dwallin

A brief search came up with trains being around 15-20x safer than driving. As with many things it matters a lot how you count but it’s a pretty huge gap.

Do you have any additional resources you recommend on this topic? I've experienced life long restless legs alongside issues with sleep quality, daytime tiredness and low energy levels so this seems like an interesting avenue to explore.

It doesn't have to be a lie to be a poor excuse. The technical limitations could be something along the lines of they couldn't find a way to provide an app that wouldn't allow third parties to bypass apple gatekeeping of apple watches. Technically truthful, but in an antitrust context maybe less of a valid reasoning, but still allows them to redirect the public narrative.

Edit: Sounds like it was about privacy concerns, so probably something like there not being a completely foolproof way of preventing a third party from grabbing health data from the android app.

As mentioned in other replies, this doesn't fully solve a lot of the cases. I think z order curves might actually be better here, they are slightly less accurate individually, but in a way that would seemingly minimize the worst case results when taking the minimum of the two. But at that point you are probably better off using a single z-order curve and using the coordinates to check the manhattan or Chebyshev distance using bitwise operations.

There's very little actual evidence for modern day patents being a value-add to society. see https://www.sciencedirect.com/science/article/pii/S004873339....

Particular industries, eg drug patenting, could be far better addressed with targeted legislation but at this point it's mostly just momentum and vested interests. It provides huge benefits for a small group and diffused costs. It's hard to imagine it being possible to get enough momentum to change things.

I thought they pretty clearly explained where the additional performance came from. If there is only 1 valid schema confirming option you can skip the LLM entirely. If there are only a limited number of possible tokens (eg. only } or ,) then you run it on smaller subset of the model. Between these two you capture a large amount of the actual token count of most json.

This seems like a very practical and well thought out approach. Turning unstructured data into valid structured data is in my experience one of the most important things for integrating LLMs deeply into a pipeline. Doing that fast and cheap goes a very long way for these use cases. Also, if you need stronger content generation than this, nothing prevents you from generating some higher quality content in another LLM and then passing it through this to structure it.

Perhaps a controversial opinion, but to me this really just shows the major flaw with our current text-file-based model for programming languages; it mixes the actual content of our programming with its presentation. It's time we begin to move past text files for storing our programs and move to directly using a shared intermediate representation (something along the lines of an AST). Then we can move past all this nitpicking about code style, indenting, etc.

The undersupply of 3 bedrooms is real, in many neighborhoods in my city you just can't find them. Anecdotally, the lack of 3-4 bedroom apartments has been the major driver of couples I know, who were aiming to start families and who would otherwise love to stay in my city, leaving the city for the suburbs. 3-4 bedroom units are also popular with younger folk, who are often happy to have roommates, and share amenities in exchange for lower rent.

> There's a shortage of detached single family homes in desirable cities and suburbs.

You have this backwards, it's essentially physically impossible for this to not be the case. Past a certain point you just cannot squeeze more detached single family homes into a reasonable distance from a city. Single family homes, suburbs, and the required car-centric transit they require are massively space and transit inefficient. If you want there to be affordable detached single family homes within a reasonable distance to a desirable city your best bet is to push for increased density within and around the core of the city, with walkable streets and excellent public transportation. The increase in livability and affordability in the center encourages more people who might otherwise be pushed out to stay and leaves more single family homes for those who really want them.

I think the article gets it wrong when it says nationalism is the biggest issue the UBI movement currently faces. Our current waves of nationalism have populist roots, and therefore is not inherently incompatible with UBI policies. However, as stated in the article, the current UBI movement is generally left-of-center; and there is a lot of reactionary opposition (atleast in the USA) to anything that comes from the other side.

I think the bigger issue is that there is a common, albeit misguided notion that UBI would create inflationary pressure. This fear was made worse by the current inflationary environment. Covid relief was made the boogeyman for inflation out of political expediency.

On the plus side, the advancement of AI, and the threat of widespread job loss creates a positive talking point that can possibly help counteract that.

You need to look at second order effects and how AI might intersect with other emerging technologies. There will be fundamental shifts that change market dynamics, we just don’t know how they will happen yet. Incumbents have a strong belief in AI being transformational but since no one truly knows how those changes will be realized, so they are mostly all hedging their bets, investing in institutional knowledge and low-hanging fruit. They cannot generally afford to make a “big bet”. If you can develop a strong hypothesis about a non-obvious shift you can get ahead of the curve in a significant way on an emerging, or transforming market. Look for things that are valuable and/or expensive that might become a commodity. Look for technologies, business models and approaches that might be considered inferior or non-viable by incumbents, but that might experience synergies with an ai-powered future that changes the base assumptions.

I think the point stands. Education at the time was highly discriminate regardless of who was considered the in-group and who was the out-group by any particular country.

> A hundred years ago college was for a rare breed of scholar.

A hundred years ago college (and especially elite colleges) were largely for white upper-class men so I would be very careful throwing around words like "rare breed". Education in general was something reserved for the elite, why would you bother educating someone who would probably be working a factory line?

The general population is far more educated, providing massive well-documented and widely-studied societal benefits. What has changed is we have moved (slightly) away from viewing education as a way of stratifying our society and more towards it as something that should be a baseline for our citizens.

> You can't legislate that someone lose more money than they gain with access to your market.

Apple would offer an app store in some form even if they were statutorily prevented from making any revenue from it because if they turned off the ability to install apps people would abandon iPhones in droves.

I put together a simple gif to roughly demonstrate the effect: https://giphy.com/gifs/JImONk8V5YK6iGBFkH/tile

While it still can't fully resolve shapes smaller than a pixel, it definitely helps with identifying forms.

One thing the article doesn't get quite right is you can't compare ppd of a virtual display directly to an IRL screen. The virtual pixels are not 1:1 to device pixels. Your perspective is constantly shifting slightly and the virtual pixels being sampled from is constantly changing. This is providing your eyes and brain a lot more data to sample from and means the perceived resolution can be higher than the actual resolution.

JetBlue doesn't fly any Boeing planes so as long as it isn't a codeshare flight you should be great.

> You create a device, and all the dev tools around it, should a government be able to force you to give it away and support people using it for free?

This is an overly broad straw man argument because no one is trying to force Apple to give away iPhones. And yes, the government should be able to force you make major concessions, especially if your device becomes integral to the daily lives of half the citizens of said government.

Apple is selling these devices at a significant profit to end users. The primary purpose of the app store for Apple is to provide value to those consumers, so they continue to buy iPhones. Even if the EU decided tomorrow to confiscate all profits Apple makes from the app store in perpetuity Apple would continue to offer the App Store.

Yeah, it's 100% a poison pill. They clearly tried to craft an alternate TOS that would look reasonable at a glance to regulators and/or judges but that almost no serious companies would accept.

Even if you are just a small indie developer that's far too much risk for anyone offering free apps (or that might want to in the future). Imagine being the next flappy bird (over 50 million downloads before it got pulled) and suddenly being on the hook for $25,000,000.

I think this point is going to be the biggest sticking point to EU regulators. Third party app stores not being able to reasonably distribute free apps is a massive competitive disadvantage.

Is this paper achieving what you mean: https://arxiv.org/pdf/2311.16728.pdf

Caveat, this field is far outside my wheelhouse. So if not, I'd love to understand more about the nuances here.

You could probably already make something compelling like a small myst-style game that uses largely static assets with existing tools, taking advantage of being able to render real world environments with high fidelity.

However for more typical games there's still a long way to go. Most of the research focus has been towards applications where existing mesh-based approaches fall short (eg. photogrammetry), but this isn't really the case for modern game development. The existing rendering approaches and hardware have largely been built FOR games and leveraged elsewhere.

Rebuilding the rendering stack around an entirely new technology is a tall order that will take a long time to pay off. That being said, the technology is promising in a number of ways. You even have games like Dreams (2020), which was built using custom splat rendering to great effect (https://www.youtube.com/watch?v=u9KNtnCZDMI).

It ignores it, but not in favor of the bottom 90%, it looks like the numbers for % of pension entitlements have dropped significantly for the bottom 90% and went up only for the top decile.

50-90% https://fred.stlouisfed.org/series/WFRBSN40178

1-50% https://fred.stlouisfed.org/series/WFRBSB50205

90-99% https://fred.stlouisfed.org/series/WFRBSN09151

The choice of 10% is extremely peculiar.

Top 1%: 40% to 54% -> +14%

Top 10%: 77% to 93% -> +16%

Top 10% minus the top 1%: 37% to 39% -> +2%

Essentially this was all the top 1% and the rest of the 10%ers barely budged in ownership percent. I'd be curious to see the results for top 0.1% and 0.01% as well.

[edit] Found the source numbers:

50-90% https://fred.stlouisfed.org/series/WFRBSN40172

90-99% https://fred.stlouisfed.org/series/WFRBSN09145

99-99.9% https://fred.stlouisfed.org/series/WFRBS99T999258

top 0.1% https://fred.stlouisfed.org/series/WFRBSTP1285

Bottom 50% barely registers, hovering below 1% https://fred.stlouisfed.org/series/WFRBSB50199

It looks like the top 0.1% barely nudged in the past two decades as well. I would guess that the top 0.1% is probably hard to account for given all the ways the wealthiest can distribute their wealth to make it opaque and hard to granularly track. Those numbers could reflect changes in choices of financial vehicles for the wealthiest households.

This is a contract violation, not a copyright issue. But if you share your audio with someone who isn’t party to this agreement, there is likely nothing stopping THEM from using it commercially.

Not all code has equivalent carrying costs. I think the ideal behind tests is that they reduce maintenance costs of the code they cover by more than they add themselves (ideally).

That seems disingenuous. No doubt there are some car owners who do care about these things, but the vast majority of drivers would likely be hard-pressed to know what type of engine their car has, or what the differences between rear and front drive are. Most people select cars based on some mix of cost, perceived safety, utility, and by what other people they know are getting.

(edit) Went to add sourcing and was surprised to find that while most of the things I listed were accurate, fuel efficiency turns out to be one I significantly underestimated: https://www.statista.com/chart/13075/most-important-factors-...

54% of the US market makes them a pretty clear market leader. https://www.statista.com/statistics/266572/market-share-held...

I can't help but feel that Stability AI is pushing themselves to irrelevancy with their business model and new policy of "research-only" models. Part of what made the original stable diffusion models such a force was the ability and willingness of others to build on top of it. This is such a force multiplier. People built all sorts of abstractions around it that it became something much larger than just the original image generation.

Locking down your models so you can sell a mediocre api and web app is not the business model you want here as a mid-sized player. You just won't be able to keep up with the biggest players and their in-house models.

I think the open license they should try using should be something akin to GPL, where derivative models need to be shared and released under the same license. This gives you legal ammunition against companies building upon your models for their own private service, while allowing for a vibrant open-source ecosystem. Then you can offer more permissive licenses to those who who can pay.