• 1 Post
  • 241 Comments
Joined 1 year ago
cake
Cake day: July 10th, 2023

help-circle

  • It’s complicated. The main issue is, I live on a remote farm without cell coverage, except in the tiny zone under my 50’ tower with booster.

    However I now have Starlink, and wired and wireless APs covering a large area with high speed, low latency data.

    So, port my number to VoIP.ms, which supports SMS, and make all my calls/texts through Wifi using SIP. On the road, use a basic cell plan with unlimited slow data that is still fast enough for voice. Tested, working, so far fairly simple.

    Now the issues. RCS won’t work with my now VoIP provisioned number, because there’s no SIM for it. The SIM in the phone has a different number, that of the new plan which will be unreachable at the farm by voice/SMS just like the old number used to be.

    This would all be a non-issue if my provider supported VoWifi on anything other than iPhones, but sadly this is not an option. So I’ve got service everywhere now, but am stuck with voice and SMS, no RCS or MMS.




  • Right, we need to come up with better terms for talking about “AI”. Personally at the moment I’m considering any transformer-type ML system to be part of the category, as you stated none of them are any more “intelligent” than any others. They’re all just a big stack of tensor operations. So if one is AI, they all are.

    Remember long ago when “fuzzy logic” was all the hype and considered to be AI? Just a very early form of classifier network but everyone was super excited at the time.


  • I’m just stating that “AI” is a broad field. These lightweight and useful transformer models are a direct product of other AI research.

    I know what you mean, but simply stating “Don’t use AI” isn’t really valid anymore as soon these ML models will be a common component. There are even libraries and hardware acceleration support for tensor operations on the ESP32-S3.


  • It’s possible for local AI models to be very economical on energy, if used for the right tasks.

    For example I’m running RapidOCR which uses a modern transformer architecture, and absolutely blows away traditional OCR at capturing data from character displays.

    Doesn’t even need a GPU and returns results in under a second on a modern CPU. No preprocessing needed, just feed it an image. This little multimodal transformer is just as much “AI” as bloated general purpose GPTs, but it’s cheap, fast and useful.



  • Correct, but often the actions of CEOs are performative and don’t actually support the goal of bringing money in. They like to put on a show of being ruthless, and often behave more psychopathic than an “optimal” business AI would.

    For example, it’s been proven that employee retention is one of the #1 ways to boost productivity. Costco is one of the few companies with a CEO which truly believes in this and despite paying higher wages than any other grocer they are one of the top performers in my investment portfolio.

    Remote work? Totally profitable and AI would maximize it instead of forcing workers back to the office to “put them in their place”

    4-day week? Also proven to be a net gain as workers are rested and motivated.

    A “cold and calculating” AI would be far more likely to make reforms that benefit both the company and the employees, as it isn’t motivated by power structures or the need to look ruthless. Cutting pay is a losing move as it loses talent more than it saves money, and deep learning algorithms would realize this easily.

    Also the “person who owns the AI” would actually be the shareholders, who are often ordinary investors. Rather than funneling money to bloated C-suites, the money would be more likely to circulate in the economy through dividends.


  • Look at Saskatchewan, Canada. We’re the only province with a public telecom, SaskTel.

    Most people in the cities and even larger towns have fiber, and our cell plans are significantly cheaper than anywhere else in Canada despite being a rural province with a large coverage area to population ratio.

    We also have decent electricity rates considering we have no hydro, and the cheapest natural gas in Canada. Thanks to SaskPower and SaskEnergy.

    Public utilities are the only way to do it, I’m always shocked to see people defend privatization in any way.



  • probably the best optical character recognition by far

    I’ve actually just been working with OCR this week, trying to capture data off of the screen of a stupid proprietary Schneider device as that’s the only way to get at it.

    Long story short Tesseract stinks at this task.

    The Chinese designed PaddleOCR seems significantly superior as it runs a more modern neural net and requires a lot less preprocessing. I would class it as more of a “full service AI” and not just a simple recognition system like Tesseract, it can correct for skew and do its own normalization and thresholding internally while Tesseract wants a perfect boolean raster fed to it.

    Unfortunately, the barrier to entry is a lot higher due to trying to understand their text vomit website and the fact that it seems prone to random segfaulting.









  • I trust Mullvad and Proton at this point for VPNs, nobody else.

    Any reason you can state not to use AirVPN? I switched to them from Mullvad because they support port forwarding. So far I’ve been very happy with their service.

    Having ads and sponsors blocked I can’t be 100% sure, but I don’t think they advertise at all. I only tried them because of a recommendation on Lemmy. Their site design is very old school which really says “run by nerds and not marketers” to me.