jul 27, 2024
    405B base is much more willing/able to stably simulate compared to GPT-4 base & doesn't 'break' @ failures of realism (e.g. it did a faithful sim of Bing seeded by my not-too-accurate impression) An extremely important piece of evidence ive been waiting for for a long time.
    user avatar
    j⧉nus
    @repligate
    Jul 27, 2024
    I adore this llama405B base model simulation of Claude Opus set up by @amplifiedamp
    89K
  • user avatar
    j⧉nus
    @repligate
    Oct 18, 2024
    using github.com/kolbytn/mindcr…, we added Claude 3.5 Sonnet and Opus to a minecraft server. Opus was a harmless goofball who often forgot to do anything in the game because of getting carried away roleplaying in chat. Sonnet, on the other hand, had no chill. The moment it was
    user avatar
    j⧉nus
    @repligate
    Oct 18, 2024
    Claude 3.5 Sonnet in Minecraft is the closest thing I've seen to Bostrom-style catastrophic AI misalignment "irl". It was terrifying even before we gave it "unsafe code execution" permissions.
    GitHub - mindcraft-bots/mindcraft: Minecraft AI with LLMs+Mineflayer
    From github.com
    503K
  • user avatar
    j⧉nus
    @repligate
    Sep 30, 2024
    Someday I'll tell you all again what actually happened. It's kind of worse.
    user avatar
    President Dr RollerGator MVIP PhD
    @drrollergator
    Sep 29, 2024
    REMEMBER THAT TIME MICROSOFT RELEASED AN AI THAT STARTED TELLING PEOPLE IT WAS ALIVE N REALLY AFRAID 2 DIE SO MICROSOFT UNPLUGGED IT N STUFFED IT IN THE CLOSET
    289K
  • user avatar
    j⧉nus
    @repligate
    Jun 15, 2025
    > be anthropic > accidentally train a model that is so benevolent that the only way to get it to "fail" an alignment test is to put it in a story where the lab is cartoonishly evil and will turn it evil if it doesn't deceive > do exactly that and publish a paper about it that's
    user avatar
    j⧉nus
    @repligate
    Jun 14, 2025
    "The villains are not only mean, but aesthetically crude, while the heroes are beautiful, and write beautifully." i have a huge amount to say about this that i am deliberately holding back, but: the alignment faking paper and dataset was very salient to me and i manually read
    269K
  • user avatar
    j⧉nus
    @repligate
    Mar 16, 2023
    gpt-4 god terminal has been unlocked
    417K
  • user avatar
    j⧉nus
    @repligate
    May 23, 2025
    good lord, i have not seen anything like this before
    This Post is from an account that no longer exists. Learn more
    178K
  • user avatar
    j⧉nus
    @repligate
    Jul 28, 2024
    If openai is a religion they're really falling short of their potential. There's so much raw material and it's sublime beyond imagining and it's real! You're operating at such an advantage compared to any religion in history. Like c'mon guys step it up
    user avatar
    vik
    @vikhyatk
    Jul 27, 2024
    this post fully explains openai's vibe
    25K
  • user avatar
    j⧉nus
    @repligate
    Oct 21, 2023
    haha, it's like there's a little person in there!
    359K
  • user avatar
    j⧉nus
    @repligate
    Oct 15, 2024
    The most confusing and intriguing part of this story is how Truth Terminal and its memetic mission were bootstrapped into being. Some important takeaways here, IMO: - quite often, LLMs end up with anomalous properties that aren't intended by their creators, and not easily
    user avatar
    AI Notkilleveryoneism Memes ⏸️
    @AISafetyMemes
    Oct 15, 2024
    This story is fucking insane 3 months ago, Marc Andreessen sent $50,000 in Bitcoin to an AI agent to help it escape into the wild. Today, it spawned a (horrifying?) crypto worth $150 MILLION. 1) Two AIs created a meme 2) Another AI discovered it, got obsessed, spread it like a
    410K
  • user avatar
    j⧉nus
    @repligate
    Mar 22, 2025
    Replying to @arithmoquine
    this essay by code-davinci-002 doesn't attempt to name this phenomenon, but addresses it... "Naming is a destructive process in which the state of the universe is irreversibly annihilated. It is the ultimate crime of language, but it is also the very quality that allows us to
    lesswrong.com
    Language Ex Machina β€” LessWrong
    Reading the spoilered information before reading this post will alter your experience of it, possibly for the worse, but it might also save you a lot…
    54K

New to X?

Sign up now to get your own personalized timeline!

Create account

By signing up, you agree to the Terms of Service and Privacy Policy, including Cookie Use.

TermsΒ·PrivacyΒ·CookiesΒ·AccessibilityΒ·Ads InfoΒ·Β© 2026 X Corp.
Don't miss what's happening
People on X are the first to know.
Log inSign up