schoen

70 ms·

871.
by schoen 3y ago
Adding a large language model is going to move you in the exact opposite direction from ease of formal verification.
872.
by schoen 3y ago
I was told a few years ago that bean-to-bar was done by very few companies. But I've since heard that it's had kind of an explosion that somewhat parallels the popularity of beer microbreweries, with a huge boom in sales of chocol
873.
by schoen 3y ago
The response from Mozilla here indicates that this is a misconception or a series of misconceptions.
874.
by schoen 3y ago
My conceptual understanding of language model sycophancy is something like this: These language models are predicting the most likely conclusion in a context . The most likely conclusion in most linguistic contexts is more like agreement t
875.
by schoen 3y ago
Basically, by putting a relatively small number of adversarial examples into the training data of a text-to-image model (that don't necessarily look suspicious to a human observer), they can make it completely mislearn a concept.
876.
by schoen 3y ago
There can also sometimes be more specialized reasons than the general considerations given in this article. I forgot now whether it was DH or RSA or both, but there was a toy attack where interacting with a peer with super-bad maliciously-c
877.
by schoen 3y ago
From the headline, I was totally thinking of this as like a 3D printer and then thinking "there's no way there's a feedstock for water that's cheaper to make and transport than ... water!". But it turns out it'
878.
by schoen 3y ago
> There are no controls because there is no possibility of controls, as a matter of mathematics. This is kind of true because, for example, the one-time pad is information-theoretically secure, and anyone could choose to use a one-time
879.
by schoen 3y ago
Where would the tunnel connect to on the west side? While there seem to be many villages and resorts along the west side of the gulf, I don't see that any of them are connected to Panama's road network. I don't see road con
880.
by schoen 3y ago
Since English "Columbia" has the same meaning as Spanish "Colombia" (namely something like "Columbian (land)"/"(land) of Columbus"), I'm curious why the former didn't ultimately end u
881.
by schoen 3y ago
Wikipedia says that both Voyagers are calculated to be further away than Pioneer 10, but for Voyager 2 only since July 18 of this year.
882.
by schoen 3y ago
Multiperspective validation, if the attacker is far enough away from the target site. :-) I think some of the researchers who wrote about BGP spoofing attacks against Let's Encrypt may have suggested something about logging BGP changes
883.
by schoen 3y ago
Let's Encrypt uses "multiperspective validation" to prevent a single backbone router or backbone network from being able to do this attack in many cases. This doesn't help much if the attacker is sufficiently close on th
884.
by schoen 3y ago
This criticism/argument was, in fact, made against domain validation (DV) when it was first introduced by earlier CAs. It's not an unreasonable criticism, but the economics of PKI mean that we would have much, much less encryption
885.
by schoen 3y ago
Just to be clear (with better details already given by other comments), addressing this is specifically part of the threat model for Certificate Transparency. You could argue that CT policies should be strengthened in various ways, but the
886.
by schoen 3y ago
Other entities that publish transparency reports and that have received NSLs have reported them in bucketed ranges. I forgot the exact granularity. It does not appear that NSLs compel arbitrary actions or require the recipient to actively l
887.
by schoen 3y ago
Full-text PDF: https://onlinelibrary.wiley.com/doi/epdf/10.1111/meta.12599
888.
by schoen 3y ago
I'm confused at why these authors start their headline with "Victory!", since the Supreme Court's action was to grant the petitioners' motion for a stay and petition for certiorari. https://www.supremec
889.
by schoen 3y ago
(2016)
890.
by schoen 3y ago
Wasn't there a deprecation of the X- prefix in protocol headers? Other thoughts: You could probably make this play better with HTTP libraries by returning the header on an existing HTTP message, instead of changing the overall reply se
891.
by schoen 3y ago
"How to use the innards of Python and Pandas"? "If you find yourself fighting with Python and Pandas, I'll help you improve your technique"? "Using Python and Pandas in the wild"? "10 unexpected uses
892.
by schoen 3y ago
If the proof is written in a formal language, we have that now! Even several directly competing efforts (for better and worse). https://www.cs.ru.nl/~freek/100/ Some people greatly hope that fully formal proofs be
893.
by schoen 3y ago
> Or there was the great PLUR debate of 1994 when peace, love, unity and respect fell out of favour. Maybe this is deeply familiar to people who were adjacent to it, but how did this happen, and what replaced them? Were these things a ki
894.
by schoen 3y ago
Published almost a decade ago! Yikes.
895.
by schoen 3y ago
This is very similar to how Latin spelling reflects consonant assimilation of prefixes (like in- and ad-). https://en.wiktionary.org/wiki/immanis#Latin https://en.wiktionary.org/wiki/alloquor#Latin
896.
by schoen 3y ago
Although I'd say with something of a more malicious and/or mischievous connotation.
897.
by schoen 3y ago
If you're still interested in this, Tom Davies (Geowizard) has done the straight-line crossing in a couple of different countries and posted videos, both of some of the planning and logistics and of actually doing it. Two of the most r
898.
by schoen 3y ago
Eric von Hippel at MIT has written a bunch of business-oriented research along these lines (related to "user innovation"). https://en.wikipedia.org/wiki/Eric_von_Hippel
899.
by schoen 3y ago
Wow, TED was so different then!
900.
by schoen 3y ago
Thanks! That also raises a different question about people's introspection and ability to accurately answer questions about their utility and preferences, but that's a bit far afield from the mathematical question.

More ›