HN user
And we should let them all in - the people who are leaving Russia now have an ardent desire to escape what is fast becoming a totalitarian state, while the Ukrainians are, of course, fleeing an invading army that has shown callous disregard for their lives.
It's worth replying specifically to highlight how wrong the original comment is, but the evidence that was cited by the Russian government to justify their invasion was clearly fabricated.
For instance, an alleged car bombing in Donetsk right before the invasion was staged using cadavers,[1] and a video used to claim that Ukrainian troops were moving aggressively into separatist territory was filmed far from its purported location.[2]
If Russia was so sure that Ukraine was committing these atrocities, they wouldn't need to rely on fake videos to justify attacking.
[1] https://www.bellingcat.com/news/2022/02/28/exploiting-cadave...
[2] https://www.cnn.com/2022/02/22/europe/russia-videos-debunkin...
I think the key distinction is that, for the most part, the current generation of cruise missiles are targeted and launched with a "human in the loop" - that is, a person actively makes the decision to fire the missile at a specific target. However, the autonomous weapons being debated have significantly less direct human control - they are potentially designed to simply patrol a certain area and attack anything that the weapon classifies as an enemy.
Of course, there are grey areas here - certainly there are existing missiles that can be launched without a defined target and programmed to aim for anything that, say, has radar emissions that match known enemy systems.
"Omicron" also has the advantage of being easier to hear and understand in english - "Nu" can cause confusion because it sounds so much like "New".
This is an intriguing concept, but given that the submarine is small and suitable for mostly high-value cargo (whisky is the example cargo given in the article) I'm not sure how it'll offer significant advantages over other forms of transportation in practice.
Although submarines are more effective navigating through storms, one could simply wait for the storm to pass when shipping non-time-sensitive cargo, then use a regular surface cargo ship (which could be automated if desired just like the submarine). Surface ships also have the advantage of compatibility with our already-established shipbuilding and maintenance infrastructure, while a submarine would require the proliferation of new skills and tools to support it.
For time-critical cargo (where one can't wait for a hypothetical storm to pass), it's likely aircraft would be a better option for most shipments - certainly in severe storms aircraft can't operate either, but in that case the very act of loading and unloading the submarine would be hazardous as well.
I think that a lot of this makes sense from Google's legal perspective, where antitrust litigation is a constant consideration and any internal document mentioning market share or competitors could be used against them.
I'm sure that there is a great deal of discussion about potential anticompetitive issues within Google and with their outside counsel, but in a context where legal privilege protects against disclosure.
I think it'd be risky for Carvana if they tried "cornering the market" on used cars - unlike houses, used cars fall in price pretty quickly and new-car production has the potential to increase as car manufacturers respond in a way that housing production does not. If Carvana buys up all the inventory to drive up prices they'll need a plan to unload as well - while keeping prices high.
I think the challenge is that if the rewards were high, Twitter employees (with the advantage of inside information) might be tempted to "tip off" an outsider in exchange for a cut of the reward, rather than just reporting the issue internally.
At the same time, there isn't much of an outside market for algorithmic bias info in the same way there for security vulnerabilities. Probably the biggest effect of this reward will be to pull some grad students who were going to study algorithmic bias anyways towards studying Twitter specifically - after all, there aren't any rewards for studying the algorithmic bias of other companies!
I think this is a really good point, and I think that if anyone is really committed to promoting free-speech-maximalist approach to the web they should be focused on building tools that make is easier for people to host and distribute their own content without relying on a centralized service.
Any business with the technical ability to censor what they host is going to be tempted (and likely pressured by other actors) to take down content that people find objectionable. Removing these "chokepoints" where a small number of people have the ability to engage in mass censorship is key if you want to promote more diverse speech on the web. (Not everyone has this goal!)
The critical thing is that this also requires a growing population - which is not guaranteed in all places/timeframes!
I think the most concerning element in the ProPublica piece is that there might be individuals inside the government who are willing to exfiltrate extremely sensitive private information and hand it over to outside parties without any public process/consent - this seems like a major privacy risk, and even more so than other entities/corporations one deals with it's very hard (for understandable reasons!) to prevent the IRS from collecting and retaining personal information, and tough to keep them accountable for the information they do collect.
Having the pay ratio be relative to the "Median San Francisco worker" might have some interesting effects - for financial/tech firms who want to avoid the tax they might just shift their ~25% lowest paid workers to an office in Oakland (though any corps with a mandatory physical presence like retail won't be able to do this).
I'm a big fan of ISL - one of the best intro machine-learning oriented textbooks out there IMO. If you're looking for book that still offers a broad survey while going a bit deeper into the math, I recommend Elements of Statistical Learning as well (they share 2 authors):
I'll add that my main hope is that the scale required for launching large satellite constellations will also make it cheaper to launch space-based telescopes, but it's far from clear that would be a major benefit (after all, launch costs will probably be on the order of 1% of the JWST's total ~$10B cost).
I'd note that while a broken unemployment system would affect statistics on the number of claims made, the official unemployment rate is calculated differently, using the Current Population Survey, and doesn't incorporate data on claims:
You can't extract any more information from the image than what the satellite originally acquired - you could make an "error correcting" model that creates an image that appears to have higher fidelity, but the additional detail is just a representation of what the model is filling in based on the other images it has been trained on, rather than an actual increase in resolution.
Possibly still a reasonable buy for him even at 3-6x what was expected if the higher price generates enough extra press coverage.
I didn’t quite get that paragraph, because flexibility of schedule is an important criteria for determining if someone is an employee or a contractor in most of the the US! (In California this isn’t true as of fairly recently, but most of the rest of the country still uses a 20-point set of factors that includes schedule flexibility)
https://www.oregon.gov/ODA/shared/Documents/Publications/Nat...
The real risk there is that it would incentivize landlords to select the highest-income tenants possible (more so than they currently do).
It's worth noting that printing money and giving it to the banks is effectively the same as taxing cash - it causes inflation and reduces the value of Yuan-denominated assets (though there are ways to manage this).
By doing this, the Chinese government is subsidizing entities that borrow from the bank (which are often SOEs) without directly taxing its citizens.
I'm really interested in seeing how this will play out if solar continues to get cheaper.
I'd imagine that at a certain point, the panels themselves will be cheap(er) compared to the installation cost (so that the cost structure changes to favor the most efficient installation/maintenance regime), and that residential-scale solar will be uneconomical compared to large-scale installations in the desert, where you can get a couple of hundred megawatts all in one place getting the advantages of optimal weather, easily-installable sun-tracking, and more regular maintenance than what you'd get on a 1,500 sqf roof in SF.
It's entirely possible that, 10-15 years from now, this ends up being a "tax" paid by all homeowners to the residential-solar-installation industry so that they can put solar panels on roofs that will never earn their money back, just so that the homeowner can displace what might be equally clean energy generated elsewhere.
The MTA is comparing single-ride fares with "effective" cost per ride of a then-$81 30 day pass, which they amortize over around 70 rides, which is a bit high for the average person in my experience (though not ridiculous).
It might be more accurate to compare the single-ride fare then with the single-ride fare at the time the photo was taken, which was either $2.00 or $2.25 (the price changed in 2009, I think).
The most frustrating thing about this is that these aren't even official government cards (they're issued by the union instead), so it's harder to crack down on them through your elected officials.
Even if you banned cards (and I'm not sure you could), unions might just shift to making "Police Benevolent Association Family Member" license plate frames - or any other token. I think the most efficient way to solve this might be to have the law fall more lightly on everyone - reducing the benefits to connected citizens relative to the their less fortunate neighbors.
I'm struck by how significantly the effect varies between states - according to table 2, 25% of the transfers (around $1.55 billion) are accounted for by Connecticut and Delaware alone (which collectively are around 2% of the US population).
Does anyone know why the effect is so strong in these states? In the notes the authors say that Delaware is one of the states with unusually strong protections - so I'd expect transfers would actually be unusually low in that case!
Same here, though that is conditional on a friendly environment for financing to reach the longer term.
And it also assumes that skill in the (relatively) low volume production of rockets can be transferred to mass production of cars.
The source is a bit misleading - the 12.8 trillion includes money the fed spent to buy assets, driving interest rates down. They've been slowly unwinding that position, and it's not as if they were giving the money away.
I think the rapid score moves are psychologically useful. For every person who slides rapidly downwards on a few unlucky hits, there's another who gets a few good shots, shoots up the board, and thinks "hey, I'm actually not bad at this!" which might make them more likely to stick around.
This is neat but sort of strange - I love the idea of a hardware package that makes fiddling with deep learning easy, but I wonder why Amazon went through the trouble of producing such a relatively niche product?
If I had to guess, this is going to be a great "user education" tool for AWS, designed to get new developers on the platform as early in the learning process as possible.
I wonder what the compliance rate was? I've found a cold-shower regimen hard to stick to - effect might actually be larger if there was a way to verify and sort out anyone who was sneaking in a bit of warmth.
I'd actually be inclined to have an escalating fee (as a % of sale price) for a higher final price (relative to an independent appraisal) if I were a seller - just to keep the incentives aligned.