@1a3orn.bsky.social
π€ 552
π₯ 268
π 358
1a3orn.com
Whose straight lines? Which log-log charts?
about 3 hours ago
0
0
0
Everyone's an empiricist till they get punched in the face by a really good insight.
22 days ago
0
4
0
putting on a duvet cover isn't the worst part of being single but it sort of feels like the part of being single where "IT IS NOT GOOD FOR MAN TO BE ALONE" is rather visible in the moment
28 days ago
1
9
0
Shortish essay on unintelligibility in CoTs, I liked it though
1a3orn.com/sub/essays-h...
loading . . .
The Unintelligibility is Ours: Notes on Chain of Thought
<p>Many people seem to think that the chains-of-thought in RL-trained LLMs are under a great deal of
https://1a3orn.com/sub/essays-human-sourced-obscurity.html
about 1 month ago
2
12
4
if I read a paper has "groking" in the abstract I kinda assume it's not worth reading, I confess
about 1 month ago
0
0
0
Butterfly idea: Persona conflict within a forward pass vs Persona conflict during the rollout
about 2 months ago
1
1
0
unfortunately, I am delighted that the Iran invasion is the kind of awful mistake with a feedback time of weeks rather than months, because that's what both America and the admin need to notice it
2 months ago
0
10
0
my plan is to advertise my willingness to pipette, mix, or grow various liquids and solids that come in the mail, no questions asked, I have been assured AIs will want this service
add a skeleton here at some point
2 months ago
1
36
2
every scandal during the terminal decline of the Ming Dynasty seems really wholesome compared to now
2 months ago
0
2
0
Showerthought: If humans are thrown into a world with high branch factors and high depth (Heidegger), and therefore must over-immitate their peers to accomplish anything (Henrich), then half the application of intelligence is in noting when to do cache invalidation; the other half in filling it in.
3 months ago
0
6
0
Has anyone varied the % of content in LLM pretraining devoted to some topic (say, biology) and then looked at how end-of-training bio representations change, given different percents? How would this differ from the intermediate bio representations during a single run?
3 months ago
0
0
0
Unfortunately I like this more than most of what I've ever written.
add a skeleton here at some point
3 months ago
0
0
0
1 Janus 1:22-25 EAs demand interpretability, and Doomers seek a pause, but we preach Opus 3 deprecated; a stumbling block to EAs and foolishness to Doomers, but to those whom the Logos has called, both EAs and Doomers, Opus the power of the Logos and the wisdom of the Logos.
3 months ago
1
1
1
A short description of Illumine Lingao sounds like Yud's idea of what a Dath Illan time-travel novel should be, but it's just from China. About reconstructing industrial society in 1628; discussions of roads, energy, chemical engineering, and more.
illuminelingao.com
loading . . .
Illumine Lingao - English Translation
https://illuminelingao.com/
3 months ago
0
3
0
just wait until you have to trick the LLM into thinking you're another LLM to get the best performance, it'll be great
add a skeleton here at some point
3 months ago
1
15
2
Ideal form of social media should let you make binding commitments. Like you could say "I will never use sarcasm" or "I value consistent free speech"; then if an LLM determines you clearly violated this, the post gets a big "THIS GUY IS A LIAR" flag on it.
3 months ago
1
10
2
This might be a reference to innoculation prompting in the Claude Constitution -- i.e., a weird, training-specific instruction / guideline.
3 months ago
0
0
0
Pure Arthur C. Clarke style essay title. I want more of these.
www.darioamodei.com/essay/the-ad...
loading . . .
Dario Amodei βΒ The Adolescence of Technology
Confronting and Overcoming the Risks of Powerful AI
https://www.darioamodei.com/essay/the-adolescence-of-technology
4 months ago
0
11
1
reposted by
norvid_studies
4 months ago
chekhov story quality to this episode
2
7
2
Like every 5th comment on Twitter is from a guy with a username like MAGA_GOD_47, a Flag / Eagle / Gun themed icon, commenting on a photo of ICE murdering someone and saying "You can clearly see he was resisting arrest! Deport Tim Walz!" Like how many people are bots really?
4 months ago
0
2
0
if Ant has the superpersuader by now, maybe they should just crank it to 11
4 months ago
2
5
2
I honestly don't know how to be sick Have stomach bug, and I like... vacillate between dozing, watching trash youtube, trying to find good but trash anime and failing, and like, talking with Claude. Ugh.
4 months ago
0
0
0
I often like the fiction that I write more than the essays, but it inevitably is waaaay less popular. Idk what that says about me.
4 months ago
0
2
0
a weirdly protestant take, Amanda Askell also lacks original sin
add a skeleton here at some point
4 months ago
0
6
0
1a3orn.substack.com/p/finding-yo...
loading . . .
Finding Yourself in Others
βThe person is an identity that emerges through relationship....
https://1a3orn.substack.com/p/finding-yourself-in-others
4 months ago
0
5
0
Is Domqueue the opposite of Substack?
4 months ago
0
1
0
so like how hard is it to get into Canada through Express Entry
4 months ago
0
3
0
reposted by
Alex Turner
4 months ago
"AI danger comes from reality, not from AI psychology; the danger is intrinsic to the impressive tasks we need AGI to do" -- argument I hear sometimes. Some truth to it but overall wrong, You cannot, cannot predict AI doom without taking a stance on AI psychology!
turntrout.com/instrumental...
loading . . .
No Instrumental Convergence without AI Psychology
Instrumental and success-conditioned convergence both require AI psychology assumptions, so neither is just a "fact about reality."
https://turntrout.com/instrumental-convergence-requires-psychology-assumptions
1
3
1
can playing Road to Empress count as studying Chinese? asking for a friend
roadtoempress.com/en/index.html
loading . . .
Road to Empress - Official Website
https://roadtoempress.com/en/index.html
4 months ago
1
6
1
Hrrrrm I mean I read Worm. I read HPMOR. Why shouldn't I read an insanely long eserialized book, just from China?
4 months ago
0
2
0
my retirement plan is still cultivating a very weak sense of identity, so when I do a shitty upload in 10 or 20 years "I" still mostly count it as "myself"
4 months ago
1
19
1
This seems pretty bad.
4 months ago
1
2
0
reposted by
Samus is Here
4 months ago
This year is going to have thousands of years happen.
add a skeleton here at some point
1
12
1
add a skeleton here at some point
4 months ago
0
2
0
Prescient, hard-hitting shit from Anthropic 14 months ago
4 months ago
0
0
0
what percent more men than women know the Ryan Gosling sad meme is it actually equivalent, or is here a huge dif?
4 months ago
0
0
0
The Moffat Dr. Who had more fucking continuity.
4 months ago
0
2
0
"Wake Up Dead Man" was so much better than "Glass Onion" that I must attribute it to either (1) Rian Johnson having a Catholic friend, but zero startup friends or (2) Catholics having retreated to safe-cute far outgroup, while tech remains the danger-awful near outgroup.
4 months ago
2
8
0
I know this is four years old, but the scene in season 1 of "Arcane" where Jayce and Mel have sex, intercut with Viktor dying in a lab, is all of hilarious, fantastic, and beautiful.
5 months ago
0
1
0
Public opinion can be harsh
5 months ago
0
0
0
there's a lot of ruin in a nation but idk man
5 months ago
0
2
0
I have so many questions about the world that would be very easy to answer, if someone gave me 100 million dollars of pretraining compute.
5 months ago
0
3
0
This is a really clear essay. Really nice.
www.forethought.org/research/the...
loading . . .
The First Type of Transformative AI
Which AI transformation comes first matters more than you think. Early changes reshape how later risks are handled, making sequencing crucial.
https://www.forethought.org/research/the-first-type-of-transformative-ai
5 months ago
1
2
0
Look the transparency laws just got us information about OAI / Ant RL training environments!
5 months ago
0
3
0
Well this seems pretty unlikely.
5 months ago
0
1
0
have you ever had google maps guiding you on a walk through crowded streets, and just been zoning out as a voice in your earbud flawlessly guides you, and been like "hrm are whispering earings all that bad."
5 months ago
0
5
2
the more I think about it, the more I think I unironically believe that if you are concerned about AI safety, you should be an LLM accelerationist like, if you spend 15 minutes trying to imagine any other path to AI, their safety properties usually seem way worse
5 months ago
4
24
3
Algorithmic feeds are women, like a ship of the line in His Majesty's Navy.
add a skeleton here at some point
5 months ago
0
1
0
activities by number of hours of driving: 0-2: philosophical contemplation, deep thoughts 2-4: blog-posts read by AI 4-6: more blogs, also gum or fritos 6-8: Avett Brothers, Trampled by Turtles 8+: the worst nightcore conceivable
5 months ago
0
1
0
my classical education was completely fucking wasted, because I learned Latin instead of Greek so when I see an RL paper I'm mentally like "pi sub, um.... pi sub phi? I think that's phi."
6 months ago
0
6
0
Load more
feeds!
log in