iNic

10 ms·

91.
by iNic 2y ago
NASA is adhering to not biting the hand that feeds it, since they can no longer build their own rockets.
92.
by iNic 2y ago
- The Fourier Transform is an invertable operator (i.e. it acts on functions, in the case of matrices both functions and operators are themselves matrices). It transforms into what we call frequency space. - This is most intuitive for signa
93.
The FFT Strikes Back: An Efficient Alternative to Self-Attention(arxiv.org)
456 points by iNic 2y ago|168 comments
94.
Making a "Miracle" HIV Medicine(asimov.press)
2 points by iNic 2y ago|0 comments
95.
by iNic 2y ago
Please do not mistake "AI ethics" for "AI safety", where the latter is about "Studying and reducing the existential risks posed by advanced artificial intelligence" [1, 2]. [1]: https://forum.effecti
96.
by iNic 2y ago
This is only ever true during "transition" periods. Chess computers at this point are strictly better than just humans or human computer teams. Of course, it is currently true for LLMs, but the question is how long?
97.
by iNic 2y ago
The tokenizer catches it: https://platform.openai.com/tokenizer .
98.
by iNic 2y ago
This is an interesting comment, because it won't matter to who? I can't tell if you are pro or anti Musk. Are you saying that Musk misinterpreted the investigation and so shut down Starlink, or that anti-Musk people will misinterp
99.
by iNic 2y ago
https://archive.ph/lItkf
100.
The Constitutional Crisis Is Here(theatlantic.com)
59 points by iNic 2y ago|11 comments
101.
by iNic 2y ago
To add to this: specifically regarding random firings and loopholes. Every marginal dollar spend on the IRS brings in $5 to $9 in return to the government [*]. Yet they are firing IRS agents. More generally, we know, and the government know
102.
by iNic 2y ago
I am happy to cut government spending and increase government efficiency (obviously). I have so far not seen any evidence that DOGE is working is this direction. As I like to point out the marginal dollar spent on the IRS brings in ~$10 of
103.
by iNic 2y ago
The following is some unstructured thoughts: I know you are being facetious, but it is not obvious (to me at least) what the best approach is in this scenario where future AI capabilities and use is unclear. I don't know what the corre
104.
by iNic 2y ago
I wonder if there's a "minimum viable connectivity threshold" in modern life - you literally cannot function below a certain baseline of digital access. You could model the failure of "delete everything" strategies
105.
by iNic 2y ago
This is a direct consequence of the housing theory of everything [1]. When land is scarce landlords can eat up any surplus produced by both labor and capital! [1]: https://worksinprogress.co/issue/the-housing-theory-of-
106.
by iNic 2y ago
The housing theory of everything is true [1]. When land is scarce landlords can eat up any surplus produced by both labor and capital! [1]: https://worksinprogress.co/issue/the-housing-theory-of-every...
107.
by iNic 2y ago
I compared by "Following" tab and "For You" Tab and they got a pretty similar health score ~40%. It wasn't able to analyze the "political diversity" of the Following tab which I will take as a win lol. It
108.
by iNic 2y ago
Yes that is true, but there are other ways to see spiky-like behavior. First, the volume of spheres (or balls rather) in higher dimensions goes to zero as the dimension grows. Said another way, to keep unit volume on a ball you need to grow
109.
by iNic 2y ago
There is a standard thought experiment where you start with a hypercube of side-length 2, centered at the origin. You then place a radius 1 sphere on each vertex of this hypercube. The question then becomes: what is the largest sphere you c
110.
Datapods – Reclaim ownership over your data(datapods.app)
2 points by iNic 2y ago|0 comments
111.
by iNic 2y ago
The kind of intuition you gain for higher dimension tends not to be visual. It is more that you learn a bunch of tools and these in turn build intuition. For example high dimensional spheres are "pointy" and most of their volume a
112.
0de5 – Keep Learning Forever(0de5.net)
1 points by iNic 2y ago|0 comments
113.
Towards Benchmarking LLM Diversity and Creativity(gwern.net)
2 points by iNic 2y ago|0 comments
114.
A new golden age of discovery – DeepMind(deepmind.google)
2 points by iNic 2y ago|0 comments
115.
by iNic 2y ago
Posting this mostly as a complaint about Apple taking so long to get this to a functioning version.
116.
Accelerated Jax on Mac(developer.apple.com)
17 points by iNic 2y ago|3 comments
117.
by iNic 2y ago
It sounds like every sentence is an ad read.
118.
Why so few Matt Levines?(gwern.net)
263 points by iNic 2y ago|141 comments
119.
by iNic 2y ago
Twitter User @_xjdr (with @doomslide) has been working on new sampling techniques and getting some early promising results. The hope is to achieve an open source version of what o1 does. The key insight is to use entropy and the variance of
120.
Entropix: Entropy Based Sampling and Parallel Cot Decoding(github.com)
2 points by iNic 2y ago|1 comments

More ›