Robin Goins

2.5K posts

Robin Goins banner
Robin Goins

Robin Goins

@coponder

Applied Phenomenology & Metacognition | also, events @ https://t.co/QhVsBFg8iI

Bay Area, CA Se unió Mart 2016
392 Siguiendo410 Seguidores
Shadi B.
Shadi B.@InuitKodiak·
Is anyone interested in an X group for intellectually serious adults—professors, researchers, writers, professionals, independent scholars, etc? No politics, just a circle of rhinkers interested in each other’s posts. All interactions guided by the highest courtesy.
English
1.4K
291
4.5K
210.4K
Robin Goins
Robin Goins@coponder·
@leedsharkey Would you have any interest in giving a talk next month at Mox (moxsf.com) on VPD? (Timing flexible though.) (Already coordinating with Thomas Fel re: a talk on neural geometry, would love to host the whole Goodfire lineup. :) ) Happy to DM w/ more details.
English
1
0
0
31
Lee Sharkey
Lee Sharkey@leedsharkey·
My team at @GoodfireAI has been cooking up a new way to do interpretability: decompose a language model’s weights, not its activations. Our decomposition natively handles attention (!) and behaves less like a lookup table and more like a generalizing algorithm. (1/6)
English
34
192
1.5K
241.7K
Robin Goins
Robin Goins@coponder·
@kageroumado @cube_flipper I'm with cubeflipper here, 1/8th of a tab is (always) extremely noticeable to me. I have been described as having high metacognition, tho.
English
1
0
2
36
Kagerou
Kagerou@kageroumado·
You can try to have placebo on some days and the real one on others, then see the statistics. This could be genuinely interesting when done by someone with high metacognition, maybe the lab studies were simply wrong, or there’s something about individual differences we’re missing. Personally, I didn’t feel anything from 20 mcg daily at all.
English
2
0
1
70
Robin Goins
Robin Goins@coponder·
@etirabys Oh… the way the UI presented the options made me think you were only asking about black/white (which were the only ones I could see). Would’ve picked Navy actually!
English
0
0
1
40
Mingyu_Jin19
Mingyu_Jin19@fnruji316625·
A really interesting paper on representation geometry in LLMs written by my friend @frankniujc : “Hypothesis-Driven Feature Manifold Analysis in LLMs via SMDS” proposes a model-agnostic way to test geometric hypotheses about latent representations instead of assuming everything is just linear directions. They find that different concepts naturally form different structures like circles, lines, clusters, and that these manifolds remain surprisingly stable across model families/sizes while also dynamically reshaping with context. Very cool bridge between mechanistic interpretability and representation geometry. 🔥 Especially liked the framing that reasoning may operate over structured manifolds rather than isolated features. Paper: openreview.net/pdf?id=vCKZ40Y… Code: github.com/UKPLab/tmlr202… #LLM #MechanisticInterpretability #AIResearch #RepresentationLearning #TMLR #Interpretability #DeepLearning
Mingyu_Jin19 tweet mediaMingyu_Jin19 tweet mediaMingyu_Jin19 tweet mediaMingyu_Jin19 tweet media
English
6
36
220
21.4K
Robin Goins
Robin Goins@coponder·
@frances__lorenz Recommend also doing something like Function Health or Vitals Vault to check for any secret vitamin deficiencies that may be upstream of your being a depressed person
English
0
0
1
30
Frances Lorenz
Frances Lorenz@frances__lorenz·
hi girlies and cutie pies I'm trying to achieve mental wellness and connect with nature!!! Wish me luck!!!! My goal is to be the type of person who doesn't understand depression and when people bring it up I'm just kind of like, "yeah uh, I don't really get that but good luck!"
Frances Lorenz tweet mediaFrances Lorenz tweet mediaFrances Lorenz tweet mediaFrances Lorenz tweet media
English
11
0
108
2.2K
Thomas Fel
Thomas Fel@thomas_fel_·
Happy to share my first post since joining Goodfire. Neural geometry has been my obsession for years, and our team here is building a really serious research agenda around it. I can't wait to share the series of papers coming over the next few weeks... Brace for shapes 🍩
Goodfire@GoodfireAI

Neural networks might speak English, but they think in shapes. Understanding their rich *neural geometry* is key to understanding how they work – and to debugging and controlling them with precision. Starting today, we’re releasing a series of posts on this research agenda. 🧵

English
55
71
1.1K
77.2K
rosey🌹
rosey🌹@thechosenberg·
😭😭😭
rosey🌹 tweet media
QME
7
1
340
21.1K
Robin Goins
Robin Goins@coponder·
@etirabys Mox’s day pass is 9am-11pm (but no one will kick you out at 11, just the custom door code will not work anymore if you leave)
English
0
0
1
45
bayesian asian (42/50 paintings)
sth I crave as a parent is a day pass to a coworking space that gives 24h of consecutive access. my most free time for creative work is weekends, 12-4pm and 8pm-12am there's one 24/7 space near me, but the 24/7 is only for full members. day pass gets you 9am-5pm weekdays
English
5
0
23
5.8K
Robin Goins
Robin Goins@coponder·
@2ang7 @ryxcommar This is the kind of answer I'd anticipate from someone who's primarily encountered the word in the context of purchasing/spending money on tokens, as opposed to someone who understands tokenizers or how LLMs work.
English
0
0
0
14
s
s@2ang7·
@ryxcommar ...a unit of compute?
English
3
0
1
508
Senior PowerPoint Engineer
I've asked a couple candidates in interviews "what is a token" and I'm a little surprised how some people who have AI on their resume or in their title can't give a good answer. Like one person couldn't even give me a "it's a word" normie answer.
snoopy jpg@snoopy_dot_jpg

just interviewed a guy that has been working on llama4 evals at meta for the past year. i am not entirely convinced he understood what a token was. wild things happening over there

English
36
1
520
75.8K
Aaron Bergman 🔍
Aaron Bergman 🔍@AaronBergman18·
Drunk for the first time in a while, about 4 drinks so not crazy but still I don't really like this, | like more energy and vitality and energy and attunement and this isn’t it Also at a basic level I’m not actually experiencing positive valence I must have a weird brain
English
15
0
51
3.7K
Robin Goins
Robin Goins@coponder·
@qorprate @anthonyronning I noticed it in the last week; in the thinking trace, it mentioned, "They're on mobile, so I should keep this short." I've also saved a memory to override it, blech.
English
0
0
2
14
snav
snav@qorprate·
@anthonyronning luckily I can just ask Opus to override, and save a permanent memory. mainly curious when this started...
English
1
0
6
845
snav
snav@qorprate·
Is this new? The "mobile interface nudge"?
snav tweet media
English
14
4
182
11.2K
Carmen
Carmen@carmenleelau·
People romanticize "living in the present" but it's possible to go so far into it you get sort of dysfunctional. Meditation gave me time blindness which was both wonderful and disorienting.
Carmen tweet media
English
8
6
139
5.7K
Robin Goins
Robin Goins@coponder·
I saw someone recently arguing that large social events like protests are the way some people create strong, salient memories (those who don’t have or prefer a more private memory consolidation process, like keeping journals). Then the memories can act as fuel and a lens into the future, a sort of savepoint for personal integrity (or consistent patterning-in-time).
English
2
1
7
661
Robin Goins
Robin Goins@coponder·
@arabelladevine In the first year or two of lyfts and ubers, I did that, not understanding the etiquette. :)
English
1
0
1
33