glup

Placidly uncaring since long ago as I stopped consuming media full stop.

Yes! This seems to be a really neat combination of 2010's Bayesian cleverness / Tenenbaumian program search approaches with the LLMs as merely sources of high-dim conditional distributions. I knew people were experimenting in this space (like https://escholarship.org/uc/item/7018f2ss) but didn't know it did so well wrt these new benchmarks.

There's this one GitS (SAC I think) UI thing that I think about pretty often: a bot needs to work faster so its fingers pop open revealing a huge number of additional cursed sub-fingers, such that it can type on a keyboard massively faster. I love thinking of this as such a wonderfully naive early 2000's view on HCI — there's no direct-access data API, the bot needs to use the keyboard so.... more actuators!! How much of technology has this same flavor?

Nah, you have to write cute little LLMs that grind leetcode for you...

And that’s an inclusive “or”, mind you-- I remember an Al-Jazeera news story that opened with the line “Some say the new measures go too far… but do the new measures go far enough?”

I'm missing how this is an advance over https://www.pnas.org/doi/10.1073/pnas.2105646118

This is a near-meaningless statistic for evaluating the health of the environment, obviously. Maximizing the coverage of intact ecologies supporting biodiversity should be the management objective. The problem here is that you can plant more trees, but the forest is monocropped and doesn't support the rich network of species it did previously. Hence, the importance of conservation (and corridors between protected lands) rather than restoration.

It's not pretty, but, everything else seems like a toy for children by comparison.

Note the question mark in the OP’s comment. Also, pretty sure the e-ink limitation is common knowledge on HN. Yep, HN is amazing.

This sort of reminds me of what happened to the Slashdot... everyone was playing video games, hacking with Arduinos, doing silly things with homelabs, arguing about OS X vs. Linux, etc. and then suddenly so many people in the community moved up through the ranks and started to interface with politics, policy, and international business, just as those things started to really change with tech. So ya, tech overflows itself.

In light of this pretty natural scope creep, I agree with the above point that the moderation feels pretty consistent, and am, for one, extremely appreciative of Dang's work.

I was vacationing in Mammoth one time and started reading about the geology... quite a shock to find that basically everything to the east is a giant caldera [1]. Also 760,000 years ago is really quite recent on geologic time scales.

[1] https://en.wikipedia.org/wiki/Long_Valley_Caldera

Besides the author’s v points about the poor working conditions, I was struck by the points about transit costs. I am curious how much of the current labor situation in the US can be explained by the combination of 1) geographic wealth segregation —- richer consumers clustering together while generally poorer service workers have to live far away —- and 2) high gas prices / underdeveloped public transit. At some point as both of these factors increase, there should be a big drop off in available workers. Is this consistent with the pattern we are seeing?

Obfuscating the total cost when the consumer is making decisions—-the menu price is like the listed price in Google Flights, then taxes and tips are kind of like the baggage and seat fees on low cost airlines.

Kiwi used to have this in their map view -- you could select a very large circular region regardless of country as the origin -- but it seems like they got rid of it.

This was by far the most useful tool to identify cheap segments and cut down the space of options.

If it doesn't suffer from gimbal lock, is it even really a monitor?

“It is difficult to get a man to understand something when his salary depends on his not understanding it.” -Upton Sinclair

Risk $1.3b on an unproven design by an eccentric amateur that has some pretty clear shortcomings in "user experience" for the *vast majority* of users? I can't believe UCSB leadership is still seriously considering this rather than seeing what a risk this is. Pro tip: build some condos with windows.

The most sophisticated resource I know of this sort is https://littlesis.org

Similarly naive follow-up: how much hay is needed per animal per month, assuming no other sources of food? (or does one always provide hay plus something else?)

I took up large format film two years ago and I’ve found it to be extremely rewarding. Film has a different response to light than digital sensors, the size of the negative lets you do things you can’t with digital, and then there are camera movements (tilt and shift). Developing film and printing are also very fun, and feel a lot more like cooking.

This was originally reported by Bloomberg and then the Boston Globe, https://www.bostonglobe.com/2021/07/29/business/wealthy-amer...

To my knowledge, Bermuda Post is not a real journalistic organization in Bermuda (the major paper is the Royal Gazette). Instead, checking their about page reveals:

'BermudaPost is a non-profit, private and self-funded, commonly-created News and info-sharing platform that enables everybody to share valuable content such as local and global news updates.'

Besides two suspiciously barebones property listings, there are no references to Bermuda.

I find nothing more disturbing than an invertebrate eating a vertebrate. It happens but it shouldn’t.

I find it interesting how this is premised on the "one big claim per paper" model, which for better or worse seems increasingly common. In my experience a lot of 60's-90's research is of the form "the results of our first study were consistent with [some specific paper], and our second experiment was not." This nuance is pretty hard for people to keep in mind, and often seems to get lost over time (as a postdoc in cognitive science at MIT, I've noticed a big difference between what papers actually say, vs. what people say those papers say 30 years on). I'm excited to see how well Scite actually works as I test it over the next week!

Exoornithologists disagree.

Completely inedible.

A lot of academics and scientists I know dislike TED talks. First the name means (T)echnology, (E)ntertainment, and (D)esign -- which are all proximal to, but quite different from, actual science. If anything, it turns science into consumer products, and gloms it with self-help and entertainment in a quasi-megachurch format (hint: "innovation" is the savior; and everything is cast as innovation, esp. technological). Second, it induces a race to the bottom in terms of simplification and overselling conclusion / significance (compare with the perspective that knowledge production requires deep backgrounds to truly understand, and that the results of most experiments / analyses / models of the real world are fairly nuanced). Third, the stuff that comes from their franchising (TEDx) contains some real garbage.

No.

I've read that quality is largely limited by optics, and optics are limited by space in the MacBook line (just a few mm). Not sure why Apple wouldn't improve it on the iMac (a few cm depth to work with!).

In linguistics, lemma is also a term of art.

I want to grow up and be this ship.