natsucks

13 ms·

61.
by natsucks 3y ago
A key advantage of US over other imaging modalities is portability. A key disadvantage is resolution with depth. Tomographic ultrasound systems exist, but if one is going to go with a bulky machine you might as well go with CT or MRI.
62.
by natsucks 3y ago
i'm astonished that no one has created a well-functioning dumbphone that has what most people would agree upon as essential - things like maps, 2FA, encrypted messaging, tethering, android auto/carplay, etc.
63.
by natsucks 3y ago
Arguably the second culture is also built for leadership. Evidence can't tell you which direction to take a company in, and the person that decides on this direction needs to project confidence in others to instill it in others.
64.
by natsucks 3y ago
this. this right here.
65.
by natsucks 3y ago
important comment.
66.
by natsucks 3y ago
"In attempts to mitigate the dangers of memory unsafe code in C and C++, many software manufacturers invest in training programs for their developers. Many of these training programs include tactics designed to reduce the prevalence
67.
by natsucks 3y ago
Can you recommend a good alternative?
68.
by natsucks 3y ago
right. if that was real time, the latency was very impressive. but i couldn't tell.
69.
by natsucks 3y ago
or smartphone use, which is correlated with emotional dysregulation.
70.
by natsucks 3y ago
"if the first thing a company has to say about a product or feature is that it's powered by AI, that is a strong signal that it isn't actually very useful." i have a product team that is totally disconnected from the eng
71.
by natsucks 3y ago
Why no multi-turn evaluation? A lot of these benchmarks fail to capture the strength of ghost attention used in Llama 2 chat models.
72.
by natsucks 3y ago
What's holding me? People writing software who are choosing C/C++ for legacy reasons only.
73.
by natsucks 3y ago
can we just move on from C/C++ already
74.
by natsucks 3y ago
recently ditched Supabase for Weaviate. I was tired of the python bindings not keeping up, no hybrid search, slower search algorithms. Also Supabase has a lot of features that i just don't need.
75.
by natsucks 3y ago
It makes no sense to me why people want to run this model when it barely outperforms Llama 2 while being over twice the size.
76.
by natsucks 3y ago
Interesting. Do you think that points to the current limitations of RAG or a mismatch in what a user truly wants from search?
77.
by natsucks 3y ago
Do you know of a good example demonstrating RAG with query understanding via structured attributes?
78.
by natsucks 3y ago
So in your opinion what are some examples of highly effective RAG systems/implementations?
79.
by natsucks 3y ago
Author has a good point, but no mention of hybrid search? ...not that hybrid search solves everything.
80.
by natsucks 3y ago
curiously, they don't mention MMLU, which is one of the most commonly reported metrics for comparison.
81.
by natsucks 3y ago
exactly. i have a hard enough time running a 70b model at fp16. no way i attempt 180 without a significant performance gain.
82.
by natsucks 3y ago
that would be great. thanks. will do.
83.
by natsucks 3y ago
Just migrated from Supabase + pgvector to Weaviate hoping to take advantage of langchain.retrievers.weaviate_hybrid_search.WeaviateHybridSearchRetriever. Unfortunately i can't get it work asynchronously :(
84.
by natsucks 3y ago
i'll take this banter. it's fun. long live emacs.
85.
by natsucks 3y ago
The native bindings are...not great. Personally i use xah-fly-keys, but evil (vim) keybindings are popular.
86.
by natsucks 3y ago
hilarious.
87.
by natsucks 3y ago
I wonder if this is a good application for recently developed multimodal/vision-language models that demonstrate "emergent" properties like an ability to classify objects unseen in their training data.
88.
by natsucks 3y ago
As a reMarkable owner, I love it. I want it. gimme gimme gimme.
89.
by natsucks 3y ago
- After seeing how merely quantizing a model can make it go berserk, I have very little confidence I can fine tune an LLM and expect similar performance benchmarks. - A RAG-empowered LLM can tell you where the knowledge used to answer a que
90.
by natsucks 3y ago
not to mention staying true to their benchmarks...

More ›