canjobear

It is indeed a very weird way of talking, characteristic of prepared statements and forced apologies.

“When I describe priming studies to audiences, the reaction is often disbelief . . . The idea you should focus on, however, is that disbelief is not an option. The results are not made up, nor are they statistical flukes. You have no choice but to accept that the major conclusions of these studies are true.”

Cats aren't as deeply domesticated as dogs but there's clearly been some breeding effect which you see when you compare a housecat against a true wildcat (felis silvestris). The main thing is that housecats are comfortable around humans as long as they're exposed to humans within a critical age window, even if they were born in the wild to a feral cat. In contrast, wildcats never become comfortable around humans and are effectively always feral.

Housecats also have a number of behaviors that seem only distantly related to things that would be useful in the wild, such as meowing for attention. Another weird behavior is when cats line up dead mice outside their owner's bedroom door like a present. There's clearly some relationship that they understand between themselves and their owner.

Fascinating. I didn’t know Aztec writing used glyphs for their phonetic value.

Indeed. Just looking at and zooming around is incredibly thought-provoking.

Is this the simplest construction of a self-simulating game of life?

It’s models all the way down.

There’s a long debate in neuroscience about whether information is encoded in timing of individual spikes or only their rates (where rate coding is a bit more similar to how ANNs work, but still different). It hasn’t been decided by any one paper, nor is it likely to be: it seems that different populations of neurons in different parts of the brain encode information through different means.

Does it explain double descent?

If so it would likely be a function of amount of training data in that language.

No, they’re multilingual.

I don't buy it. The word "heart" is literally cognate to "core". This suggests a very old view that the heart is central.

ChatGPT's got it.

Yo, listen up, as I paint the universe broad and deep, Sky's dark, Earth's yellow, secrets ancient they keep. This life, this hair, rooted in the core, Four elements, five virtues, what we livin' for.

Cities shining, from east to west we're ruling, Grounded in the earth, it's agriculture we're schooling. Lose yourself in books or market's charm, Life's treasures, keep close, safe from harm.

Sun and moon, their dance of light and dark, Nature's cycle, in harmony, leaves its mark. Respect for life, ain't no damage done, From mountains high to rivers deep, under the same sun.

etc.

It's still impressive---imagine writing 1000 words of English without reusing any function words like "the", "it", "of", "is". And just to show off, the Qianziwen holds a bunch of those words for the very last line!

It's like doing billions of code reviews based on billions of build artifacts, so that you start to notice that certain patterns are more common than others, and start to be able to reverse-engineer the compiler and source language. The best predictive model for executable binaries would be a generative model over possible source codes and then a model of the compiler.

The Bayes-optimal text predictor would have to implicitly model the generative process underlying texts, that is, human minds. So it’s not crazy to think this approach could lead to an intelligent model.

The people that don’t like IQ tests usually deny that “general intelligence” is a thing at all.

Compression and intelligence are closely related. http://prize.hutter1.net/

On the other hand they’re about as good as humans on a text version of Raven’s matrices, another IQ test. https://arxiv.org/pdf/2212.09196v1.pdf

Furthermore Copernicus’s system had lower accuracy and more epicycles. There wasn’t any real reason to prefer heliocentric models until Kepler.

Aristotle’s physics is correct for objects immersed in a fluid. https://arxiv.org/abs/1312.4057

You can get it with full gradient descent though... https://www.nature.com/articles/s41467-020-14663-9

Honestly the fact that there doesn't seem to be a good explanation for this makes me think that we just fundamentally don't understand learning.

Have they figured out what causes double descent yet?

Sometimes it's nontrivial to show that a theory predicts something, even if the thing is already observed.

The first step is to propose a model that explains stuff. Then you test it on new predictions.

If that’s true then it should be possible to predict the precession from special relativity alone.

The "when applicable" gets them out of nearly anything.

The Fremen in Dune are clearly inspired by Muslims but not necessarily Arabs. They are more closely modeled on Chechens. Fremen words like "sietch" and "kindjal" are from the Caucasus.

There has been tons of sound change over the last 100 years. In old radio broadcasts you hear lots of sounds you don't hear any more, especially in regional dialects that have either changed or been levelled out of existence. Check out this spellbinding speech from Calvin Coolidge in 1924: https://www.youtube.com/watch?v=xfF-xv7rc20

We've got...

* Non-rhotic dialect with "earnings" as [əjnɪŋz]

* Final /ɪ/ as in British English, not /i/ as in modern American English. "Country" is [k@ntrɪ].

* /p/ is clearly unaspirated

* "paid" is [pɛ:d], not [peɪd] as in modern American English

* "grant" is [grɑnt] not [grænt]

As another example, the Northern Cities vowel shift (for example "Chicago" pronounced [ʃɪkægo]) only really got going in the 1960s.

Would you say the Holy Roman Empire was really Roman?