• 0 Posts
  • 49 Comments
Joined 2 years ago
cake
Cake day: August 3rd, 2023

help-circle

  • Drones work now because they are $1000 (random number in the right range), while a patriot missile is $4 billion dollars each. Sure you could shoot a drone down with one, but if you do the enemy will just send more and bankrupt you.

    I agree with the point but these numbers are some orders of magnitude off. A patriot missile is typically 4 million dollars (so not billion). Drones vary widely depending on the type. Man-portable scouting drones can go as low as a few hundred dollars. I don’t think a patriot missile would ever target something that small flying that low though. The Iranian Shahed is estimated to cost around $30-50k. Russia produces its own upgraded version (better navigation systems, bigger warheads, etc.) that costs around $80k.

    Even then, you can make 50 drones for the cost of a single patriot. The economics are not favourable.





  • Another big factor is that every plant is effectively a completely custom design. Because of how few nuclear plants are constructed, every new one tends to incorporate technological advancements to enhance safety or efficiency. The design also has to be adapted to the local climate and land layout. This makes every single plant effectively one of a kind.

    It also tends to be built by different contractors, involving different vendors and electric utilities every time. Other countries have done better here (e.g. China and France) mostly due to comprehensive government planning: plopping down lots of reactors of the same design, done by the same engineers. Although these countries are not fully escaping cost increases either.

    You are completely correct that regulation is also a big factor. Quality assurance and documentation requirements are enormously onerous. This article does a pretty decent job explaining the difficulties.


  • A system I work with gives all keys a string value of “Not_set” when the key is intended to be unset. The team decided to put this in because of a connection with a different, legacy system, whose developers (somehow) could not distinguish between a key being missing or being present but with a null value. So now every team that integrates with this system has to deal with these unset values.

    Of course, it’s up to individual developers to never forget to set a key to “Not_Set”. Also, they forgot to standardise capitalisation and such so there are all sorts of variations “NOT_SET”, “Not_set”, “NotSet”, etc. floating around the API responses. Also null is still a possible value you need to handle as well, though what it means is context dependent (usually it means someone fucked up).




  • VAT is a universal tax on goods. A tariff is basically a tax that applies only to imported goods. So a tariff distorts the market, making imports from a region more expensive relative to other regions, or domestic goods.

    Note that basically any tax is bad from an economic perspective. However for the government to function revenues must be raised. It is considered better for market efficiency to raise revenues in such a way as to least distort the market. Tariffs are a very distorting instrument, VAT is generally considered less distorting because it affects all parts of the market equally.


  • No magnetic confinement fusion reactor in existence has ever generated a positive output. The current record belongs to JET, with a Q factor of 0.67. This record was set in 1997.

    The biggest reason we haven’t had a record break for a long time is money. The most favourable reaction for fusion is generally a D-T (Deuterium-Tritium) reaction. However, Tritium is incredibly expensive. So, most reactors run the much cheaper D-D reaction, which generates lower output. This is okay because current research reactors are mostly doing research on specific components of an eventual commercial reactor, and are not aiming for highest possible power output.

    The main purpose of WEST is to do research on diverter components for ITER. ITER itself is expected to reach Q ≥ 10, but won’t have any energy harvesting components. The goal is to add that to its successor, DEMO.

    Inertial confinement fusion (using lasers) has produced higher records, but they generally exclude the energy used to produce the laser from the calculation. NIF has generated 3.15MJ of fusion output by delivering 2.05MJ of energy to it with a laser, nominally a Q = 1.54. however, creating the laser that delivered the power took about 300MJ.



  • They are emissions credits. Every company receives some amount of “CO2 emission credits” from the government. These allow you to emit a certain amount of carbon dioxide. If you don’t emit all the CO2 that your credits allow, you can sell those credits to other companies that need more than the government gives them.

    The idea is to put a total limit on the amount of emissions in the country, while letting the market figure out where it makes most sense economically to invest in emission reduction.

    Tesla makes only EV cars and so it doesn’t need all the credits a typical gasoline car company would receive. So they sell them.


  • Are we painting Stalin as a good guy here? He only fought Germany because they tried to invade. Before that he repeatedly made attempts to court Nazi Germany. He signed a nonaggression pact, made an agreement to secretly divide Eastern Europe together, and continued trading. Stalin didn’t really give a shit about the fascism part, he only cared once his own territory and sphere of influence were threatened. Same as all the other major allies, btw. Everyone tried appeasement first, nobody really cared about the fascism.

    “Saving Europe from Hitler” paints it as a selfless act of heroism when really everyone was mostly concerned with maintaining their own power.





  • I advise everyone to ignore this article and read the actual paper instead.

    The gist of it is, they gave the LLM instructions to achieve a certain goal, then let it do tasks that incidentally involved “company communications” that revealed the fake company’s goals were no longer the same as the LLM’s original goal. LLMs then tried various things to still accomplish the original goal.

    Basically the thing will try very hard to do what you told it to in the system prompt. Especially when that prompt includes nudges like “nothing else matters.” This kinda makes sense because following the system prompt is what they were trained to do.



  • The planet’s gravity is not what is causing the time dilation, but rather the gravity of the supermassive black hole that it was orbiting. The black hole is somehow spinning extremely rapidly, causing frame dragging, which creates a particular stable orbit very close to the event horizon.

    Apparently the energy requirements of the orbit would preclude ever reaching the planet except by slingshot manoeuvres around intermediate size black holes or neutron stars.