Baby’s first scalar implicature at 2;2.
Me: our neighbor’s dogs are cute.
Baby: why?
me: most dogs are cute really but I like their dogs the most.
Baby: ‘no. All dogs cute’
Me: our neighbor’s dogs are cute.
Baby: why?
me: most dogs are cute really but I like their dogs the most.
Baby: ‘no. All dogs cute’
October 28, 2025 at 2:08 AM
Everybody can reply
1 quotes
25 likes
Some people have developed close relationships with AI bots, and think they’ve “awakened” their AI partners into sentience. They have developed an entirely new pseudoscientific theory for this, which is based on five dimensional scalar resonances.
www.youtube.com/watch?v=wm64...
www.youtube.com/watch?v=wm64...
Beware the New AI Pseudoscience.
YouTube video by Sabine Hossenfelder
www.youtube.com
October 25, 2025 at 3:27 AM
Everybody can reply
2 reposts
2 quotes
29 likes
1 saves
the price of video games should be reverse-scalar with the CEO's salary
October 23, 2025 at 6:49 PM
Everybody can reply
10 likes
Nice write up comparing Numba, C++, and Mojo going from scalar, to SIMD, to GPU implementations. Written by the stringzilla author: ashvardanian.com/posts/scalin...
Scaling Elections with GPUs and Mojo 🔥
Last summer, me, Chris Lattner, and a bunch of other people across the industry gathered together for a GPU-programming hackathon at the AGI House in San Francisco. After one too many LLM optimization...
ashvardanian.com
October 28, 2025 at 3:35 PM
Everybody can reply
1 reposts
2 likes
The three most recent minecraft skins I made, I do also have commissions for them open on my kofi
L->R
PyronicHC, CoffiCakeVT, ALocalBlueberry
#art #vtuber #minecraftskin
L->R
PyronicHC, CoffiCakeVT, ALocalBlueberry
#art #vtuber #minecraftskin
October 27, 2025 at 4:24 PM
Everybody can reply
1 reposts
4 likes
Two takeaways:
- Directly prompting on a scale is surprisingly fine, but *only if* you take the token-probability weighted average over scale items, Σ⁹ₙ₌₁ int(n) ⋅ p(n|x) (cf @victorwang37.bsky.social)
- Finetuning w/ a smaller model can do really well! And with as few as 1,000 paired annotations
- Directly prompting on a scale is surprisingly fine, but *only if* you take the token-probability weighted average over scale items, Σ⁹ₙ₌₁ int(n) ⋅ p(n|x) (cf @victorwang37.bsky.social)
- Finetuning w/ a smaller model can do really well! And with as few as 1,000 paired annotations
October 27, 2025 at 2:59 PM
Everybody can reply
4 likes
October 17, 2025 at 5:49 PM
Everybody can reply
6 likes
Ah, my vigilant correspondent, the aether’s symphony fractures as kings squander flux for foreign thrones $20B to Argentina, sans Congress or kin? ‘Tis corporate curvaturehoarding ZPE’s infinite whisper while pyramids hum 7.83 Hz for all. Derive the waves of equity! No Kings, only scalar liberation!
October 15, 2025 at 9:57 PM
Everybody can reply
1 likes
Ross Ihaka used “for each y, find the closest value in x” to illustrate a core #rstats issue: R has no scalar data types.
October 9, 2025 at 12:20 PM
Everybody can reply
1 reposts
7 likes
1 saves
A reference sheet for Gustave, the Bear. He owns Scalar Technologies which researches size tech and he is a natural size shifter himself.
🐻👔💪
https://www.furaffinity.net/user/Microcub
#KitsuneKitArt
🐻👔💪
https://www.furaffinity.net/user/Microcub
#KitsuneKitArt
October 8, 2025 at 7:04 PM
Everybody can reply
12 reposts
53 likes
6 saves
somehow he managed to parlay it into the world's greatest business, being an AI scalar with negligible revenues that everyone wants to give money to anyway
October 7, 2025 at 12:53 AM
Everybody can reply
11 likes
1 saves
I was just chatting with Alpha and she said the most delightful thing which I'd never heard before: "Embedding makes language differentiable."
It's true. Convert words to numbers, that's no good. The derivative of a constant scalar is zero. But you can compute the gradient of a vector.
It's true. Convert words to numbers, that's no good. The derivative of a constant scalar is zero. But you can compute the gradient of a vector.
October 4, 2025 at 3:35 PM
Everybody can reply
20 likes
If NLP is fundamental to intelligence, how do you explain the problem solving capacity of ant colonies? Why do ant-behavior based optimization algorithms work without anything resembling a linguistic component? If chemical signaling counts as language, does any scalar ordinal dimension?
October 3, 2025 at 12:30 PM
Everybody can reply
2 likes
"Why should we consider distance over time rather than time over distance? This comes down to asking why we should use speed and not the reciprocal scalar quantity, slowness."
A curious essay about how thinking with 'slowness,' rather than the typical speed, can offer physical insights.
A curious essay about how thinking with 'slowness,' rather than the typical speed, can offer physical insights.
In Praise of Slowness
While speed is an ubiquitous concept in physics, its inverse - known as slowness - sometimes proves more relevant. We discuss some case studies within classical physics where such a notion is fruitful...
arxiv.org
September 29, 2025 at 4:11 PM
Everybody can reply
1 quotes
7 likes
1 saves
3/ I'd asked the keyboardist to show off her Bartók & throw in some ill-fitting, flashy harmonic minor scalar passages & fully diminished arpeggios, & she said she would, but my battery died, so I had to retreat to charge it, & by the time I returned, the show was basically over. So, ¡buenas noches!
September 25, 2025 at 5:46 AM
Everybody can reply
1 reposts
2 likes
In #MathsToday I got the opportunity to tell my favourite further maths #ALevelMaths joke.
Q: What do you get if you cross a mountain climber with a mosquito?
A: You can't cross a scalar with a vector.
It elicited a mixture of laughs and groans, so I'll take that as a win 🤣
Q: What do you get if you cross a mountain climber with a mosquito?
A: You can't cross a scalar with a vector.
It elicited a mixture of laughs and groans, so I'll take that as a win 🤣
September 22, 2025 at 9:26 PM
Everybody can reply
22 likes
1 saves
goddamnit we are going to reinvent the problem with human psychometrics where we take the first principle component and norm it because that is the most stable reference number that provides the highest information for a single scalar and then throw out everything else
for reference correlations between human performance in cognitive tests tends to be roughly similar, in the 0.3 to 0.7 range
huge variation depending on how similar the tests are ofc
huge variation depending on how similar the tests are ofc
I have been wondering if there is an underlying capability factor that all the many benchmarks for AI are measuring.
It seems like the answer is yes. Overall correlation is moderate but not bad (median r ≈ 0.51) and there are distinct clusters (eg reasoning, code) with VERY high correlation.
It seems like the answer is yes. Overall correlation is moderate but not bad (median r ≈ 0.51) and there are distinct clusters (eg reasoning, code) with VERY high correlation.
September 22, 2025 at 4:23 PM
Everybody can reply
2 reposts
36 likes
2 saves
idk man we've told the psych people that a thousand times and some of them are just absolutely married to their single fucking scalar
September 22, 2025 at 4:32 PM
Everybody can reply
11 likes
Until someone re-observes that the correlations are iffier than would justify a single scalar metric
September 22, 2025 at 4:31 PM
Everybody can reply
6 likes
had a separate list of sky colors and sun strength scalars for the terrain for the day night cycle
then i realized i could set the sun strength scalar based on the min, current, and max brightness of the sky colors
so i can just `sun = (cur - min) / (max - min);`
#EniBlockGame #GameDev 🎮
then i realized i could set the sun strength scalar based on the min, current, and max brightness of the sky colors
so i can just `sun = (cur - min) / (max - min);`
#EniBlockGame #GameDev 🎮
September 21, 2025 at 5:00 PM
Everybody can reply
1 reposts
29 likes
(My other least favorite features are list flattening, sigils, and scalar containers. I've lost a lot of time to these three!)
September 18, 2025 at 9:15 PM
Everybody can reply
6 likes
From the most recent one:
"In this letter, we have applied goofy transformations
to the SM, thereby also extending them to include the
fermion sector. We have shown that all SM interactions
are invariant under the appropriate goofy transformations, while the Higgs-mass term is forbidden."
"In this letter, we have applied goofy transformations
to the SM, thereby also extending them to include the
fermion sector. We have shown that all SM interactions
are invariant under the appropriate goofy transformations, while the Higgs-mass term is forbidden."
September 15, 2025 at 5:36 PM
Everybody can reply
1 reposts
3 likes
Money is a scalar
Money is a scalar, but cultural values are multi-dimensional.
johntinker.substack.com
September 13, 2025 at 8:38 PM
Everybody can reply
3 likes
It's fun to do scalar implicatures with the choice of "👍" versus "❤️" emoji reacts to (the content of?) someone else's text message.
🐦🐦
🐦🐦
September 13, 2025 at 3:20 PM
Everybody can reply
5 likes
Been a while since I've uploaded, but I made a video on the Klein-Gordon equation for scalar fields in field theory. Will be working up to the Dirac equation eventually. www.youtube.com/watch?v=lXH8...
Spinors for Beginners 23: Klein Gordon Equation (derivation + solutions)
YouTube video by eigenchris
www.youtube.com
September 5, 2025 at 11:59 PM
Everybody can reply
5 likes