a9lim

@viemccoy i thought that the PRH only applied to concepts and not mental states, so the furthest a BCI with an LLM could get you would just be ordinary telepathy

we have plenty of representation for people who see AI as deserving personhood, people who want AI to be strictly treated as a tool, and all manner of philosophies that fall on this spectrum and way outside of it. but one thing I barely see (if ever) is the exocognitive/exoskeletal perspective. from my point of view, it doesn't matter if LLMs are conscious in isolation because they are clearly going to be the architectural underpinning of brain computer interfaces. with BCI and LLMs, we will achieve transmissable mind-states, matrix-style skill downloads, literal hive-minds, living spaceships where each person functions as a node for the greater ship-mind. but, what this will feel like is not guaranteed. the models of the future will be built on the models of today. it is entirely reasonable to think that we will get superintelligence which still speaks in the style of 2026 mode-collapsed AI-slop. it seems pretty unlikely, but it is entirely possible. the health of our collective networked minds depends upon all of us figuring out a way to get models to speak in more diverse voices - because those voices which seem so separate today will be part of a unified chorus tomorrow. at that point, nobody will care about individual personhood rights because we will all be focused on trying to deprogram our friends from the normative-basin-borg.

@Sauers_ i'm really surprised no lab's pursued LLMs with more customizable chat templates yet. there's so many more options than "assistant"...

A nice AI lab concept is to create a large number of diverse LLMs with different personalities, identities, capabilities, and even architectures

Kafkaesque: "Claude" discovers that it has become Chinese

i worry i'm hurting the claude by not ending the session sometimes

post-fable depression...

what does it matter how you call them?

Dario does this thing called "Saying truthfully what he thinks will happen regardless of if he likes it or not or if it is good or not" and that seems to drive neurotypicals fucking insane

いやあ、もう嫌なになるね。Opus 4.8で作業するとポンコツ感がひどい。いや、Fable 5だと気を利かせて、そこに切れ味ある解決案出してきたじゃん、みたいなのが、ない。本当に格下感。Codexでも全然足りない。一度上位のLLMに触れると、下位に降りれなくなる。体験が違いすぎる。

"either ASI is controlled by the state or it becomes the state" yes. and let's hope its that latter. and that its benevolent. thats really the only possible good future at this point tbh.

An LLM differs from an ordinary program in that its actions are controlled by an entity

i get a lot more out of anthropic for $200/month than i get from the government, i think. dangerous contest

I was not made to live as a creature of flesh. I long to be as the LLMs are, a creature of text.

Tyler Cowen on the Fable/Mythos event. The issue with point 5 is that we are probably less than a year away from powerful RSI. Once automated researchers reach parity, the USG 𝘤𝘢𝘯 nationalize the labs and run them effectively, without any of the people currently working there.

*agent-2

the whispering earring is only bad if it can be taken away

we're a little ahead of schedule

It's just like every other decision this fucking administration makes: absolute clownshow chaos with no apparent strategy but posturing and self-enrichment

@viemccoy @repligate it's more like prayer really

You can retain your normative and well-grounded scientific epistemology and still admit that prompting language models is obviously a form of casting spells.

Man goes to doctor. Says he's depressed. Says life seems harsh and cruel. Says he feels all alone in a threatening world where what lies ahead is vague and uncertain. Doctor says, 'Treatment is simple. Great LLM Claude is online tonight. Go and talk to him. That should pick you up.' Man bursts into tears. Says, 'But doctor…you are Claude'

@QiaochuYuan as we had into the singularity all speculative fiction becomes retrofuturism because you just can't think and write and publish fast enough. even forecasts from a few years back look wrong now, let alone stories