One answer is to stack 3 LEDs on the die, each tuned to the center wavelength of human photo-receptors. Existing RGB LEDs emulating "white" suffer from shadowing effects where shadows have a color based on the parallax between the red, green, and blue elements in the light bulb. This is no longer an issue if you co-locate the emitters.
No change. Just because your radio is "software defined," doesn't change physics and signal processing math. There are 'best' ways to use a given piece of spectrum, and while a SDR can make your hardware more flexible, it can't change that the most efficient way to use spectrum is usually to have a dense set of subcarriers modulating amplitude and phase.
By avoiding a strict multiple-access protocol on a band, everything just gets worse; probabilistic back off is inefficient vs hard-scheduling.
Every wireless system I've worked on (satellite radio, WiFi, LTE) has been very spectrum constrained, so I dont really believe in this future where spectrum is cheap.
The piece that always bothered my about Airbnb is how it incentivizes hosts to cheat in small (but not insignificant) ways. If my only recourse for not-as-described when I've arrived is to accept the keys, or abandon the booking, that's not a real choice. Most people in an unfamiliar place will just put up with the exaggerated "fully stocked" kitchen, false pictures, etc, and are left with no ability to claw anything back from the host. I think much of this would go away if guests could get partial refunds if they can prove the cheating.
Definitely the peak would be higher. My laptop has a ~60Wh battery and gets ~6hours of battery life, so I was just assuming 10W average for web browsing.
I'd probably use closer to 10W as the device power level (a standard laptop, what I'm using right now), so the number for me would get close to 0.2g, but your point stands; the article is answering a different question than I was assuming, but regardless it seems that the majority of the energy is clearly the local device.
I'm having a very hard time rationalizing these numbers. Something has to have some very poor assumptions:
Let's assume each page load is ~1mbit, and it's being served from 1000km away. 1mbit may be low, but most of it will presumably be content delivered by a CDN which is far closer than 1000km.
The reference design (which I work on) for 400G DWDM systems are ~100W for 2 transceivers + EDFA + switches, for a standard 80km link. Assuming we hop every 80km (which should be conservative here), we get 3mJ for the energy through the network. The previous generation is probably 10x worse, so ~30mJ.
Let's add the endpoint too; Wifi router + cable modem (12W or rather 3W marginal power since it's running anyway/ 200mbit) = 15mJ
And finally for the server. I'd hope the server could do 1k requests per second, and that should amortize against backends, etc (2k RPS, but frontend + DB backend server, maybe), maybe 150W for the server, so 150mJ.
This gives us 200mJ for a page load. Using California's current (1:28am) emissions for power generation of 0.274 mT/MWh (it's 1/3rd of this during the day), this yields 15ug/request.
Where do the remaining 5 orders of magnitude come from? What am I missing? If TCO/production energy of the server was included, that couldn't more than double it (capex in a datacenter < opex, and if 100% of the server cost was for energy, we could double, but this is wildly conservative)
If the border tax is just for countries without a similar tax, let's consider one example, say steel. Foundries in countries that fall under the taxed regime then have an incentive to produce the cheapest (and dirtiest) steel, since they get no benefit vs the other foundries (you could rightly retort though that this is largely the way it is today). Maybe the tax then should try to evaluate the relative carbon intensity of each supplier, rather than just by region/country, but good luck avoiding cheaters around the world.
I guess overall I have a hard time imagining that you could create a fair tax/duty for carbon intensive industries that accounted for all of these potentially perverse incentives unique to each industry (just take a look at the US tax code), given how hard it is to write and pass policy without succumbing to special interests.
How does this deal with the border-adjustment problem? I used to be a strong supporter of carbon taxation, but ultimately concluded that a good policy (that doesn't have perverse incentive consequences for imports/exports) would be an extremely difficult policy to implement, and beyond the current scope of support for climate policies.
A major point missed by the other answers is that these rovers are exceptionally specialized for their mission, so building a second rover wouldn't be able to accomplish many additional scientific goals, without major modifications. The Curiosity rover and Perseverance share a lot of components, but pretty much the entire science suite is replaced or upgraded. As such, the expensive parts weren't shared, even though the huge value of the bus/bogie system and the skycrane were shared. Even the EDL systems have been changed a lot; the entry system on Perseverance does localization during some phases with a camera (a first for a spacecraft afaic, maybe the shuttle counts though since a human eyeball steered it to a runway). If you don't have additional science objectives that a second rover could accomplish, it will be very hard to justify that budget, even if it is only 20% more (for example).
This being said, clever mission design might take advantage of this cost structure where multiple units are cheaper to build, intending to build two cheaper spacecraft instead of one expensive one; eggs in two baskets. In this case, we've seen this happen with spirit and opportunity, which was a wildly successful program.
Oh boy, at least it's only annoying. Not sure if this is just because I use 3rd-party bluetooth headphones (bose QC35) and inadequate Apple software testing, but if I try to use headphones for audio and switch to the internal macbook mic, Facetime either freezes or crashes every time.
Sure, many of the operations in the list can be pipelined as you mention. Something like G-Sync would also allow you to sync the destination display to the arrival of the (start of) frame.
Why exactly do existing video streaming solutions use such small amounts of bandwidth and have terrible quality as a result? Does anyone have a deep dive into why this is the case? It seems that it would be a killer feature to make better utilization of bandwidth.
Even over wifi, speedtest shows 4ms/100mb/100mb on my internet connection, but Zoom, FaceTime, and others never use more than about 0.8Mbit/s for a video stream, and the resulting quality of audio and video is...understandably poor.
Latency too totally feels like a software problem, perhaps with too many layers of abstraction. (60fps->16ms for the camera, ~10ms for encoding with NVENC/equivalents, 35ms measured one-way latency from my laptop to my parents 4000km away, ~10ms decode, 16ms frame delay = 87ms one way). Maybe I'm asking for too much from non-realtime systems (I'm used to RTOS, extensive use of DMA, zero-copy network drivers, etc), but it seems that there is a lot of room to improve.
I would also add computational complexity. We're increasingly willing to throw more transistors at video decoding and encoding due to their reduced cost and increased performance, so the algorithms can become more complicated, have larger perform larger searches, etc.
If you think about it from a power perspective, the power cost of a bit of data transmitted to a mobile device is pretty large. I would expect to see video compression formats increase in complexity until the power cost of compressing away one more bit approaches the cost of transmitting one more bit over the network.
Soyuz still uses UDMH and nitrogen tetroxide as its propellants. It simply jettisons the propulsion module before reentry so the reentry module has no thruster system and therefore no (toxic) fuel.
With the return of a SpaceX vehicle, the recovery crew don't have suits on, that would be very unwieldy at sea, and they would just back off if trace gas detectors measure a toxic leak.
RCS engines are practically always hypergolic since you need to be able to pulse on and off very easily/quickly, and propellents need to be storable (non-cryo). If you look at the choices remaining, they're all various levels of extremely toxic.
Depending on where you are, that's pretty close to what's happened. Right now (4pm), the California ISO is generating 73% from renewables+hydro, 7% nuclear, and the balance gas. Naturally, the night replaces solar with gas.
This is largely how California has reduced electricity emissions to their current levels.
Power markets are more complicated than most people realize. One thing to note is that solar power can be cheaper than gas, but still not be economic. The fact of the matter is that an intermittent kWh is not as valuable as an on-demand reliable kWh to a utility who's number 1 priority is reliability. Even as solar is acquired at lower and lower prices, if evening power is generated by expensive and inefficient gas turbines, the customer might not see costs go down (and emissions might not go down either!). Solar is clearly economic in many markets, but we'll never get grid emissions in a place like California down much more without storage.
This is a relatively good point. In the California electricity market, the increase in renewables results in a very large power-ramp rate in the late afternoon (nothing unique to California). In general, more-efficient power plants either cannot, or lose efficiency, when changing operating power levels. A ~60% efficient combined-cycle gas plant cannot ramp power very much, which results in the grid building and running more ~40% efficient gas turbines, which can ramp their power output in order to meet the early-evening power ramp.
Interestingly, at this point, adding more Solar to the California grid results in very little emissions reduction, since the additional solar displaces efficient baseload generation with inefficient ramp-able load. The solution, of course, is storage.
I can't speak about wind (not my expertise), but for solar, the energy payback is about 1-year in operation for current installations, with a predicted 30-year lifespan. Compare this against estimates for the energy cost of bringing gasoline to market which can exceed 30%.
Power engineer here. Something that is not apparent is that many small power supplies are limited in power handling not by the components, but by maximum safe touch temperatures for meeting regulations (probably around the 10W level). Therefore making a more efficient power supply means that you can package it smaller since the heat needs lower surface area to dissipate safely. Right now, gallium nitride MOSFETs are gaining popularity allowing for higher efficiencies in power conversion, and some wildly high power densities for USB chargers. I'm a fan of ravpower's 30W USBc supply which is about one cubic inch with folding prongs.
Another side to consider is that often said Senator has 40-years of experience being a senator (or similar), and ~0 years being and employee or business owner. I often feel (anecdotally) that many representatives have spent their entire lives in government such that they have never experienced what it's like to be in the private sector, and therefore can be tone deaf to issues that matter to me.
In addition to energy, power use is another reason; parking a core will allow it to cool down thermally, so that when it is put back in use (milli)seconds later, it can run at a higher clock speed for longer.
EDFA's: erbium doped fiber amplifier. They're optical signal amplifiers. Due to the attenuation of the fiber, they need to be installed every so often on the fiber, under the sea. Since the equipment is cheap in comparison to the cost of repair, modern undersea amplifier boxes often contain multiple amplifiers and switchgear to MUX to a new one after it fails.
While rooftop solar does allow you to go completely off-grid, if you still want to have a grid around, it will have to be centralized. An underappreciated aspect of power grids is just how much advanced control is required to keep the entire thing stable and functioning, think armies-of-PhD's-difficult.
When grids get too small, or don't have sufficient control (either knobs, or software/people turning the knobs), the results are frequent blackouts.
Interestingly, the risk begins increasing again as your grid gets too large: you have too many knobs to reasonably control centrally, and it certainly won't work without synchronized control. As a case study, see how China is breaking up some of their huge electricity grids (in to still-large ~1000mile or so regions), as maintaining reliability and preventing faults from propagating across the country became too difficult.
I use very good PCB inspection xrays and ct scanners, and it is critical to use energy levels which do not resolve the silicon, e.g. are not absorbed by the die. Otherwise, it is frustratingly easy to give the part a total ionizing dose which breaks it.
I wouldn't want to use an xray to do more than verify the die is the same shape and has the same wirebond layout.