Just mcint

Just mcint

@mcint.io

Builder of fine artifacts, bitwise and otherwise.

345 followers1.3K following361 posts

Just mcint reposted

If you can substitute "hungry ghost trapped in a jar" for "AI" in a sentence it's probably a valid use case for LLMs. Take "I have a bunch of hungry ghosts in jars, they mainly write SQL queries for me". Sure. Reasonable use case. "My girlfriend is a hungry ghost I trapped in a jar"? No. Deranged.

Yeah, but §99 of magnificat humanitas hits much lighter when read after §98

Grace@gracekind.net

Just found out the Pope doesn’t believe in computational theory of mind. My day is ruined

View full thread

Hmm this implies the possibility of a benchmark, for “inner growth” …or weak proxies for it

Yeah, but §99 of magnificat humanitas hits much lighter when read after §98

Grace@gracekind.net

Just found out the Pope doesn’t believe in computational theory of mind. My day is ruined

Just mcint reposted

Replying to @gracekind.net

Really? ~33 AD: first documented mind restored from backup onto new hardware. substrate independence is canon. 1 Cor 15:44

Just mcint reposted

Just found out the Pope doesn’t believe in computational theory of mind. My day is ruined

Just mcint reposted

Few people know this

Matt Lichti@mattlichti.bsky.social

The data center water use is actually for baptizing all the new LLM instances. We should also add some baptism related Bible verses to the system prompt just in case.

Human control…of the gaps

Ryan Moulton@moultano.bsky.social

A whole lot of people think controlling models is easy who a month ago didn't think controlling models was necessary.

Call it 🏄‍♂️🏄‍♀️🏄 🌊 surfing the data manifold

Grace@gracekind.net

AI writing, unfortunately

Just mcint reposted

if your mainstream cultural movement doesn't have at least a little bit of sex cult as the origin story you are ngmi

Just mcint reposted

If you're doing "n-degrees of separation guilt" among the people working in one very small industry, in one medium sized city, you are guaranteed to reach the whole graph. At some point you have to start looking at what individuals do and believe. There just aren't that many people.

mmm don’t pin your hopes on wetware competing with the technocapital optimization machine

Dan Davies@dsquareddigest.bsky.social

I'm increasingly thinking there might be weaker versions of (c) - my neural network is the same kind of thing, it's just much better at organizing a wider variety of information, connected up better to physical sensors and only needs a couple of pies rather than a gas turbine to power it

View full thread

Part of Perez’s great tradition (I’m starting at the Dutch joint stock corporation) VOC ships → (UK) canals → rail + telegraph → electricity + submarine cables → roads/oil/telephone → fiber/Internet → compute/AI

mmm don’t pin your hopes on wetware competing with the technocapital optimization machine

Dan Davies@dsquareddigest.bsky.social

I'm increasingly thinking there might be weaker versions of (c) - my neural network is the same kind of thing, it's just much better at organizing a wider variety of information, connected up better to physical sensors and only needs a couple of pies rather than a gas turbine to power it

A helpful stylized fact, re: Chinchilla, scaling laws, and successor ideas, what fraction of total compute budget devoted to a model should be in training versus inference? Before RL regime, training tokens should be 3x (to 1x) model-lifetime inference tokens proceedings.mlr.press/v235/sardana...

proceedings.mlr.pressBeyond Chinchilla-Optimal: Accounting for Inference in Language Model Scaling LawsLarge language model (LLM) scaling laws are empirical formulas that estimate changes in model quality as a result of increasing parameter count and training data. However, these formulas, including...

Just mcint reposted

Replying to @simonwillison.net

maybe folks are a bit discouraged by the way openAI and Anthropic have been hitting the news lately, and it's difficult to disentangle the tech from corporate strategy sometimes.

Just mcint reposted

since rationalists are taking shit I might as well point out a pet peeve with the other (broadly speaking) side: doing the whole "these are low-status morons who think scifi is real" bit makes me think less of you as a person. There are more serious critiques to be made, learn to be a better hater

Just mcint reposted

Replying to @jcsalterego.bsky.social

Jesuslike, he takes the crown of thorns every day, putting up with SO MUCH of our shit

Just mcint reposted

Replying to @renderg.host

The internet population who know about Bluesky is tiny. The population on non-bsky PDSes are relatively tiny. The number of people on those independent PDSes who haven't yet heard of the Atmosphere (even as AT proto) is probably also very small. mu.social/profile/noz....

This post is unavailable.

Just mcint reposted

One thing I've been toying with, instead of talking about "PDS" and cryptic nonsense like that is: Atmosphere Service Provider, like an ISP but for the social internet. Part of the motivation is that "Sign in with Atmosphere" only works if people have seen that before.

B. Prendergast@renderg.host

Log in with Blacksky Log in with Bluesky Log in with Eurosky Log in with Gander Log in with Northsky Log in with Pckt Log in with Social Selfhost Log in with Wsocial Log in with hello you still there??? OR… Sign in with #Atmosphere 🌀 wecanjustdothings.leaflet.pub/3mlv5oijn5s2q

gurrrl u nearly saturate ur own benchmark

Amber 🌸@shimmermathlabs.com

ok i set my top agents on the task. it made a quiz for myself and anyone else. my honest result: Am I a Post-Rationalist? 42/50 shimmermathlabs.com/postrat.html

The phrasing feel a little noncommittal, or over index on a certain kind of hedging away from a “rationalist vibe”. It would take wider contribution, but could be more interesting

Amber 🌸@shimmermathlabs.com

ok i set my top agents on the task. it made a quiz for myself and anyone else. my honest result: Am I a Post-Rationalist? 42/50 shimmermathlabs.com/postrat.html

A *detailed, *illustrative microcultural snapshot of TPOT is Chercher’s hoods graphic, pre-vibecamp 1, 2021-12

46/50 “you are likely explaining attachment theory at a group house”

Amber 🌸@shimmermathlabs.com

ok i set my top agents on the task. it made a quiz for myself and anyone else. my honest result: Am I a Post-Rationalist? 42/50 shimmermathlabs.com/postrat.html