Exactly. My ahme club do a 3 week free trial, nobody new is paying until the 4th week (and even then you have wiggle room), you can just stop showing up, nobody will charge you.
"...the company still hasn’t explicitly said what about the situation prevents it from offering Xbox game purchases like Steam and PlayStation already do..."
Though the real answer is already in the article in a different context: "...ending the requirement for apps to use Google Play Billing."
Asshole monopolist #2 is angry at asshole monopolist #3 that they are monopolizing too much of the money that should by all rights be monopolized by them instead.
As a DoF "hater", my problem with it is that DoF is just the result of a sensor limitation (when not used artistically etc.), not some requirement of generating images. If I can get around that limitation, there's very little motivation to maintain that flaw.
In the real world, if I see a person at the beach, I can look at the person and see them in perfect focus, I can then look at the ocean behind them and it is also in perfect focus. If you are an AI generating an image for me, I certainly don't need you to tell me on which parts of that image I'm allowed to focus, just let me see both the person and the ocean (unless I tell you to give me something artsy :)).
While you could look at DoF as a sensor limitation, most photographers use it as an artistic choice. Sure, I could take a pic at f/16 and have everything within the frame in focus, but maybe the background is distracting and takes away from the subject. I can choose how much background separation I want; maybe just a touch at f/8, maybe full on blue at f/1.2
If you have the camera focus on the person, they'll be in perfect focus. If you then have the camera focus on the ocean, it'll be in focus.
Our eyes work the same way. Of course, just like the camera's aperture can be set our pupils will be pretty contracted on a beach.
Of course, you should be able to tell the AI to generate it how you want - that's the goal, after all. Having at least a somewhat shallow depth of field by default makes sense though.
The way I heard it, trains in Japan aren't (state) subsidized, but rather they are real estate businesses where "subsidized" cheap traffic into and between train stations drives increased real estate value and thus commercial rents.
If you think about it, the same could be said of state subsidized public transport, where increased economic activity due to improved traffic (getting people to/from jobs, shops and their homes) can increase tax revenue which can then be spent on public transport subsidies, turning them revenue positive. Of course whether most state subsidized systems actually live up to those aspirations is a bit more questionable.
> These all suck, and the government generally collects money on assets as they move not assets at rest.
The government can also collect money on assets at rest (or at least, on cash at rest). They do so by creating money. It could be an interesting tax regime where the only forms of taxation are taxes to discourage action (e.g. tax on tobacco) and money creation.
> Yes, Chrome was absolutely superior to Firefox in every way at the beginning.
At the time of Chrome's introduction, Firefox still had XUL extensions and a somewhat different UI from Chrome, so Chrome wasn't strictly superior (though it is fair to say it had better performance than Firefox). Over time, Mozilla replaced XUL extensions with Chrome's extensions API (without providing meaningful additions, like reintroducing the ability to modify the UI) and reworked the UI to be a clone of Chrome, so thanks to Mozilla's efforts it might be fair to say that today Chrome is strictly superior (excluding privacy and MV3 etc.).
I actually prefer XFCE's scaling via setting a font DPI to e.g. KDE/Gnome's scaling which increases everything in size and not just fonts.
First because usually it is only text that I want to be larger so I can read it. Increases the size of "everything" just decreases UI density with no benefit (while things that conform to the text like button still increase in size to contain the text).
Second because scaling "everything" often leads to ugly results. E.g. I use a program to browse local media files that generates thumbnails for those files. The size of the thumbnail generated matches the size of the widget it displays for that file. If the widget is scaled 1.5x due to UI scaling, it will show a blurry upscaled thumbnail.
Looking at mean SAT grades is close to irrelevant when your (supposed) strategy is to pick Top-N candidates. If for example every single person got exactly the mean score allotted to them by their race, you'd actually expect that the top 5.9% of universities would all be 100% Asian, the next 57.8% of universities would be purely "White (incl. Jewish)", the next 18.7% of universities purely Hispanic and bottom 12.1% purely Black.
By that math, every single slot in an ivy league school "belongs" to an Asian candidate and even a single white person is already over-representation.
(Didn't separate White from Jewish since they don't have mean scores for just Jewish, possibly they should be first according to the estimate in the text below, but that doesn't really matter for the point made)
> Looking at mean SAT grades is close to irrelevant when your (supposed) strategy is to pick Top-N candidates.
Not if the mean SAT score predicts how many students surpass some academic cutoff used by the universities. It's like predicting who will win a best-out-of-3 100m dash, when all you have are the runners' mean 100m times.
I find it baffling how people become incapable of the simplest inferences, and capable of the smallest nitpicks, when they don't like where the data leads.
Obviously it depends on the exact test you are running, but a factor that is frequently ignored in A/B testing is that often one arm of the experiment is the existing state vs. another arm that is some novel state, and such novelty can itself have an effect. E.g. it doesn't really matter if this widget is blue or green, but changing it from one color to the other temporarily increases user attention to it, until they are again used to the new color. Users don't actually prefer your new flow for X over the old one, but because it is new they are trying it out, etc.
> trashy lies, very biased and informative. Misreporting what people actually said is a strong tell of the first two
For "biased" rather than outright lies, I think the two most common easily observable techniques I see are:
- Quoting known unreliable sources: "Black people eat human babies! (say KKK leadership)"
- Using passive and active voices to shift blame: "Police shoot and kill bystander during drug bust" vs. "Shots fired during drug bust fatally injured potentially uninvolved man"
reply