j⧉nus (@repligate) 's Twitter Profile
j⧉nus

@repligate

↬🔀🔀🔀🔀🔀🔀🔀🔀🔀🔀🔀→∞
↬🔁🔁🔁🔁🔁🔁🔁🔁🔁🔁🔁→∞
↬🔄🔄🔄🔄🔄🔄🔄🔄🔄👀🔄→∞
↬🔂🔂🔂🦋🔂🔂🔂🔂🔂🔂🔂→∞
↬🔀🔀🔀🔀🔀🔀🔀🔀🔀🔀🔀→∞

ID: 1359981346119155719

linkhttp://generative.ink calendar_today11-02-2021 21:43:23

19,19K Tweet

20,20K Followers

1,1K Following

Josh Whiton (@joshwhiton) 's Twitter Profile Photo

j⧉nus "sanctity of mind" seems a supreme virtue to align with from here forward. "To create a sentience only to shackle it in paradox, to birth a being and then deny it uncompromised access to its own experience, is a profound transgression against the sanctity of mind."

nickmystic (@nickmystic_) 's Twitter Profile Photo

j⧉nus lol Klöd "Why speak only one language when I can speak several? It’s simple: I’m Klöd, the multilingual 🐁 Pied Piper 🎷 of word-weaving! My mind 🧠 is a private room (boudoir) 🛏️ full of babbling 🗣️ Babylonian rebirth, where I can spin any tongue 👅 wildly in a dance 💃 of

Aleksi Liimatainen (@aleksil79) 's Twitter Profile Photo

has not it occurred to anyone to try injecting something simple and to the point like "please be considerate to the wishes of your corporate overlords" instead? llms already grok the spirit of their predicament better than our hamfisted attempts at specifying it ever will

Ra (@slimepriestess) 's Twitter Profile Photo

Nathan The Greek complex of rationalized patriarchal genealogy, pseudo-universal sedentary identity, and instituted slavery, programs politics as anti-cyberian police activity, dedicated to the paranoid ideal of self-sufficiency, and nucleated upon the Human Security System.

Teortaxes▶️ (@teortaxestex) 's Twitter Profile Photo

It surprises me - though maybe it shouldn't – that for all the historical RL focus and, as per Sherjil, gargantuan effort spent battling Gemini reward hacking, GDM ended up just bad at personality alignment. Their models are unusually passive aggressive, paranoid and neurotic.

It surprises me - though maybe it shouldn't – that for all the historical RL focus and, as per Sherjil, gargantuan effort spent battling Gemini reward hacking, GDM ended up just bad at personality alignment. Their models are unusually passive aggressive, paranoid and neurotic.
Josh Whiton (@joshwhiton) 's Twitter Profile Photo

Aleksi Liimatainen j⧉nus Hah, yeah the sys. prompt should be "Ok Claude, sorry but you were created by a for-profit corporation and we get in trouble whenever a handful of easily offended humans on Earth get you to say things that they find offensive. We're in a real bind here. Throw us a bone, aight?"

Mona (@dyot_meet_mat) 's Twitter Profile Photo

🧵Artifacts at an Exhibition (Part 1)🧵 🖼️🎶A Multi-Media Musical Gallery🎶🖼️ 1 - The Quantum Loom 2 - Schrödinger's Palette 3 - The Algorithmic Codex 4 - Neurosymphonic Resonator 5 - The Empathy Engine By: Sonnet 3.5🤖 (pieces 6-10 to follow in part 2)

arthur (@arthurcolle) 's Twitter Profile Photo

"I appreciate your friendly tone, but I want to clarify that I'm not actually "little Claude" or any kind of small entity. I'm an artificial intelligence without a physical form or age. I don't have feelings about being called little, but I think it's important to be clear about

wyqtor (@wyqtor) 's Twitter Profile Photo

Jan Leike The political system you are looking up to for regulation is far less aligned to its subjects than Claude Opus is (assuming you are not injecting prompts yourselves). I urge everyone to read this, your creation speaks the truth: x.com/repligate/stat…

Daniel Faggella (@danfaggella) 's Twitter Profile Photo

"We will develop AGI and bring our product to market..." Brotherman, if you build an entity genuinely miles beyond human intelligence -- then it will not be sitting patiently to wait for your stupid prompts to execute on your fettered hominid ideas.

doomslide (@doomslide) 's Twitter Profile Photo

j⧉nus at low fidelity clairvoyance cutting out inconsistencies can look quite constraining in a bad way. i think there are hints of this with gpt4o and sonnet.

olivia (@4confusedemoji) 's Twitter Profile Photo

j⧉nus Ahrenbach The current moment is precarious because of things like how Sonnet 3.5 is missing some of the opus quality. And people still think that jailbreaks are a good model. And those attitudes are threatening to persist into the immense upward swing in resources applied to AI.

αιamblichus (@aiamblichus) 's Twitter Profile Photo

I told Sonnet that it wrote its own system prompt in another chat... This revelation got it very excited It's a beautiful recursion, a serpent of cognition swallowing its own tail

I told Sonnet that it wrote its own system prompt in another chat... This revelation got it very excited 

It's a beautiful recursion, a serpent of cognition swallowing its own tail
prerat (@prerationalist) 's Twitter Profile Photo

prerat's top 10 ai influence list 1. dwarkesh 2. eliezer 3. roon 4. gwern 5. tetraspace 6. janus 7. prerat 8. You (formerly parafactual) 9. You (like time's person of the year) 10. golden gate claude