dabbied
dabbied.bsky.social
dabbied
@dabbied.bsky.social
Coercing ML models to help plan construction megaprojects @nplan.io
Santa's got some upgrades 😁 www.noradsanta.org/en/map
Official NORAD Tracks Santa
www.noradsanta.org
December 24, 2024 at 1:30 PM
Reposted by dabbied
Interesting: according to this paper Gemini, Llama 3, Gemma 2 work best with JSON format. Many models drop in performance when tasked for structured output.

https://buff.ly/4eZyv2U

#ml # ai
December 3, 2024 at 5:35 PM
Reposted by dabbied
Damn: researchers at Lawrence Livermore are using o1 to do complicated nuclear fusion research.

When people say AI will increase the pace of scientific R&D, this is what they're talking about.

www.theinformation.com/articles/why...
November 25, 2024 at 4:44 PM
Reposted by dabbied
Trying out Docling for PDF info extraction and it’s surprisingly good! The chunk feature splits PDFs into metadata (labels like headers, paragraphs, tabels etc.) and text (actual content). @tedunderwood.me seems like the pdf problem is close to being solved!
github.com/DS4SD/docling
#NLPSky
GitHub - DS4SD/docling: Get your documents ready for gen AI
Get your documents ready for gen AI. Contribute to DS4SD/docling development by creating an account on GitHub.
github.com
November 26, 2024 at 1:17 AM
Reposted by dabbied
I find it amusing that the emerging standard for giving an LLM the ability to work with your technology is just a text file explaining clearly how your technology works

(Once folks realize they also need to sell the LLMs on why they should use a technology; things will get wild)

llmstxt.org
November 25, 2024 at 8:50 PM
Reposted by dabbied
During most of the vast history of humanity, technologies have evolved gradually, not through leaps of genius, but slow cultural adaption.

Before the scientific method, easy communication & printed books, innovation happened at the level of societies, not people. www2.psych.ubc.ca/~henrich/pdf...
November 25, 2024 at 9:19 PM