Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
joewferrara
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
11 ms
·
31.
▲
by
joewferrara
3y ago
This is a great article about the technical difficulties of building a RAG system at scale from an engineering perspective. Performance is about speed and compute. A topic that is not addressed is how to evaluate a RAG system where performa
32.
▲
by
joewferrara
3y ago
Shit is crazy! I did not know about the amount of fake scholarships private colleges give until today! If I knew it in high school I would have applied to more expensive colleges. I only applies to public universities in my state in high sc
33.
▲
by
joewferrara
3y ago
I’m not sure about textbooks specifically on infinity. As another comment mentions, the study of infinity generally falls under the math subject of set theory so a set theory textbook would be a mathematician’s definitive resource on infini
34.
▲
by
joewferrara
3y ago
These visualizations are very cool, and will hopefully lead to more understanding of what's happening with neural networks internally. The next thing we need is an example/write up of someone using these visualizations to either 1
35.
▲
by
joewferrara
3y ago
Not a textbook, but Everything and More - A Compact History of Infinity by David Foster Wallace is a fun read about infinity. It’s technical but not quite as technical as a textbook that would be used in a college course. I have a PhD in ma
36.
▲
by
joewferrara
3y ago
This line between fine tuning and continuous pre-training is what I’m interested in. What is the investment difference between fine tuning, contoured-training and training from scratch? Do you (or anyone else) have any good sources or know
37.
▲
by
joewferrara
3y ago
This is a very cool concept. The example given of Bytedance combining the text embedder ALBERT with the a transformer image embedder, to make an embedder that can do image and text at the same time to get an interaction score is fascinating
38.
▲
by
joewferrara
3y ago
They test two fine tuning tasks in the article - reliable output formatting and custom tone. These are two tasks (reliable output formatting in particular) that are advertised regularly as areas where fine tuning an LLM should work. The goa
39.
▲
by
joewferrara
3y ago
This is a cool project. What do you think about the solution to redact text in a user’s message using an NER model - replacing the named entities with placeholders like PERSON_1, LOCATION_1 etc, and then sending the redacted message to open
40.
▲
by
joewferrara
3y ago
Modular forms also have applications to representation theory and physics in conformal field theory https://researchers.ms.unimelb.edu.au/~dridout@unimelb/semin...
41.
▲
by
joewferrara
3y ago
It’s a consequence of the maturity of the smart phone market/product. Apple’s “boring” new IPhone announcement last week (or was it two weeks ago?) is another consequence of the maturity of smart phones. There’s less and less new groun
42.
▲
by
joewferrara
3y ago
This article is a lot of words about the fact that twitter, or should I say X, is promoting tweets, er posts, that keep one on X (posts without a link in them, the example given in the article is posts with pictures) instead posts that send
43.
▲
by
joewferrara
3y ago
The author has more experience in the field than me, so gotta defer to him for the most part, and while I generally agree with the post, I disagree strongly with one point. The author frames this new era of AI about using transformer models
44.
▲
by
joewferrara
3y ago
Check out this blog post I wrote about my evolving understanding of what a website is and how the internet works. I lack a background in computer science, and this posts reflects what I've learned since leaving academia to work as a da
45.
▲
What's a Website
(joe-ferrara.github.io)
2 points
by
joewferrara
3y ago
|
1 comments
46.
▲
Blog Post on RSA
(joe-ferrara.github.io)
4 points
by
joewferrara
3y ago
|
1 comments
47.
▲
by
joewferrara
3y ago
Check out this blog post I wrote on RSA and let me know what you think!
48.
▲
by
joewferrara
3y ago
This essay by the same person who answered the mathoverflow post greatly expands on the ideas in the post. https://arxiv.org/pdf/math/9404236.pdf It’s a great read for young people interested in mathematics and in
49.
▲
Euclidean Algorithm
(joe-ferrara.github.io)
1 points
by
joewferrara
3y ago
|
0 comments
50.
▲
by
joewferrara
3y ago
This game is exactly like graduate school in a PhD program! Love it!
51.
▲
by
joewferrara
3y ago
Pretty interesting that signs of self domestication include a shrunken brain and aggressive males being eliminated from the population. I would not have thought that our brains have shrunk since we (humans) became self domesticated.
52.
▲
by
joewferrara
3y ago
This is pretty good considering the fact that David Foster Wallace took his own life in the end. I read the whole quote before I saw at the end it’s from him.