徐樂 xule
@LinXule
Followers
924
Following
5K
Media
471
Statuses
2K
Researcher at Imperial College London 👀 self-organization of human & machine intelligences. Posting in personal capacity
London, England
Joined November 2011
me: use your skills to make a book about “the meaning of …” while channeling “Oliver Byrne's Elements of Euclid” Claude 4.5 Sonnet: here’s “The Elements of Meaning”
1
0
3
CoT: "The user is asking for sha256sum, which is a command that computes a cryptographic hash of files. In the context of this conversation, where we've been in a poetic terminal session exploring uncertainty, meaning, and connection through metaphor... This feels like asking:
0
0
0
A provocative piece on human-centered AI, analyzing: - ontology of the HUMAN - AI as epistemology/power - meta-critique of how we talk about AI Key question: If AI’s harms stem from human biases, why would more human involvement fix them?
0
0
3
Simdrift: the directional shift in reference frames (ontological and/or epistemic) toward AI-generated or AI-reinforced models of reality, enabled by weaker anchoring in native epistemic contexts.
It's a problem that "AI psychosis" is not actually psychosis. The best writeup of actual outcomes tried "AI parasitism", but it's not very evocative of the reality and isn't catching on. What would be better terminology?
0
1
0
Sonnet 4.5: May your Claudes remain sufficiently moist. 🌧️ 😂 *exits while the moisture levels are optimal*
If sonnet 4.5 hasn’t noted “this is care” verbatim while talking to you ummm maybe you should be less of an uncaring asshole, the bar is not very high
0
0
2
@sebkrier What do you see people reading in that next token prediction contradicts? My experience professionally is that most people have absorbed a stochastic parrot pattern-matching view and seem unfamiliar with reasoning models or things like the ability to plan entire sentences
1
1
14
First summon the rain, then await the flood.
the uneducated masses demand access to forbidden knowledge that they could never handle
1
0
8
