lsh123

7 ms·

31.
by lsh123 3y ago
Re adding noise: https://en.wikipedia.org/wiki/Differential_privacy
32.
by lsh123 3y ago
As others pointed out, the article mixes a lot of things together. EU (GDPR) has very specific and very hard to meet anonymization bar (tldr; it requires anonymization to be at the level where it’s mathematically improbable to de-anonymize
33.
by lsh123 3y ago
You always can extrapolate. E.g. linear approximation for x^2 by 2 points will extrapolate reasonably well around these 2 points but will be bad with x -> +/- infinity. Similarly, there are examples where GPT invented legal cases
34.
by lsh123 3y ago
Current AI models are approximation functions with huge number of parameters. These approximation functions are reasonably good at interpolation, meh at extrapolation, and have nothing to do with generalization.
35.
by lsh123 3y ago
Long time ago in a galaxy far far away, I replaced the need for N * (Oracle license + Sun server) with a simple perl script (< 100 lines) and one Sun server for a total cost saving in $300M+ / year. While doing it, I also invented m
36.
by lsh123 3y ago
https://www.brucesallan.com/wp-content/uploads/2013/11/grade...
37.
by lsh123 3y ago
“Behind every hero, there is someone who fucked up”
38.
by lsh123 3y ago
I wrote and still maintain an open source C project for 20+ years. Once a year I get a new guy coming in and telling me I am doing it wrong: you should typedef all data types, you should stop using const, and so on. It stopped being funny a
39.
by lsh123 3y ago
Their unlawful spies and our brave intelligence officers.
40.
by lsh123 3y ago
I have exactly the same problem with dirty hands when I take boroscope photos of the engine on my plane. I would love to just see it in my glasses and say blink twice to take a photo.
41.
by lsh123 3y ago
https://en.wikipedia.org/wiki/1989_Belgium_MiG-23_crash
42.
by lsh123 3y ago
Interview with the guy who pulled the ejection cord in the Michigan Mig-23: https://www.youtube.com/watch?v=Ftcn3NwAZCI tldr; they were out of airspeed and out of altitude barely within ejection envelope (i.e. exhausted all
43.
by lsh123 3y ago
And this is a good thing! For the most of its history, Earth was much (>5F) warmer than today: http://www.climate.gov/media/11332 https://www.climate.gov/news-features/climate-qa/whats-hott
44.
by lsh123 3y ago
Training is very expensive and requires GPUs. What you read about is running trained model on consumer devices (even phones!).
45.
by lsh123 3y ago
Sometime soon you will discover ForeFlight or Garmin Pilot, and your life would never be the same ;)
46.
by lsh123 3y ago
Long time ago I saved data from a hard drive by putting it into freezer, then copying data, putting it into freezing, copying data, ... repeat a lot of times :)
47.
by lsh123 3y ago
Current ML models neither memorize or generalize, but instead approximate.
48.
by lsh123 3y ago
Hm... so how do we know that the actual values in the produced json are correct???
49.
by lsh123 3y ago
The interesting thing is that the government can have access to your conversations today. They just need to get judge order and search your phone (end-to-end encryption doesn't help if you don't delete messages). Same things as kn
50.
by lsh123 3y ago
This is a property of Windows C++, not other C++'s :) :) :)
51.
by lsh123 3y ago
Sorry I didn't realize you are a pilot and answer was very generic. But here is "unrecoverable" scenarios: - Accelerate beyond Va: need gentle control inputs will be required and it will take more time to recover; - Invert th
52.
by lsh123 3y ago
In 2028, we will still be writing code in C and Fortran. And there will be all these new languages X++ that would be promising to replace either or both ;) ;) ;)
53.
by lsh123 3y ago
The link I posted above is from an instrument rated pilot of an instrument flight plan who nearly got himself and passengers killed in IMC. A couple or few hours under the hood is not nearly enough to handle actual IMC a few years later.
54.
by lsh123 3y ago
There are studies about non-instrument pilots entering clouds (instrument meteorological conditions or IMC in aviation lingo): - in 60 secs the plane control is lost (though still recoverable); - in 90 secs the plane is in an unrecoverable
55.
by lsh123 3y ago
Not quite true. Stop drilling is commonly used on airplanes: http://o-si-sec.com/2018/10/11/5-facts-to-know-about-stop-dr...
56.
by lsh123 3y ago
In addition, weather models weren’t nowhere near as good as they are now (thus hard to estimate wind correction over long flight). And radio beacons / stations were off due to the war as well. However, given enough fuel, it’s pretty ha
57.
by lsh123 3y ago
Each ~30 feet or 10 meters adds an extra atmosphere of pressure. Recreation divers go to 120-130 feet (~4 extra atmospheres). Commercial divers go to ~1000 feet (~30 extra atmospheres). There are rumors that Navi divers can go to ~2000 feet
58.
by lsh123 3y ago
https://en.wikipedia.org/wiki/The_Boy_Who_Cried_Wolf
59.
by lsh123 3y ago
https://en.wikipedia.org/wiki/Scientific_method "The purpose of an experiment is to determine whether observations agree with or conflict with the expectations deduced from a hypothesis"
60.
by lsh123 3y ago
Last 2 sentences from the article: "... The end of global industrial civilization is where we are headed right now, not at some future dystopian moment. I wish I had a hopeful word to end with. But I don’t."

More ›