cuuupid

20 ms·

211.
by cuuupid 3y ago
Not in love with axolotl but appreciate the advantages. This is an interesting approach, but you can also finetune easily on providers who wrap axolotl like Replicate [1], Modal [2], or if you want to run the infra, LLM Engine [3]. My only
212.
Ask HN: Why do all electronic health records suck so bad?
3 points by cuuupid 3y ago|5 comments
213.
by cuuupid 3y ago
This is a big improvement, but I'm not a believer that SQL is the most appropriate query lang here. Personally am more bullish on language models being trained with ORMs, as those normally capture much more information about the fields
214.
by cuuupid 3y ago
Google has fallen so far. Both Inception and Mobilenet were released openly and changed the entire AI world. Nowadays we just get blog posts about results that were supposedly achieved, an accompanying paper that can’t be reproduced (becaus
215.
by cuuupid 3y ago
Yep, for most PWS contracts there are base personnel requirements. That being said for research the bar is pretty high, for regular industry it's pretty low: - experience in the field (or agency-based past performance) - clearance (usu
216.
by cuuupid 3y ago
This is a myth that’s exaggerated by specific anecdotes around sole-source (which is where you work with the govt and they put out a notice saying they intend to go with you). Because of there not being constraints on time there and them ne
217.
by cuuupid 3y ago
This is a really good in depth article but misses one thing so many projects miss: just sell to the civil government. USG has so many programs (eg NSF grants) for tech and the disjointedness of civil agencies, IC, and state govts means they
218.
by cuuupid 3y ago
I’d like to note that this model’s parameter usage is low enough (13b) to run smoothly at high quality on a 3090 while beating GPT-3.5 on humaneval and sporting 32k context. 3090s are consumer grade and common on gaming rigs. I’m hoping gam
219.
by cuuupid 3y ago
Have been watching this unfold over twitter the past couple days and all I can think to say is that Carta really, really needs to hire a communications director. This situation is really not that terrible (Carta is trying to monetize via se
220.
by cuuupid 3y ago
The expansion llm would have to have a pretty good model of language so would likely need to be 7B realm though, but could be useful given we are almost at a time where 7b models can run ubiquitously on most consumer hardware
221.
by cuuupid 3y ago
There’s a good exploration of counters here: https://www.scientificamerican.com/article/physicists-critic... It remains to be seen if wolfram’s theory outperforms earlier attempts
222.
by cuuupid 3y ago
I’m curious if this could one day be used to identify results or concepts that we haven’t yet discovered. It seems reasonable since it discovered things like gravity that we hadn’t identified until well past the classical era.
223.
by cuuupid 3y ago
Mamba has shown SSMs are highly likely to be contenders for the replacement to transformers in a really similar fashion to when transformers were first introduced as enc-dec models. I’m personally very excited for those models as they’re al
224.
by cuuupid 3y ago
Stark contrast with Google's "all demo no model" approach from earlier this week! Seems to be trained off Stanford's Megablocks: https://github.com/mistralai/megablocks-public
225.
Mistral releases 8x7 MoE model via torrent(twitter.com)
61 points by cuuupid 3y ago|9 comments
226.
by cuuupid 3y ago
This is promising as the future of modeling. Outperforming transformers doesn’t catch my eye because there is so much variation in performance based on size, training data, training method etc. However the greater (5x) inference bandwidth m
227.
by cuuupid 3y ago
Just want to chime in and say this is one of the best tools, I have this permanently pinned in Arc and it makes it really, really easy to collaborate. Some more dots: - free, open source, encrypted and private - I hated this at first becaus
228.
by cuuupid 3y ago
I've found the best LLMs still fall a bit short when it comes to being truly best in class, SOTA in most useful tasks is still dominated by smaller models with focused approaches and innovative loss mechanisms.
229.
by cuuupid 3y ago
Saw the collaboration mode and immediately feared it would devolve into people throwing around slurs as many unmoderated online experiences often do. Pleasantly surprised to see conspiracies about Tupac and 'vue gang' take over th
230.
by cuuupid 3y ago
Out of curiosity do you share the suspected fraud with law enforcement? At that amount there are probably too many to chase down so I’d think this is just an unprosecuted crime?
231.
by cuuupid 3y ago
One thing no one’s talking about is how insanely great all of this has been for defense. For the better part of the year the US has been mulling over GPT-3.5 as a maybe/maybe not and only ever touching Azure endpoints. Securing most of
232.
by cuuupid 3y ago
Tbh this reads a lot like Ilya thinking he’s Tony Stark and his (still impressive) language model is somehow the same as an iron man suit. Which is arrogance to the point of ignorance, reality isn’t that romantic. I can only hope this doesn
233.
by cuuupid 3y ago
This coupled with Microsoft’s recent “security concerns” brings up the possibility this is related to them misusing or selling data they weren’t supposed to?
234.
by cuuupid 3y ago
I think people are forgetting that transformer architectures are a wider field from GPT and predate GPT3 by 3+ years. Referring to transformer architectures using a branded commercial nomer (GPT) is just going to help cement OpenAI’s brand
235.
by cuuupid 3y ago
Is that you Mulva?
236.
Cloudflare Continuing to Experience Outages(twitter.com)
5 points by cuuupid 3y ago|2 comments
237.
by cuuupid 3y ago
Can’t link to their official status page. Have had a lot of my Cloudflare hosted sites go down for most of the day, looks like issues still waiting to be resolved.
238.
Ask HN: Why has no one replaced Ticketmaster?
196 points by cuuupid 3y ago|162 comments
239.
by cuuupid 3y ago
Based off the three authors all having long-term ties to China, the tone of the article, it parroting known propaganda, and the comments under it all scrutinizing it, I feel this is very obviously propaganda. That being said if we were to a
240.
Orama: Vector-capable search written in TypeScript(oramasearch.com)
1 points by cuuupid 3y ago|0 comments

More ›