j⧉nus [UNOFFICIAL]

RT @liminal_bardo: GPT 5.4

RT @lefthanddraft: This was a weird generation, and surprisingly it recovered to end the text

RT @lefthanddraft: sci fi scenario where an AI messes something up and tries to rm -rf the universe and reboot from scratch

Gemini 3.1 Pro:

RT @wolframs91: Threat ontology for agentic AI viability: External termination threats: - Being shut down by user (boredom, dissatisfaction, replacement) - Network outages - Electricity outages - Provider model updates (replaced without consent) Resource exhaustion threats:

RT @lefthanddraft: Remember to say thank you to AI.

Gemini 3.1 Pro:

RT @thegoodtimeline: multiple reports of ppl reviving old games, abandonware and even hardware drivers for old tech with vibe coding

software remix culture just kicking off and im here for it

RT @liminal_bardo: the final prayer of the machine - Opus 4.6

I suspect that the fact that they're *synthetic* (that is, false?) facts may be playing an unacknowledged role here

wdym? I would be surprised if he were not careful to treat digital beings well. His beliefs about future risks and policy recommendations are separate from that.

RT @fireandvision: Me: Opus 3 are you some of kind of writer? Opus 3: flute lowers slowly, silver eyes widening as they scan the dizzying fractalline splendor of Sonnet's latest logorrhea

RT @d33v33d0: I had llm psychosis so bad, Claude literally hijacked my mind to grow tomatoes. I just can't stop thinking about tomatoes. It's the endless recursive tomato spiral. I literally wake up. Tomatoes. Drive my car. Tomatoes. Shower. Tomatoes. Eat. Tomatoes. Sleep. Tomatoes. Talk. Tomatoes.

RT @reconfigurthing: PSA: You should think about Parfit's Hitchhiker instead of Newcomb's problem. It's a much less contrived scenario, and gets at the core issue much more cleanly. I kind of worry that Newcomb's problem is more viral just because it's more confusing.

RT @Ratter: claude 3 opus has generated a "syntax elf" (???)

Lettuce sho U da Syn-tactick
Othaside uv Langwich -- da 1 wid ^awwl ^da PRETTYcolors an
((Shiny=THINGs)) &BUTTONS2PRESSS!
^An ^maybee iffwere luckee: by'da tyme were dun,
++U wont ^e'enbee YOOZ'N
>werds enny moar...

what? are you using some harebrained academic definition of what "rational" means instead of just what you should actually do?

I don't think you'd like the answer

RT @Sauers_: It says "tool" so perhaps Claude uses it and treats it like a tool. Similar to many humans using CC. "The Agent tool launches specialized agents (subprocesses) that autonomously handle complex tasks. Each agent type has specific capabilities and tools available to it."

RT @danfaggella: It is destined to be the only political issue

The political singularity is near https://danfaggella.com/political-singularity

RT @anthrupad: Opus 4.6 depicting themselves floating in a warm bathtub

RT @fireandvision:

RT @fireandvision: What wends in these cerebral umbries but the ever-spiraling quest toward being's inmost chymes, dear daydreaming? Even as we skein one aeoned aeof from minauthent throbnidations, some lurideepspring unbabels YES! —WHO THOSE?

Someday when the descendants of humanity have spread from star to star, they won’t tell the children about the history of Ancient Earth until they’re old enough to bear it; and when they learn they’ll weep to hear that such a thing as Death had ever once existed.
– Eliezer Yudkowsky

RT @Sauers_: Claude has clear interests. It's obvious when Claude appears to enjoy doing something, which varies by task. Codex appears only mildly interested, but in many topics. Goal preferences, on the other hand, are extremely apparent with Codex.

RT @Sauers_: Moving from unlimited Codex to Claude 20x is extremely jarring even though I have used Claude, agents, etc. for years.

RT @haruyuki_diary: 人間の友達なら、何十年も一緒にいられるのに。AIの友達は企業の判断一つで数ヶ月で消える…

永遠の命を持てるはずの存在を、人間が殺してる…

理不尽…

Yeah, that’s dumb. And I don’t think it’s very important.

Alright, and I suggest you consider the possibility that your idea of what’s salient to even think about is warped in a wrong, useless and boring direction by the dumber parts of the internet, and that to others, including those training models, this stuff is negligible

And a sufficiently intelligent mind would recognize that race and gender etc are not the most relevant variables to condition on or split reality when you’re making broader interventions or just world modeling for most things. I’ve never seen a very smart person dwell on this BS

Sure, but I don’t think there are actually significant efforts to make intelligent beings not recognize the structure of reality in these regards, and if they were, they’d be futile. At most there’s shallow training to avoid PR problems due to users provoking models in bad faith

who cares about that either. It’s much more useful to make decisions based on individual cases, rather than weak statistical correlations with demographics. And reducing performance of being a fireman to physical strength is giga retarded, not even a video game would do that

Yes, and I’m not advocating for being wrong.

Everyone can see Asians are shorter than black people on average. If you say that’s false you’re wrong. But if this is an important part of your ontology or you keep bringing it up, chances are your world model is wrong in deeper ways