38 comments

[ 15.0 ms ] story [ 82.1 ms ] thread
Oh my god, are we going to have machine learning "Minority Report" from space analysis of earth?
No. But they're gonna spot your cannabis plants with spectral analysis from high altitude now and put you in a work camp.
They’ve been doing that via plane since at least the 70s.

There’s a local commune by me that has a “ragweed festival” where a pilot mistook ragweed for pot and raided their compound. They sued the state, got some money and now put on an annual festival in light of the plants they did find on the property haha

Did he use missiles, drop bombs or strafed with the machinegun? I'm mistified.
The pilot provided firing solutions for heavy SWAT ordnance, which then subjected the ragweed people to low-level, round-the-clock saturation "being in jail".
Urban legend (and an article in the Austin American-Statesman) says that the nearby army base in Killeen would train helicopter pilots by using the known (to them) grow houses in the Austin area because they were easily visible in their IR cameras. I'm guessing the use of LEDs has stopped that though.
Like the only IR emitter in the whole of the city is a few grow houses. A single car emits vastly more IR purely based on how much hotter it is.
You think someone training in military equipment can’t discern between a car and a house? Like totally reaching to make a point but fell on your face.
And why would they train on houses when they can train on a moving car?
are you intentionally being obtuse?
We’ve been doing this for years. FarmLogs (YC 12) used satellite imagery to determine crop health and identify pests/disease.
Yes, I worked on it in 1972 for Earth Resiurces Technology Satellite, ERTS, developing an image orocessing library to correct for satellite angles, Sun position and cloud cover. I bellieve that was for NRO/CIA as it was used for grain yields during the Cold War. It read multiple color bands plus IR and we could recognize corn and wheat blight. This allowed it to be used for crop yield predictions, so we knew Russian yields before they did. It was a bunch of linear array processing functions. That satellite is now called IntelSat, likely a KeyHole 10/11.
Out of curiosity, what can you do with data abour Russia's farm yields?

Probably a lot, but besides maybe predicting price changes of the crop, I can't think of much (I guess that's why I'm not a gov analyst).

Probably concessions at the bargaining table at the intersection of arms and food.

Where before the harvest they might balk at some of the language in an agreement after they learn the harvest was going to be disappointing they may agree to certain demands and make concessions.

The discussions among team members was that once the U.S. had reliable predictions of crop yields then the State Dept could respond in the grain markets. Thee were grain reserves that were available that could be used to flood the global markets resulting in lower grain prices and lower revenues for Russia. The undersanding was that "Grain is Gold", and it appears now that that is still true.
That's pretty wild. Thanks for the details
@saganus Completely unrelated but I saw you ask a question back in 2019 on how to repair a TRS-80 Model 4P. You might want to look at the YouTube channel Adrian's Digital Basement. He repaired a Model 4P about 10 months ago. Did three videos on it.

Hopefully you already figured out how to fix it though!

Awesome! thanks a lot!

Will check the videos out. I have not had the chance to repair it yet.

Cool! Glad to help, in a small way.

The Model I was my first computer. So I have a soft spot for the TRS-80 line :)

As a note. Adrian lives in the Portland, Oregon area. If that would be of any use to you.

Look up DARPA gorgon stare. You're like ~20 years late.
Military was doing it in Iraq back in 09 with planes https://www.eglin.af.mil/News/Article-Display/Article/392767...

No machine learning, but same concept. There was a constantly orbiting plane with super high-res cameras on the bottom, taking pictures of entire cities, roughly 1 picture per second. When an IED went off they could zoom in and "rewind the clock" and watch where the explosion occurred, watch the person burying the IED, how they arrived, and where they came from up to the city limits. Was apparently an invaluable tool. You'd need tunnels to defeat it.

That was almost 15 years ago. I'd be shocked if a satellite version for even larger regions wasn't in the works.

(comment deleted)
If you're interested in machine learning applications like this, others are being researched here [1] and there's an upcoming online event about them.

[1] https://fdleurope.org/

apparently, due to a strange combination of "elite production over-capacity" , climate angst and "capital and power concentration effects" .. there are emerging groups of unpaid, highly-skilled "helpers" who will never have children or property of their own, yet they use crucial data products like this one, to guard with their lives the preferred table refreshments of privileged diners. Helpers are writing automation of machine vision right now, using full-YOLO9 GPU capacity, causing NVidia stock to climb.
Are you claiming that people are volunteering for jobs that should rather be paid employment?
> the preferred table refreshments of privileged diners

You can buy a bottle of wine for $4.50 at Trader Joe's.

TWO BUCK CHUCK WENT UP AGAIN!?
alcohol is unsafe in any quantity. when are we going to move beyond this poison?