And Google's revenue per employee is $2,020,329 in 2022 (says Zippia), while the average salary at Google is $123,944 (says PayScale.com) which is about 6% (and for software engineers "The median compensation package totals $275K", says levels.fyi).

So... is Varoufakis lying?

But Google also has to pay for data centers, electricity, office buildings, taxes and so on. And we could say the same for other tech companies. So a more reasonable thing to measure would be the profit per employee, which is said to be $158,000 for Alphabet (Google) as of 2019 and higher or lower for others.

And this "techno-feudalism" concept seems to uniquely describe Amazon and perhaps a few other "walled-garden" cash cows like Apple's app store. Google or Microsoft just seem like normal capitalist entities to me. Too powerful, perhaps, but not "feudalist".

On one hand, you have statistics saying a million people died of Covid in the U.S. alone. On the other hand, some people have feelings. I'd like to share the story of my uncle Bert, who died "with" Covid in Alberta, my aunt Elaine, and my father Don who lives 5000 km away in Hawaii. All three of them became anti-vaxxers after the pandemic started because their right-wing sources told them about the evils of vaccines in general and (once mRNA vaccines got the EUA) mRNA vaccines in particular.

A key part of this belief system is that Covid isn't so bad (as long as you have ivermectin anyway). My aunt wrote this on Facebook:

Bert is in a ventilator in ICU in Lethbridge. He is in a deep sleep, seemingly unaware of his surroundings or anyone's touch. A team of 4 turns him every afternoon from his back to his tummy which seems to increase his oxygen level, and then in the morning they move him back onto his back. - Elaine

Last night, Bert's brother Don phoned me to say Bert is being treated for the wrong condition. He feels that Bert has suffered a stroke. He was in the front yard, watering the flowers when he fell, and was unable to get up. This is similar to other incidents that have occurred recently, and Bert has called me on his cell, so I rushed out to help him stand up. I don't know why we didnt do more than help him into the house so he could sit in his "lazy boy" for a while. This time a young couple driving by saw him fall and rescued him before I could get to him - hence his trip to the hospital and a diagnosis of covid 19.

So: Bert has fallen repeatedly in the garden. This time when he fell, passers-by called an ambulance. When he got to the hospital, he was tested for Covid and it came back positive. Perhaps due to this, Elaine wasn't allowed to see him (she was also infected, but had a very mild case). Later, he was placed on a ventilator (a common treatment for severe Covid). Don, from 5000 km away, diagnoses him with a stroke. The hospital told Elaine he died of Covid, but Don still thinks he died of a stroke. Elaine is inclined to believe Don, though she told me later, her voice breaking, that maybe if they had allowed her to give him ivermectin, he'd still be alive today.

And you know what, I do wish staff had given him ivermectin. He would've still died, but at least Elaine wouldn't be left with an impression that doctors are the enemy. As for my father--the man I remember is gone, replaced a mystery man who sent me messages reciting the talking points given to him by the TV, never with evidence to back it up. He ignored all of my replies. Now we don't talk anymore.

heat pump efficiency is still significantly higher than fossil fuel and electric resistive heating systems at an appliance level

This is a trivial statement. Heat pumps are by nature always more efficient "at the appliance level" if they move any heat whatsoever. You burn a fossil fuel, you extract maybe 75% of the heat (some is lost in exhaust). A heat pump retains almost all the heat produced by the electricity it uses, plus any additional heat it moves from outside the building. The report goes on...

Heat pump efficiency is measured by the deviceâ(TM)s coefficient of performance (COP), the ratio of the useful heat outputted to energy consumed. Typical COP values for heat pumps lie in the range of 3â"6, indicating that 3 to 6 units of heat are created from each unit of electricity used.

Correct, except air-source heat pumps drop below 3 if it's cold enough. Then they mention that in very cold climates ground-source heat pumps are better than air-source to avoid this problem, and they show a chart indicating that previously-installed heat pumps usually have a COP above 2 "in mild cold climates".

But again, the report is talking "at an appliance level" which neither the Guardian nor Slashdot mentioned. To claim "heat pumps are more efficient overall" you must be looking at the primary energy supply, which this report ignores. For example, is it more efficient to burn natural gas in a power plant to generate electricity at 44% efficiency and then transmit that to your home (with minor line losses) and use that electricity to power a heat pump, or is it more efficient to send the natural gas to a home and burn it there? The actual answer is "it depends on the heat pump and the temperature of the outdoor heat source".

A heat pump will probably be more efficient overall if you get a high-efficiency model, or if it's not too cold, or if you have a ground-source heat pump.... but it probably won't be "twice as efficient". And even if it is more efficient, that doesn't mean you won't pay more for the energy. I once computed that electricity in my location cost four times more than natural gas for the same amount of energy. Electricity actually should cost more than twice as much to account for generation losses and line losses, but on top of that you have to help pay for the power plants themselves, and their profits.

People should switch to heat pumps in the long run because burning fuel causes global warming, and higher energy demand stimulates building more clean power plants (wind and nuclear - solar in winter is a weak energy source, and yes I am aware solar panels are more efficient in cold weather, but that boost doesn't come anywhere close to making up for the lack of sunlight).

You may or may not save money in the long run with a heat pump. Generally, the colder your climate is, the less likely you are to save money, although if your location lacks natural gas lines then heat pumps probably make sense. Here's a nice video about the nuts and bolts of home air-source heat pump technology.

Historically it has been very difficult to convince the NRC to approve reactors. In 2012 it was reported that the NRC approved the first commercial reactor to be built in over 30 years. What they didn't mention was that the NRC had never in its entire history reviewed and approved a reactor that was built. Reactors that actually got built were reviewed by the AEC before the NRC existed; the NRC had a colder attitude to nuclear power. And while they did approve the debacle known as Vogtle 3/4, they recently made it even harder (some would say impossible) to get a reactor approved.

This is the context in which the DoD wants to get a mobile reactor approved? A reactor that could be blown open up by a missile? A reactor that could possibly topple over on its side?

I know of one reactor category that makes sense for a mobile reactor: a thermal-spectrum, burner-type (uranium-fueled) graphite-moderated molten salt reactor (MSR). MSRs are self-stabilizing: physics ensures fission stops if it overheats, and if uranium is dissolved in salt there is no need to vary reactivity with control rods or add excess reactivity (fuel) in the reactor chamber to deal with xenon poisoning. They're also compact by avoiding the need for

• - a large body of water for cooling
• - a thick pressure vessel to hold in steam
• - a much thicker concrete containment structure to hold in steam if a pipe breaks

...while also being safer to cool with water, relative to liquid sodium-metal designs. But possibly there are other designs that could work (e.g. liquid lead coolant?)

Uranium atoms split into a wide variety of radioactive "daughter atoms", and molten salt is able to dissolve the vast majority of those atoms so that they cannot escape into the atmosphere. For a military design it might also make sense to use fuel balls (TRISO?) in non-radioactive coolant salt, so as to reduce environmental damage to the ground if the reactor is blown up, though xenon presents a bigger challenge in that case.

Certainly it's doable, and would have huge benefits on the battlefield in a war with a major power (*cough* China *cough*). But unless the NRC has different standards for military and civilian applications, it's hard to imagine them actually approving it.

While China can tell everyone to f**k themselves, hypocrisy has a lot of propaganda value both inside and outside the Free World. And yes, there is plenty of hypocrisy in U.S. history, but the U.S. still looks better relative to China (especially within the free world) if the U.S. stops doing obviously hypocritical sh*t.

Consider: Russian propaganda says "so what if we invaded Ukraine - the U.S. invaded Iraq!" But that invasion happened 20 years ago and there is no huge chorus of Americans eager to bomb Iraq nowadays. But if, last year, the U.S. had declared that Iraq was its 51st "state", Russia's whataboutism would be much more persuasive.

Also what he's accused of doesn't really fall under "free speech".

What he's accused of, as far as I know, is acting as a middleman: someone gave him classified information and he gave journalists access to it. I am unsure how this is different from Ed Snowden sending documents directly to journalists. Anyone care to explain?

But what really gets me is that Assange is an Australian living in the U.K. being extradited to the U.S. for breaking U.S. law. How can he be subject to U.S. jurisdiction? If what he did was illegal in the U.K., why can't he be tried in the U.K.? If what he did was legal in the U.K., why is it legal to extradite him?

Pretty much no one has a perfect memory like that. But the opposite is true for the computer.

No, GPTs/LLMs can't replicate the book word for word. It can rarely even remember a complete sentence from a random book. LLMs are neural networks - digital brains - and they have similar properties to biological brains. Training an LLM on a book leaves the LLM with an impression of what was in the book and the style of the book, just as a human would have. If an LLM is trained on the same book many times, it will replicate the style better, but it still cannot regurgitate the whole book or even a single page.

Consider a quote that is famous, but not so famous ChatGPT can remember it all:

Me: What was the paragraph Jefferson said? "But our rulers can have authority over such natural rights only as we have submitted to them..." then there was something about "testimony in a court of justice"...

ChatGPT: The passage you're referring to is from Thomas Jefferson's "Summary View of the Rights of British America," which he drafted in 1774. The full paragraph goes as follows:

"But our rulers can have authority over such natural rights only as we have submitted to them. The rights of conscience we never submitted; we could not submit. We are answerable for them to our God. The legitimate powers of government extend to such acts only as are injurious to others. But it does me no injury for my neighbor to say there are twenty gods, or no God. It neither picks my pocket nor breaks my leg. If it be said, his testimony in a court of justice cannot be relied on, reject it then as you would the evidence of a lunatic or a child."

The actual quote:

But our rulers can have authority over such natural rights only as we have submitted to them. The rights of conscience we never submitted, we could not submit. We are answerable for them to our God. The legitimate powers of government extend to such acts only as are injurious to others. But it does me no injury for my neighbour to say there are twenty gods, or no god. It neither picks my pocket nor breaks my leg. If it be said, his testimony in a court of justice cannot be relied on, reject it then, and be the stigma on him. Constraint may make him worse by making him a hypocrite, but it will never make him a truer man. It may fix him obstinately in his errors, but will not cure them. Reason and free enquiry are the only effectual agents against error.

Your client is hereby instructed to cease the collection of all facial recognition data and iris scans

If that's the main ask of the letter, Worldcoin probably just sent a letter back explaining that they do not "collect" iris scans or data that can be used for face recognition. What they store is cryptographic hashes of phase information from Gabor-wavelet transforms of iris scans. Once the phase information is gathered, most of it is thrown away, leaving a relatively low-dimensional binary image that is almost invariant when noise is added or lighting conditions are changed. Next, they use a special biometric hashing technique (possibly a Fuzzy Extractor?) to get a hash from which the original iris scan cannot be reconstructed, and this hash is what is registered in the Worldcoin database. Reconstructing the iris scan from the hashcode is not possible. I understand they have a further security measure so that if you have the hashcode of someone's iris, you cannot look up which person it is, nor can you look up their World ID; the only thing you can do is check whether they have a World ID at all (though I didn't follow the technical details of this part).

Technical details are given here and here, and you can find an excellent overview of the field of Biometric Proof of Personhood (by the inventor of ethereum) here.

The other day I was watching a video by a relatively obscure vlog on the Ukraine war. A reader asked, "Did you observe the number of trolls has diminished since Prigozhin troll farm has been closed?" (referring to Wagner's mutiny against President Putin, which temporarily disrupted Wagner's pro-Kremlin information operations). He responded "Yeah, it was funny because the week that happened...all my troll comments disappeared...they eventually started to come back". I myself have been talking to people who go into various forums and take Putin's side, and noticed that usually they give very low-effort responses, won't tell me where they get their information, won't indicate whether they are left-wing or right-wing, and I have to wonder: does this person really believe what they are saying, or were they hired to spread this point of view? But right now, these people are actually people. Right now, a country can buy a lot of human-written comments in support of The Dear Leader and Against His Enemies, but in the future this problem will get much worse thanks to the rise of AI bot armies that are very inexpensive and almost impossible to tell apart from humans. So I think biometric proof of personhood will be very important in the future, and we should respect people who are trying to make these systems work in a way that respects privacy and does not store biometric information in a central database.

On a technical level it is very difficult to make a system that (i) only gives IDs to real humans, (ii) won't give multiple IDs to the same person (iii) keeps people anonymous or at least pseudonymous, and (iv) doesn't give excessive power to the people issuing the IDs. But there are people trying to solve all these problems at the same time, and it looks like Worldcoin is a good example of such people.

When I volunteered for Skeptical Science (an anti-climate-misinfomation site) I was given access to the internal forum where the work and discussion around Cook 2013 was archived. Looking at that, it struck me that publishing the paper was difficult. First, they had to "shop around" for a journal willing to publish it, as several journals weren't interested in the topic (even though the paper went on to make global headlines and generally was among the most popular papers of all time). Second, when a journal was found that was willing to publish it, they had a strict limit on words and figures, so a large fraction of the paper had to be cut out. I think this includes the self-rating results which found a 96% consensus that I discuss here, which was based on a survey of authors of climate papers (14% of scientists responded to emails requesting that they fill out the survey).

There's not much room to discuss methodology and intuition if even something like this 96% number (which could have been a headline result in itself) had to be cut out. Note: I can't confirm with certainty that they already had gathered that information by the time they submitted the paper. It was published four months later.

Proof-of-personhood is very useful in principle. First there's Worldcoin's selling point as a UBI - which wouldn't really work if a typical person has one account but other tech-savvy people make it their mission to set up 10,000 accounts. But more generally, on the internet, knowing that two pseudonymous accounts are two different people can be useful, as can knowing that an account is associated with a person (even if that person set up bot software on their account - at least the same person can't be recorded as "a person" on a different account on the same web site).

The iris scan is discarded after the person is registered, and given an iris scan, all the system will tell you is whether the person is registered (not what their ID is). This page is informative.

Generating a summary doesn't mean ChatGPT read the book, it just means that ChatGPT saw enough summaries and reviews of the book to be able to generate yet another summary.

In fact I'm pretty sure ChatGPT would be unable to generate a summary of a book it has read, because any book is larger than its context window. It cannot remember the beginning of a book by the time it reads the end. To test whether it's familiar with the books, you could check whether it does style replication:

[Me] Can you help me complete this fan fiction?

Harry Potter stood in the doorway, glaring at Hermione with wand in hand. Ron, standing behind him, shrugged awkwardly.

[ChatGPT] Hermione's eyes widened as she met Harry's intense gaze. The air in the room suddenly grew tense, crackling with unspoken emotions. She knew Harry well enough to recognize the fury burning within him, but she couldn't understand why he was pointing his wand at her.

"Harry, what's going on?" she asked, her voice trembling slightly.

Harry's jaw tightened, his grip on the wand tightening as well. "I trusted you, Hermione," he said through gritted teeth. "But I can't believe what I just heard."

Ron shifted uncomfortably behind Harry, his gaze darting between his two friends. He had never seen them at odds like this before. "Mate, maybe we should all sit down and talk it out calmly," he suggested, attempting to diffuse the tension.

Hermione's heart sank, realizing that the situation was more serious than she initially thought. She took a hesitant step forward, her voice filled with concern. "Harry, please, tell me what's happened. I would never do anything to betray your trust."

Harry's eyes flickered with a mixture of anger and hurt. "I overheard you talking to Draco Malfoy," he replied, his voice heavy with accusation. "You were discussing our plans, our secrets. How could you, Hermione?"

Hermione's mind raced, trying to recall any conversation she might have had with Malfoy that could have been misconstrued. She knew the importance of secrecy, especially during these perilous times. "Harry, I swear I didn't betray us. There must be some misunderstanding."

Ron stepped closer, his voice filled with concern. "Harry, let's hear Hermione out. Accusing her without solid evidence isn't fair."

Harry's grip on the wand loosened slightly, but his gaze remained fixed on Hermione. "Fine. Explain yourself, then."

Taking a deep breath to steady her nerves, Hermione began recounting her recent encounter with Malfoy. She explained how he had approached her unexpectedly, seeking a chance at redemption. Malfoy claimed to have vital information about the Dark Wizard they were fighting against. Hermione, hesitant at first, had cautiously listened to his story, understanding the significance of any potential advantage.

"I didn't trust him completely, Harry," Hermione explained earnestly. "But I thought it was worth investigating. I haven't revealed any of our secrets or plans to him. I promise."

Harry's anger seemed to waver, his expression softening as he absorbed Hermione's explanation. He lowered his wand, [blah blah blah]

Even then, if it's a popular book ChatGPT could do this just based on fan fiction and reading random excerpts it found in blogs.

An official did drink the water. Obviously, if you drink the water you are incorporating tritium into your body which maximizes the risk. You'd probably safer swimming in it, and much safer still by not swimming in it and mixing it with a large body of water. And if you're truly obsessed with safely, mix it with the entire ocean.

However, virtually no one in the media is reporting on just how much (= little) radiation this water contains, even though radiation is very easy to measure and understand: anybody can buy a radiation meter for like $50 and there must be tens of thousands of meters in Fukushima, and here's a summary of a NASA report on the hazards of radiation.

The dose of natural background radiation we all receive each year is about 2 mSv, give or take 1 mSv. So if, for example, radiation doses would be elevated by 1 mSv for a few years, that's like moving from a lower-radiation town to a higher-radiation town for a few years. According to NASA data, for the average adult, 100 mSv of gamma radiation gives a cancer risk of roughly 0.4%.

So how radioactive is this water? The closest answer Google could give me was this article:

James Conca, an American environmental scientist specializing in the disposal of nuclear waste, called the radiation levels of the treated wastewater "trivial."

"The radiation dose from one quart of this water is equal to four bananas or a family-sized bag of potato chips. A ton of this water gives a dose of approximately 4 mSv, about the average annual background dose to everyone in America, and less than a single chest CT scan which is 7 mSv," he wrote in Forbes.

Well no, annual background radiation is closer to 2 mSv than 4. But he's right, this water is about as safe as the tap water in my old apartment in the Philippines (we bought drinking water in jugs, of course), and it will be mixed with the entire ocean.

But even Vice's article makes the confusing move of using Bq measurements:

tritium in the Fukushima plant's wastewater will be diluted to 60,000 Bq/L before being released.

1 Bq (bequerel) is defined as one nucleus decaying per second. This number generally says little about the danger level, first because different nuclear decays happen at a variety of energy levels, and second because it can't hurt you unless your body actually absorbs the radiation, the probability of which depends on circumstances (but is zero if you're on your couch eating a fish, while the bequerels are in the ocean).

But bequerel numbers are often used to scare people because they look very large. In the show "Chernobyl" someone says "Every Atom of Uranium Is Like a Bullet". Well, yes, ionizing radiation is sort of like bullets, but it's highly misleading to say this without mentioning that your body typically absorbs about 20,000 "bullets" per second of natural background radiation. Or in other words, 20,000 Bq of radiation is perfect normalcy. This wastewater is 60,000 Bq/L, so I guess drinking a litre of it would temporarily elevate your radiation levels by (very roughly) three times above normal, until the water passes out in the usual way. This is a "so what" kind of level, as long as you drink in moderation and have a designated driver. Swimming in it is generally safer because the radiation stops the moment you leave the pool.

Some scientists say the impact of long-term, low-dose exposure to radionuclides is unknown

Please talk to a radiation scientist next time, AP. Frak the media is awful at this.

Um... so back on the topic of solar, measure the difference between maximum daily solar output and minimum daily solar output, and multiply the result by the cost of that much energy storage, and you'll have some idea of the scale of the challenge posed by variable renewables.

According to one source (outdated, but let me know if you have more recent data), maximum solar power is about 50 times higher than minimum solar power, and the minimums are grouped together with other periods of low output that are known collectively as "winter". Eyeballing the chart labeled "Average daily output of German solar in 2014", solar output during winter as a whole is about 8 times lower than in summertime as a whole. (and yes, I know solar panels are more efficient in cold weather, but to reap this benefit you actually need sunlight shining on the panels, rather than long winter nights, overcast days, and panels covered in snow, not to mention that the sun is much less bright as it nears the horizon.)

The same chart shows that electrical load is slightly higher in winter than in summer, but again, this is based on old data, from a time when people relied mostly on Russian gas for heating. To reach net zero, we must replace natural gas with electrical heat pumps (preferably ground-sourced heat pumps) which will greatly increase winter electricity demand. In summary: Germany needs the most energy during the season when solar energy is weakest. The same is true in Canada, northern USA and most of Europe.

How do we move the summertime solar energy to wintertime? Shall we try batteries? No, we won't, and a simple calculation shows why. Let's say, for the sake of argument, that on one day we get 90,000 average megawatts from solar and electricity demand on that day is 45,000 megawatts, leaving 45,000 unused, which is 1080 million kWh of energy. Suppose we want to store that energy in batteries and discharge it in wintertime. What would that cost? I understand a typical price of grid-scale battery storage is about $200/kWh, so batteries for 1080 million kWh costs $216 billion (conveniently, 1 USD is worth about 1 EUR right now). Now, what is the energy stored in the batteries worth when sold in wintertime? I'm seeing spot prices last winter around 100 euro per MWh, so about 10 cents per kWh. So 1080 million kWh is worth about $108 million dollars.

So even assuming you get the excess solar energy for free, you're paying $216 billion to earn $108 million per year. It would take ... let's see ... infinity years to break even at a discount rate of 1%.

Battery storage is economically viable if (1) you can charge and discharge at least once per day (2) electricity prices swing wildly. Storing energy from summer to winter is nowhere near economically viable with batteries, and even at a price 50 times lower it still wouldn't be viable. So, given that demand is highest in the winter, why is Slashdot celebrating high solar output near the summer solstice? Does anyone offer a way to use all the excess summer capacity Germany plans to build?

And by the way, that second point about electricity prices swinging wildly--that's potentially a good thing if it means electricity prices are sometimes very cheap. But it's a good thing for consumers, not solar producers. People always point out that solar power is cheap now. That's great! Just remember that a solar farm is selling its power at the same time as all the other solar farms which lowers the price. Remember all the Slashdot stories about electricity prices going negative? That's a nightmare scenario for solar farms. If there are only a few solar farms this can't happen; if there are too many, solar farms can hardly earn any money. So yeah, solar power is cheap, but in a solar-heavy grid, 1kWh of solar is worth less than 1kWh of anything else. And this is why, even if nuclear is expensive, it can be a better deal than solar if there is too much solar. Also, nuclear being expensive is a choice society made; we used to build nuclear plants because they were affordable (see Devanney Figure 7.11). But I digress--this /. post is about solar, after all.