jetrink

It is an accurate depiction of how Chicago police operated, unfortunately. In fact, one Chicago detective who tortured suspects went on to work as an interrogator at Guantanamo Bay[2]. It's terrible that the series would glamorize that behavior.

1. https://chicagoreader.com/news/the-police-torture-scandals-a...

2. https://en.wikipedia.org/wiki/Richard_Zuley

That sounds like a reasonable prediction to me if the LLM makers do nothing in response. However, I'll bet coding is the easiest area for which to generate synthetic training data. You could have an LLM generate 100k solutions to 10k programming problems in the target language and throw away the results that don't pass automated tests. Have humans grade the results that do pass the tests and use the best answers for future training. Repeat until you have a corpus of high quality code.

I just looked it up. The process involves modifying a fertilized egg or embryo to create a founder organism. The gene drive is designed to be inherited by more than 50% of the organism’s offspring and propagate through the population over successive generations. Because humans have fewer offspring and longer generation times compared to insects, it would take many generations and potentially hundreds of years for a gene drive introduced in a single human to spread widely through the population.

Jevon's Paradox: https://en.wikipedia.org/wiki/Jevons_paradox

"In 1865, the English economist William Stanley Jevons observed that technological improvements that increased the efficiency of coal use led to the increased consumption of coal in a wide range of industries. He argued that, contrary to common intuition, technological progress could not be relied upon to reduce fuel consumption."

> With web search, Claude has access to the latest events and information, boosting its accuracy on tasks that benefit from the most recent data.

I'm surprised that they only expect performance to improve for tasks involving recent information. I thought it was widely accepted that using an LLM to extract information from a document is much more reliable than asking it to recall information it was trained on. In particular, it is supposed to lead to fewer instances of inventing facts out of thin air. Is my understanding out of date?

> nor makes it easier to search for online.

Searching for 'pebble core 2 duo' already comes up with a page of results only related to the watch[1] (including this very comment thread, ironically[2].) Search engines are very good these days.

1. https://imgur.com/TE3aaGY

2. https://imgur.com/l4aBszK

> Strengthen immunity by increasing contact with natural environments

My mom is a family and children's photographer. She likes to photograph kids in their backyards or in local parks, sitting in the grass. She often encounters small children who are unnerved by the feeling of grass as they have never touched it before. Also, parents will ask if their child could sit on a blanket, because the ground is dirty. I'm really curious how this attitude started. It is so alien to me.

> Fitting 3 meals into an 8h window would be … tricky.

That's the main secret to time restricted eating, in my opinion: You don't have enough time to eat as much as you normally do so overall calorie intake tends to decrease. I think the other reason it's helpful for some people is that eating nothing for one meal takes less self control than restricting yourself to a small portion, leading to better adherence than normal calorie counting.

> they'd have to arrange some way for a photographer (or themselves) to take a better picture and then CC license it.

That sounds like something a publicist could easily do if it were a priority. Actors and publicists have no trouble managing their IMDB pages and IMDB even charges for the ability to contribute.

The article that follows that one is also very interesting and worth reading. It shows how a coastline that existed 100 million years ago and cut through modern day Alabama is still visible in demographic and election maps. (By a crazy coincidence, I just dug up that article and sent it to someone not even half an hour ago, since we were talking about geology and I remembered reading it. Then I go to HN and see this posted.)

1. https://starkeycomics.com/2021/06/11/how-a-coastline-100-mil...

This is not the type of asteroid that would end all life on earth. It has the ability to destroy a city, but even if it hit the earth, it would probably land in the ocean or a remote area where few people would be harmed. We should not be alarmed about a 2% x 0.28% probability that a city or town is destroyed in ten years.

The Hindenburg was one of only two passenger airships, so imagine if 50% of passenger aircraft eventually crashed. There was also no obvious route towards making them safer at the time. Even if the problem of flammability could be solved, they were very vulnerable to bad weather and it was still decades before modern weather forecasting.

Increasing hurricane risk yes, but also a dysfunctional insurance market. There's a business model where roofing companies will approach homeowners telling them they have hurricane damage and are eligible to receive a free roof. The hurricane might passed through years before, but any damage that was overlooked is still the responsibility of the insurance company. The homeowners are surprised to learn that their roof has damage, since they haven't noticed anything wrong (and there is in fact no damage), but free is free, so they sign the contract. The roofing company files a claim, the insurance company denies it, the roofing company sues, and since a lawsuit costs more than a roof, the insurance company backs down. Then after this has happened repeatedly, the insurance company raises their rates or leaves the state.

The linked book is Dealers of Lightning: Xerox PARC and the Dawn of the Computer Age by Michael A. Hiltzik, and I can second the recommendation.

And they didn't just say that NVidia was overvalued, but that it was overvalued for precisely this reason. I've also read people here and elsewhere warn that export controls could quickly become ineffective as Chinese researchers find clever ways to train models with lower-end hardware. I think if there had been a prediction market, this outcome would have been given a better than 50% probability here, though maybe the speed it happened has surprised people.

Don't underestimate an always on display. It's very nice to be able glance down to see the time or other info without making an exaggerated gesture first. It was what I appreciated most after I switched from Fitbit.

As to the Fitbit in particular, they just aren't well made anymore and have a habit of dying just out of warranty. My last one got stuck in a reboot loop 14 months after purchase. I tried everything with support from factory reset to letting the battery fully die, but nothing helped. Google told me to throw it out and buy another. My previous one had battery issues after 2.5 years, despite never letting it go below 20%. They are considered disposable devices now and aren't built to last.

When Microsoft decided to advance straight to Windows 10 from 8, it also necessitated skipping the 2015 fiscal year. They have been stuck a year in the future ever since.

Look at what foreign adversaries are already actively doing: working to turn Americans against each other. Social media is the perfect tool to spread discord. Russia has troll farms that create fake news stories, manipulated photos, and incendiary memes targeted at both sides of the political spectrum. They've even orchestrated in-person protests and counter-protests to those protests, though those efforts have been less successful. Now imagine that instead of merely using fake user accounts to this end, an adversary controlled an entire social network, including its algorithm and its content guidelines and could tailor manipulative content on an individual basis.

Stephen Fry is one source of the story.

> I like to claim that I bought the second Macintosh computer ever sold in Europe in that January, 30 years ago. My friend and hero Douglas Adams was in the queue ahead of me. For all I know someone somewhere had bought one ten minutes earlier, but these were the first two that the only shop selling them in London had in stock on the 24th January 1984, so I’m sticking to my story.

1. https://www.stephenfry.com/2014/01/mac-at-30/

All clients today (and probably back then) have options to limit bandwidth consumption including throttling, scheduling, and total data transfer caps. For serving mostly HTML and images, dedicating even 10% of a home broadband connection to serving content would allow many, many people per day to access archived pages.

I think we need a version of the New Battery Technology Checklist[1] for this type of article. (Though I understand that the research itself often just aims solve specific tensions in cosmology, and it is only the reporting that over-hypes it as a full replacement for dark matter.)

Dear alternative dark matter theory claimant,

Thank you for your submission of a proposed revolutionary theory to replace dark matter. Your new theory claims to be superior to dark matter models and will transform our understanding of the universe. Unfortunately, your theory will likely fail, because:

[ ] It cannot explain galaxy rotation curves across all galaxy types.

[ ] It fails to account for gravitational lensing observed in galaxy clusters.

[ ] It cannot explain the Bullet Cluster observations where dark matter appears separated from normal matter.

[ ] It is inconsistent with the cosmic microwave background anisotropies.

[ ] It cannot explain the large-scale structure and formation of the universe.

[ ] It introduces arbitrary parameters without physical justification.

[ ] It lacks a sound theoretical foundation or violates established physics principles.

[ ] It fails to explain the observed velocity dispersions in dwarf spheroidal galaxies.

[ ] It cannot account for empirical relations like the Tully-Fisher relation.

[ ] It cannot be tested or falsified by current or near-future experiments.

[ ] Your claims are unfounded or exaggerated.

1. https://news.ycombinator.com/item?id=26633670

It probably doesn't work if you're primed to look for hidden details. I took the test along with my Psychology 101 class of about 30 people and no one noticed anything amiss.

A friend who lives in North Carolina sent me a video of the raging floodwaters in his state- at least that's what the superimposed text claimed it was. When I looked closer, it was clearly an Indian city filled with Indian people and Indian cars. He hadn't noticed anything except the flood water. It reminded me of that famous selective attention test video[1]. I won't ruin it for those who haven't seen it, but it's amazing what details we can miss when we aren't looking for them. I suspect this is made even worse when we're casually viewing videos in a disjointed way as on social media and we're not even giving one part of the video our full attention.

1. https://www.youtube.com/watch?v=vJG698U2Mvo

I think Missouri's policy is the most correct.

> Missouri prohibits such ballots from being counted if it is proven that the voter died before polls opened and the ballot envelope has not yet been opened.

It seems reasonable to me that to be an eligible voter in an election, you must be alive. A ballot filled out in advance is merely a piece of paper until it is officially cast by an eligible voter on election day. If a voter dies before the polls open and there is a system in place to detect this, their vote should not be counted because they are no longer eligible. Conversely, if the voter is alive even one second after the polls open, their vote should be counted.

Ctrl-F in that document for 'hashing'. That step reduces the audio information to a sparse collection of key points, one for each of four frequency ranges per time segment. I would assume that everything up to that step is done on the phone and only the key points are sent to the server.

It's pretty common in the US for a name that is traditionally associated with boys become unisex (Taylor) or even prominently female (Lauren). Typically it is an uncommon male name (so not, e.g. John or Brian) that sees a sudden increase in popularity and becomes unisex. Some other examples include Leslie, Casey, Jordan, Morgan, Addison, and Riley. I can't think of any examples of the reverse off the top of my head.

A chart of this phenomenon for Taylor:

https://engaging-data.com/baby-name-visualizer/?n=taylor&sex...

The music fingerprinting on my Android phone works in airplane mode, so it would be possible with modifications. Also, it's likely that Shazam is sending a "hash" of the audio rather than an audio stream in most cases.

I loved living around the corner from a small grocery store. It made it possible to pick up ingredients and fresh vegetables for dinner on the walk home from the train or when taking the dog for a walk with my wife. It was a bit more expensive, but that was offset by the fact that buying only what you need for the next couple of days leads to less spoilage and waste. Overall, our food budget has remained about the same since we had to move away.

Where I live now, it takes 15-20 minutes to drive to the grocery store, so it's only practical to shop once per week. Shopping is a big chore and a fight with traffic, not a nice walk. It means I need to dedicate more of my apartment to food storage and more of my time to meal planning. It's certainly not an efficiency gain for me personally. On the plus side, the large store does have 73 different flavors of potato chips.

> I it's fair it's fair to claim that by definition, if AI can make it all by itself, then it's not good art, because if AI can make it all by itself, then it's already in the "been there, done that" pile. As a tool however, a good artist can use AI to make good art.

A scenario that blurs the line between these two is where an AI is used to randomly generate 100, or 1000, or 100,000 images and an artist (editor?) chooses one for some artistic purpose. If the generation process is random enough, then the AI is bound to create something new by accident eventually, just like a room full of monkeys banging on typewriters will eventually write Hamlet. At the end of this process though, did the AI create art by itself through randomness, or is it just a special case of an artist using a tool and the act of choosing is what created the art?

To me the Seinfeld project was funny in the same way that Markov Chain generators often are: It was uncanny, absurd, and filled with non sequiturs. That's not to say there wasn't artistic merit in the project as a whole, but the model's contribution was only ever unintentionally funny and I never felt I was laughing with the model, but at it.