He’s very good.

  • 0 Posts
  • 48 Comments
Joined 1 year ago
cake
Cake day: June 20th, 2023

help-circle






  • Your scenario 1 is the actual danger. It’s not that AI will outsmart us and kill us. It’s that AI will trick us into trusting them with more responsibility than the AI can responsibly handle, to disastrous results.

    It could be small scale, low stakes stuff, like an AI designing a menu that humans blindly cook. Or it could be higher stakes stuff that actually does things like affect election results, crashes financial markets, causes a military to target the wrong house, etc. The danger has always been that humans will act on the information provided by a malfunctioning AI, not that AI and technology will be a closed loop with no humans involved.


  • to my knowledge, Bluetooth doesn’t work with airplane mode

    The radio regulations were amended about 10 years ago to allow both Bluetooth and Wifi frequencies to be used on airplanes in flight. And so cell phone manufacturers have shifted what airplane mode actually means, even to the point of some phones not even turning off Wi-Fi when airplane mode is turned on. And regardless of defaults, both wireless protocols can be activated and deactivated independently of airplane mode on most phones now.

    an airplane full of 100 people all on Bluetooth might create some noise issues that would hurt the performance

    I don’t think so. Bluetooth is such a low bandwidth use that it can handle many simultaneous users. It’s supposed to be a low power transmission method, in which it bursts a signal only a tiny percentage of the time, so the odds of a collision for any given signal are low, plus the protocol is designed to be robust where it handles a decent amount of interference before encountering degraded performance.




  • If construction is delayed by an injunction

    Can you name an example? Because the reactor constructions that I’ve seen get delayed have run into plain old engineering problems. The 4 proposed new reactors at Vogtle and V.C. Summer ran into cost overruns because of production issues and QA/QC issues requiring expensive redesigns mid-construction, after initial regulatory approvals and licensing were already approved. The V.C. Summer project was canceled after running up $9 billion in costs, and the Vogtle projects are about $17 billion over the original $14 billion budget, at $31 billion (and counting, as reactor 4 has been delayed once again over cooling system issues). The timeline is also about 8 years late (originally proposed to finish in 2016).

    And yes, litigation did make those projects even more expensive, but the litigation was mostly about other things (like energy buyers trying to back out of the commitment to buy power from the completed reactors when it was taking too long), because it took too long, not litigation to slow things down.

    The small modular reactor project in Idaho was just canceled too, because of the mundane issue of interest rates and buyers unwilling to commit to the high prices.

    Nuclear doesn’t make financial sense anymore. Let’s keep the plants we have for as long as we can, but we might be past the point where new plants are cost effective.









  • Video encoding has several tradeoffs:

    • Bitrate
    • Resolution/frame rate
    • Perceived quality
    • Computational complexity of encoding
    • Computational complexity of decoding

    The cell phone encoding chips for video encoding on device make sacrifices to preserve speed of encoding and preserve battery life (higher computational complexity costs more processing cycles and tends to use more power). So it’s simpler encoding, in exchange for inefficient bitrate compression.

    YouTube (and all the social media sites) have huge server farms with highly specialized encoding chips for making the videos more efficient with bitrate for quality. That makes sense because videos designed to be watched millions of times could benefit from even a very slight improvement of bitrate in exchange for a one-time cost of complex encoding. It’s also why YouTube tends not to convert to AV1 (very efficient in bitrate for quality, but computationally complex to encode) until a video has a few hundred views, because it’s not clear whether that tradeoff is worth it until they know a lot of people will be watching it.

    Netflix customizes even further for a per-video basis and looks for even more specialized tricks on a scene-by-scene basis, because every single one of its videos only needs to be encoded once for each quality/format but will be watched millions of times.

    In other words, it’s like any other engineering problem. The engineers choose different tradeoffs based on context, which means that the cell phone applies a different set of tradeoffs compared to the social media site’s server farm.