canjobear

Do you have examples for complex grammar?

Extremely uncharitable reading. Plausibly they were in a foreign country where they didn't speak any of the language and didn't know how anything worked. This kind of situation was never easy for anyone.

The author is a high school student! Very impressive.

There's no problem reconciling the quantum with the Newtonian. Quantum mechanics recovers Newtonian mechanics in the appropriate limit. The problem is reconciling the quantum and the Einsteinian.

Right there in the second sentence. The sentence is ungrammatical because the n-gram model can’t handle long term dependencies. “To obtain such a large medical database of the joints are risk factors” — the verb should be singular and the predicate should be something else, like “to obtain such a large medical database … is difficult”

I've never encountered these words until now.

They gave the example of the verb yiwei in Mandarin. If you say “ta yiwei X” it means “s/he thinks X” with a strong connotation that X is in fact false. The Spanish equivalent is supposed to be the verb creerse [1], like if you say “Juan se cree que lo van a ascender” it means “Juan thinks that they are going to promote him” but with a strong connotation that he won’t in fact be promoted. English doesn’t really have a verb for “think” with the connotation that the belief is false. The claim (for what it’s worth, I am skeptical) is that English speakers are slower to learn the concept that someone can have a false belief, because English lacks such a verb.

[1] according to https://ojs.ub.uni-konstanz.de/sub/index.php/sub/article/vie... for example. I don’t know enough Spanish to say if the verb really works this way. Verbs like this are called “contrafactive”

In grad school I got to attend a talk by one of the researchers who was involved in the discovery of melatonin as a sleep aid for humans. He said that his team had hoped for it to become a prescription medicine dosed at 500 mcg, because anything higher gave paradoxical effects and actually made sleep worse. But it ended up being classified as a supplement in the US rather than a drug, so they had no way to control the dosage on the market.

The other useful thing I learned is that melatonin isn't primarily involved in falling asleep, its main function as a hormone is in staying asleep. I've started taking it sporadically if I wake up in the middle of the night, to make sure I get back to a deep sleep and stay there, and it seems to be super effective for this.

Who is getting rewarded for uploading tons of stuff to the arXiv?

Interestingly Columbine was also supposed to be a bombing, but it didn't work. I guess that was the tipping point.

What discoveries in anthropology make you think that cuneiform is unlikely to be the oldest?

Writing has only been invented independently a few times in history, so it seems reasonable that cuneiform could be the first.

Why would it not be? You can use big data and neural nets to fit Bayesian models (variational inference).

This is less because of any special attitude toward governance, and more because the only university degrees you could get during the Cultural Revolution were in engineering.

The other key element is holding kids back a grade until they pass reading tests, rather than allowing them to progress.

They have explicit quota systems and affirmative action for the official 55 non-Han minorities, and also for people from less wealthy provinces. Sometimes people fake residency in a poorer province to game the system.

Could this be solved by applying some small amount of noise to the image before downsampling?

> Even 6 year olds don't understand humour at all. Very similar to LLMs they know the shape of a joke from hearing them before, but they aren't funny in the same way LLM jokes aren't funny.

For further examples see a great deal of documentation here: https://www.tumblr.com/badkidsjokes

Seems like this was a better solution to the same problem https://www.evanmiller.org/attention-is-off-by-one.html

I think the point is that, despite the game being massively logically underspecified, people are still able to home in on an answer in practice. Which shows that they have a shared fuzzy semantic space.

On the contrary, academics love to navel gaze and complain about academia. What’s lacking is realistic alternatives.

I’m not surprised that Apple has been struggling to integrate LLMs. By their very stochastic nature they go against Apple’s philosophy of total vertical control of every aspect of their products. There’s no way to guarantee that an LLM will do anything and so they would be ceding control of the user experience to a random number generator.

The world that we need to be expressive about is changing.

This made me laugh out loud

This was absolutely not the norm I experienced in the 2000s.

It seems like if this really works, then it should do well on ARC.

Before ARR existed you could just send the paper in again and get new reviews.

I think it’s equally likely that the second version just got a different set of reviewers who randomly liked it more, and the revisions didn’t make a big difference. Having submitted lots of papers to conferences like this I basically think of the reviewer ratings as noise.

The technical claim about Inuit is indeed about roots, not words.

If there were theory that led to directly useful results (like, telling you the right hyperparameters to use for your data in a simple way, or giving you a new kind of regularization that you can drop in to dramatically improve learning) then deep learning practitioners would love it. As it currently stands, such theories don't really exist.

There’s a blog post series from a guy who did this.

https://vienna-pyongyang.blogspot.com/2008/04/how-everything...