Ben

@data-nerd

Data Stuff @pinatacloud

41 following43 followers

Latest casts

Ben@data-nerd·
Reply

This is a really good read. Thanks for sharing.

Ben@data-nerd·
Reply

Otters are the best, some of my social media feeds are just filled with videos of otters being little goof balls lol. 8 $DEGEN

Ben@data-nerd·

From @MistralAILabs: Announcing `mistral-finetune`, the official repo and guide on how to fine-tune Mistral open-source models: https://github.com/mistralai/mistral-finetune

Ben@data-nerd·
Reply

I've seen lots of studies trying to take the same approach to improve readability for medical reports for patients as well, cool stuff. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10524809/

Ben@data-nerd·
Reply

Agree with Foundry, the fuzz testing helps give you guarantees in your code that are hard to match.

Ben@data-nerd·
Reply

So true! Having excellent data is one of those things that is overlooked until you try to build ML / AI models on real data. You can get so much signal from good feature engineering and leveraging domain knowledge to add fields that matter to the data.

Ben@data-nerd·
Reply

Interviews with John Carmack are the best, especially hearing about the pre-doom days. There's a good Lex podcast with him. For those interested 30 of the 40 papers he mentions in the article can be found below. https://arc.net/folder/D0472A20-9C20-4D3F-B145-D2865C0A9FEE

Ben@data-nerd·
Reply

The way I usually find good papers is by looking for heavily cited papers or researchers from top ranked schools. There is a ton of well written nonsense in the space but these are good filters.

Ben@data-nerd·
Reply

Great thread, this is very much sums up a common experience in analytics work. (having done analytics consulting for a while) Also, Underrated Comment - "the hard work was less engineering and data science, and more social and organizational - aligning people around a common language" 6 $DEGEN

Ben@data-nerd·

Not mine, but too cute not to post!