joelburget

7 ms·

1.
On-Policy Distillation(thinkingmachines.ai)
5 points by joelburget 11mo ago|0 comments
2.
US AI Action Plan(ai.gov)
426 points by joelburget 1y ago|618 comments
3.
Showh HN: Microjax – JAX in two classes and six functions(github.com)
46 points by joelburget 1y ago|1 comments
4.
ASI existential risk: reconsidering alignment as a goal(michaelnotebook.com)
3 points by joelburget 1y ago|0 comments
5.
by joelburget 2y ago
I wrote an OCaml implementation of this paper a few years ago, which I've now extracted into its own [repo]( https://github.com/joelburget/constructive-reals/blob/main/C... ) The link in the paper to
6.
by joelburget 2y ago
And more recently, [Language Models Use Trigonometry to Do Addition]( https://arxiv.org/abs/2502.00873 )
7.
Writing Einsum in Depth (In OCaml)(joelburget.com)
2 points by joelburget 2y ago|0 comments
8.
by joelburget 2y ago
This is a good idea, though one problem is that Einsum notation (as realized in Numpy and Pytorch) doesn't support the notion of co-contravariance, and the site is based on their Einsum notation. I could potentially add the variances f
9.
Einsum in Depth(einsum.joelburget.com)
87 points by joelburget 2y ago|31 comments
10.
by joelburget 2y ago
A couple of these I'd like references on if anyone happens to have them. 1. "current science suggests that the actual health impact from consuming most types of plastic might well be essentially zero" 2. "the (weak) evid
11.
Underrated reasons to be thankful IV(dynomight.net)
2 points by joelburget 2y ago|1 comments
12.
DeepSeek-R1-Lite-Preview is now live(api-docs.deepseek.com)
1 points by joelburget 2y ago|0 comments
13.
Quantum error correction below the surface code threshold(arxiv.org)
1 points by joelburget 2y ago|0 comments
14.
by joelburget 2y ago
o1 is an application of the Bitter Less. To quote Sutton: "The two methods that seem to scale arbitrarily in this way are search and learning." (emphasis mine -- in the original Sutton also emphasized learning ). OpenAI and o
15.
by joelburget 2y ago
Vision Transformers do a shocking amount of compression in the tokenizer. In the [Chameleon paper]( https://arxiv.org/pdf/2405.09818 ) they say the tokenizer "encodes a 512 × 512 image into 1024 discrete tokens from
16.
by joelburget 2y ago
Vision transformers should be our default guess as to how GPT-4o works, yet this article never mentions them.
17.
Disrupting the Deepfake Supply Chain(openletter.net)
2 points by joelburget 3y ago|0 comments
18.
by joelburget 4y ago
It works on all human languages, just inefficiently. I ran it over a sample I found on wikipedia: sample = "ฟองมันฟันหนู, ฟันหนูฟองมัน, ฝนทองฟองมัน" len(sample), len(enc.encode(sample)) This returns `39, 40` so it&#x2
19.
by joelburget 4y ago
A few interesting findings: * the cl100k_base tokenizer has ~100k tokens -- previous tokenizers had ~50k. (enc.n_vocab gives 100277 but some numbers in that range don't work, starting at 100256) * it has exactly 1110 tokens which are j
20.
Kelvin Versioning(jtobin.io)
2 points by joelburget 4y ago|0 comments
21.
Branch Cuts and Common Lisp(johndcook.com)
3 points by joelburget 5y ago|1 comments
22.
by joelburget 5y ago
For those looking to run this on a Mac, the following seems to have worked for me (M1, Big Sur 11.2.3): ``` brew tap gcenx/wine brew install --cask --no-quarantine wine-crossover brew install winetricks winetricks corefonts vcrun6 vb5r
23.
by joelburget 7y ago
You have a keen eye -- thanks for the close reading.
24.
by joelburget 7y ago
Thanks for the heads up. It never occurred to me the svg might render differently. I'll see if we can fix this.
25.
by joelburget 7y ago
I'll use a bit of Haskell in my response, since that's what I'm most comfortable with and what we actually use for our work. Hopefully that's okay. We can define a data type of expressions in our language, with addition,
26.
by joelburget 7y ago
Stick around for part 2. We haven't gotten to the good stuff yet :)
27.
by joelburget 7y ago
Co-author here. We're planning to turn this into a series, with additional posts on how symbolic execution works at a lower level, how to present counterexamples to users, and ideas for future programming tools built on symbolic execut
28.
A gentle introduction to symbolic execution(blog.monic.co)
190 points by joelburget 7y ago|27 comments
29.
Google Fonts and 한국어(googlefonts.github.io)
4 points by joelburget 8y ago|2 comments
30.
by joelburget 9y ago
Similarly, I once wrote a triangular maze generator https://www.khanacademy.org/computer-programming/triangle-ma...

More ›