civilized

18 ms·

31.
by civilized 2y ago
> RAG is a technology to augment search engines with vector search Based on the words in the acronym, that seems exactly backwards. The words suggest that it is a generation technology which is merely augmented by retrieval (search),
32.
by civilized 2y ago
People have tried to port core R data science packages like dplyr to Python many times now and all that's happened is Python has fallen further behind. It's not just a matter of manpower and funding, which Python has more of. Pyth
33.
by civilized 2y ago
Note: it may not be immediately apparent, but all constant-coefficient linear recurrence relations of any complexity can be straightforwardly solved with this technique.
34.
by civilized 2y ago
In which case you have a trivial point rather than a backwards argument: "the output seems like it knows what's it talking about, and the easiest way to explain that is if it really knows what it's talking about."
35.
by civilized 2y ago
> Point is that the simplest way to excel in next token prediction in the way human consider correct - which is rated by how people feel the predictor mimics a human understanding - is to actually have a world model and other components
36.
by civilized 2y ago
> humans tend to agree that in a vast array of cases state of the art LLMs do predict tokens very well This argument is backwards. Humans don't measure the next token prediction ability of the agents they speak to, human or AI. We r
37.
by civilized 2y ago
> Science as an insistution has all those hallmarks: worship, authority and rituals. I'm not trying to make a case for religion here, but wanted to point out the shallowness of the whole write-up. He's talking about science as
38.
by civilized 2y ago
Allowing duplicate rows is the only practical decision for the following reasons: 1. Most of the ways you can mess up in SQL involve failing to respect unique keys in joins, and this can be done just as easily whether or not duplicate rows
39.
by civilized 2y ago
I don't think a single one of these claims is true: 1. SQL is not mostly used as a compiler target these days. 2. ORMs do not hide SQL nastiness. 3. Relational calculus, however well-designed, does not manage to avoid the practical pro
40.
by civilized 2y ago
And that's because of Medicare? How do you figure?
41.
by civilized 2y ago
Is "hard" a bit of an overstatement for problems like "I'm using a library that mangles the query"? Couldn't you search for the literal text the user inputs? Maybe let them use regex?
42.
by civilized 3y ago
Your math may well be correct but I can't follow it without a spreadsheet, so let's do a simpler example with two different appreciation scenarios. Scenario A: I buy a $100k house with $50k down and a $50k mortgage. House prices d
43.
by civilized 3y ago
Appreciation is a good deal for speculators and flippers, but for normal homeowners it isn't worth much. If all house prices double including my own, the money I make selling my house just goes into buying a new one.
44.
by civilized 3y ago
Target? Terrifying warfare connotations. Client? Smacks of historical patterns of subjugation (client states of empires). Where does this end? Are we to become linguistically homeless, evicted from our verbal environment every time someone
45.
by civilized 3y ago
How are folks so comfortable with "target"? Doesn't that have strong connotations of warfare and attack? (I don't actually care about this, just illustrating how much language is walled off by this kind of hypersensitivi
46.
by civilized 3y ago
I agree that CRDT spreadsheets like Google Sheets have similar features, but we only got those recently as well.
47.
by civilized 3y ago
I'm going to look on the bright side of this. Look at how far we've progressed in collaboration! We complain about Jira etc but it does have a lot of great features that an emailed, weekly manually reconciled Excel spreadsheet doe
48.
by civilized 3y ago
I suspect it's way less true than tech folks who hate the US defense industry think. The correlation between liberal opinions and problem-solving intelligence is nonzero but it's not all that high.
49.
by civilized 3y ago
Hosting is a long-solved problem. High-quality repositories have provided free, universally accessible document hosting to the academic world for decades. The arXiv is a nonprofit organization run by Cornell that hosts preprints, and I wo
50.
by civilized 3y ago
I would like to offer a different perspective than the one you are getting from a couple other folks. Not being in the room it's hard for me to be sure, but my guess is that both of the following are true: 1. The person you tried to he
51.
by civilized 3y ago
Both problems come from hiding assumptions in an attempt to focus on a conclusion, which is presumed to be the only relevant information. That's not a bad thing to do in general -- we always have to be selective what we communicate. It
52.
by civilized 3y ago
These people are AI "experts" and they're still unfamiliar with LLM abilities to detect jokes and incongruities? Every interested layman knew about this within 48 hours of ChatGPT being released. It doesn't imply higher-
53.
by civilized 3y ago
It seems premature to worry about graphs when most languages don't even have good support for tables .
54.
by civilized 3y ago
Follow-up: I did a little legwork and found some Julia folks who were also inspired to implement tidyverse as faithfully as possible: https://github.com/TidierOrg Here's how dplyr-style chains look in their system:
55.
by civilized 3y ago
I was excited about Julia too but it now seems to be a relatively niche HPC language. It's about saving CPU time more than user time. My sniff test for a successor language to R is whether it can replicate the tidyverse API with 100% f
56.
by civilized 3y ago
There's clearly been a ton of progress in this area; the only issue is that feature development is even faster :) I'll keep an eye out for specific issues that seem helpful to raise. The biggest one I have right now is a little ni
57.
by civilized 3y ago
> We found R has a really hard time integrating into data pipelines and was best used as a standalone tool by individuals In my org we have several 100% R teams (including mine) that have been developing and maintaining business-critical
58.
by civilized 3y ago
Also AFAICT the only one that is fully integrated with a general purpose programming language with an excellent IDE.
59.
by civilized 3y ago
Many of the most important time series prediction models are called "autoregressive", meaning they are regression models predicting the target from (prior values of) itself. This suggests that statisticians don't really share
60.
by civilized 3y ago
This is extremely interesting. You might be onto something.

More ›