andrewla

39 ms·

721.
by andrewla 3y ago
This would be a secondary source -- the primary source would be the manufacturers themselves. Secondary doesn't mean that there is no original research; in fact it's the opposite. Primary contains facts but no analysis or synthesi
722.
by andrewla 3y ago
That's one sense in which it is used, but the other sense is the opposite -- "no evidence that covid vaccines cause people to become magnetic" is not a "yeah, we haven't checked for the evidence yet, but we'll
723.
by andrewla 3y ago
"The author" here is none other than Scott Alexander, formerly of Slate Star Codex. For what it's worth, I wouldn't worry too much about his axe grinding; if there's anyone you can take at face value he would be the
724.
by andrewla 3y ago
The article addresses this exact point, and I would even go so far as to say that your conclusion, "'No evidence' is a phrase that gets used in untruthful ways" is entirely the point of the article. > Is there "n
725.
by andrewla 3y ago
If they do, then that is a significant defect in the testing itself. Most IQ tests are designed with the intention of being stable over time.
726.
by andrewla 3y ago
Yes. If I had access to the full article I would be able to perform a more nuanced analysis of their (likely garbage) methodology, but for the moment all I have is the abstract and there's already a significant error there.
727.
by andrewla 3y ago
My intuition is that this finding is correct. This is dangerous; in the social sciences, results that match your intuition will most likely have garbage methodology that does not support their conclusions [1]. I haven't read the study
728.
by andrewla 3y ago
While there are people that do this, I do not think that Whitney is one of them. This code is not obfuscated; it uses macros and strategically defined functions to allow writing code in a style similar to APL that appears natural (-ish?) to
729.
by andrewla 3y ago
I'm in way too much of a bubble. I worked at Microsoft for years and continued to run Windows for some time afterwards, sometimes dual booting with Linux. I've tried but always hated MacOS. But Windows 8 completely broke me -- thi
730.
by andrewla 3y ago
I think there's a clear differentiation between how we behave towards individuals vs. how we view a large-scale phenomenon. For an individual, the attempt to improve yourself is usually admirable. But I am cynical about New Year's
731.
by andrewla 3y ago
When my kids were smaller, they would often request that I make mazes for them on restaurant placemats, etc. Since I was working in crayon, many of the traditional maze generation algorithms, which relied on removing walls, were inapplicabl
732.
by andrewla 3y ago
I stand corrected. I made the same mistake that I am attributing to others; to think that n-bit arithmetic operations are O(1) instead of O(n).
733.
by andrewla 3y ago
This is frequently cited by law enforcement and criminal justice officials, but this is an impossible bar to set, because they have not been convicted of those other unspecified crimes. The excuse "he maybe didn't rob this store,
734.
by andrewla 3y ago
I don't think this categorization is technically correct, and even in the places where it is partially correct I don't see that the "off-by-one" phrasing adds any value. Off-by-one errors are a very specific type of erro
735.
by andrewla 3y ago
And since exponentation is logarithmic in the exponent, this algorithm is computing Fib(n) in O(log n) time. So while more compact than the matrix method or the field extension method or the doubling formula method, this reduces to the same
736.
by andrewla 3y ago
> The loss function is the feedback Please expand -- as I understand it the LLMs use cross entropy on next token prediction as their loss function and I fail to see how this gives any feedback about translation, even given parallel texts
737.
by andrewla 3y ago
For general-purpose LLMs, there is no indication given during training that texts are parallel -- while there are parallel texts in the training data, the model is not trained on generating one text from another. That is, it does not get an
738.
by andrewla 3y ago
Not necessarily -- the embedding into semantic space is largely unsupervised; LLMs do not learn to translate through explicit examples of parallel text. Largely it's just a matter of having a large enough corpus of whale speech in a fo
739.
by andrewla 3y ago
I was really hoping that this would be a more detailed breakdown of the different methods, like the egg article [1]. This just presented conclusions rather than trying a bunch of avocados and evaluating (objectively or subjectively) their s
740.
by andrewla 3y ago
KeePass is a password manager. People use other options because their usability is better. Backup & password generation seem to be the big ones. Secondary is sharing of password and cross-device sync. Third tier is probably browser inte
741.
by andrewla 3y ago
By 2005 Google had such an insurmountable edge that there was barely a discussion. Microsoft started development in what would become Bing before that, and despite all their efforts were never really able to catch up to anything that Google
742.
by andrewla 3y ago
That's not the usual sense of "moat". A moat is usually a structural or regulatory protection that prevents a superior product from competing without first crossing that moat. For example, once upon a time building a superior
743.
by andrewla 3y ago
> Eventually, Google figured out how to create a real moat, by using click data to improve search result ranking This is revisionism -- Google was far superior to any competing search engine long before Microsoft embarked on its search e
744.
by andrewla 3y ago
How does that reduce the incentives? Better to win a human-only division by cheating than to win the augmented division by playing legitimately. "Better" being a kind of sociopathic view of it, but certainly a player who wins a hu
745.
by andrewla 3y ago
I understand the desire to restrict your consumption to human generated content. But I don't understand what is meant by "human verified".
746.
by andrewla 3y ago
I think I know what "human-generated" means, but I don't know what "human-verified" means.
747.
by andrewla 3y ago
I don't know what it means to be human-verified in any useful sense.
748.
by andrewla 3y ago
If you follow the link they expand on the policy. I don't think they will enforce this; thus the "Please". They say that you can use AI tools for assist (but discourage it). Feeding the text into an AI and getting an answer t
749.
by andrewla 3y ago
A lot of comments in here seem to be saying that "setenv is rarely used, so we can make setenv threadsafe even if it is costly", but I think that's missing the point. The thread-safety is in the operators getenv/setenv (
750.
The complex William Jennings Bryan (2007)(ideonexus.com)
2 points by andrewla 3y ago|0 comments

More ›