Imagine in hitler had access to this tech and simply purchased a database of "Likely Jewish" facial data?
Do you genuinely believe those drones flying over every major city taking high resolution photos of everyone within 30 miles are there for fun?
Short term: camera resolution on the average cell phone is often better than an array onboard a satellite, so the MVP is a native app that window seat passengers use to record in-flight. Because we know the flight route before takeoff and GPS works in Airplane mode, we load in target coordinates and ping the user to begin recording when flying over a target coordinate. They receive free in-flight wifi for their effort, but we also use the wifi connection to stream the video feed. That gets us more exposure to the 77,000 daily flights across the U.S. each day (also boosts general awareness: "Why are you recording out the window?")
Long term: we're developing a hardware solution much more powerful than a cell phone camera with a 6-camera array and many more sensors, including an infrared dome for near-field cloud penetration and onboard preprocessor for compressing the stream before sending it back down to earth. The form factor is akin to a headphone case (I actually used a Bose headphone case for the prototype) and is mounted to the window within the cabin, mitigating the need for FAA (but not FCC) approval. There are some teaser images on our Instagram if you're curious - @notasatellite.
Plus video quality is usually pretty crappy and you'd need heaps of post-processing to stabilize the video/remove dirt spots on the window, etc. I find it difficult to believe that this would work well with random people filming with their smartphones out a plane window.
There are still a ton of variables that affect image quality. Type of aircraft, weather conditions, altitude, camera, seat position (sitting forward of the wing is slightly more advantageous because the inference engine doesn't have to correct for the exhaust blur in the aft), angle of view, residual engine vibration/dampening, time of day, flap/aileron position, the list goes on. Part of the go-to-market strategy involves making sure the hardware product and the inference engine handles a number of those limitations. Neither satellite nor aerial imagery is perfect, but you have a vastly higher chance with ~80,000 flights/day vs. the 300 commercial imaging satellites in orbit, with automatic cost savings in the millions that we can pass on to our customers.
Not-so-subtle marketing pitch--you can see a bunch of real-life examples on our socials @notasatellite.
- container volume in and out of ports (which are usually in industrial zones right next to airports)
- daily crop harvest rates from the midwest states
- a small undergrad research team in Oregon is analyzing the mist and water turbidity of Willamette Falls to determine flow rate and the impact to salmon runs
- estimating the amount of oil moving in and out of Cushing, Oklahoma by measuring the shadow cast on tanks over 3,500 times per day
- wildfire detection and monitoring on non-daylight flights
There are a bunch of other demos and use cases on our social accounts as well--we're @notasatellite on all the platforms.
On the anthropology side, it might also be useful for looking at statistical information about populations/settlements that's difficult to collect other ways due to either the remoteness of the area and the frequency of other data sources. Things like nomadic migration patterns or land use in areas like Siberia / Mongolia, Myanmar, or northern Canada.
Our goal is to even the playing field when it comes to accessing spatial data and I feel that's adequately reflected in the core feature of our mapping software; whether you're a forensic analyst at Interpol or the neighbor kid looking up BMX trails, you'll always be looking at the most up-to-date map data. Paid customers receive features of course, but what appears on the map is the same regardless of whether or not you're paying.
Your stakeholders have all been burned before either personally or by other vendors and you will have to somehow built trust against those mental barriers.
I personally treat any company who can access my information as if they are willing to undermine their previous statements, have non-public contracts which sell/trade my information with disreputable companies, can pivot in a moment of desperation to do everything they previously promised not to do, may be M&Aed in such a way that all previous contracts are significantly modified, could have terrible security controls of their data, or may not actually delete all copies of data when they say they do.
The mental barrier is difficult to overcome, and you can't win them all. Distill the message too much with "We make money by selling data" and you'll never see a customer, but the opposite is true as well; "The startup doth protest too much, methinks"
Assuming 30m intervals and a 1m GSD I can know when someone is or isn't home based on whether or not a car is in their driveway. For people living in the vicinity of an airport where the GSD and intervals would presumably be much higher I could track individuals to and from their home or office from the comfort of a coffee shop.
Either of the above capabilities has ramifications for things like:
- stalking and harassment (no need to follow someone physically)
- home invasion and theft (can determine when someone is out of the house)
- targeting of dissidents (can track who showed up at a meeting)
- kidnapping and rendition (can know when someone is isolated without committing physical surveillance resources)
Those are just a few of the things I can come up with off the top of my head.
Even if you limit your tools to governments and businesses what prevents illegitimate organizations from using shell companies [2] or other means for establishing legitimate accounts to your services, and what prevents individuals within legitimate organizations from accessing the tools for personal means? [3]
Calling this a "PR issue" grossly understates the potential damage a technology like this can cause in the wrong hands.
[1] "Aerial Imagery Based on Commercial Flights as Remote Sensing Platform" 03/2020 https://www.researchgate.net/publication/339988238_Aerial_Im...
[2] https://money.cnn.com/2015/12/09/news/shell-companies-crime/...
[3] https://www.reuters.com/article/us-usa-surveil-lance-watchdo...
*edited for formatting
I'm more or less a team of one at the moment and the original instinct when I had the idea was along the lines of "If I'm having to address privacy concerns, I must be doing something right", implying some degree of public interest in the product. There's now quantifiable interest and the waitlist hasn't stopped growing. Comments like yours make me realize I need a rock-solid set of first principles before enlisting outside help. The eternal optimist sees infinite use cases and it's easy to discard the bad and the ugly ones for the good, but the ramifications you've listed will kill the product before it even truly starts.
I'd love to chat more about this side of product development and throw some questions your way if you've got some time to spare (chris@notasatellite.com), but thank you again for the thoughtful response and reading material.
People voluntarily post their minute-frequency gps location, workplace, political and religious leanings, sexual preferences, views on current events, and an exhaustive list of friends, family, coworkers, and other associates. Not to mention hundreds, if not thousands, of geo-tagged pictures of themselves over the years, along with a list of the other people in the frame, as well as a written discussion of the events photographed.
A friend is a detective who essentially looks for and forms relationships with people with security cameras, especially cameras that can be linked to a reference time source (like a red light camera)
The cost of things like LPR and soon facial recognition is so low (and going lower) that you’re going to have many federated networks tracking movements of cars and people.
The red light camera is taking video footage with a trusted and calibrated time stamp.
If you see an object appear in frame in both the red light camera and another camera, you can use this moment to establish a timestamp for the other camera.
So perhaps you see a subject walking in the red light camera (establishing time), but another camera lets you observe details about that subject.
Seems more likely than not that it was done at recent protests, no?
edit: I'm not suggesting that you're skipping out on the toll. They may be taking a photo at the same time as the attempt to scan the RFID sticker. And then sending a ticket to the scanned plate number if it fails to scan a RFID.
This basically happened with lower-tech stuff; one of the ways that the Holocaust was facilitated was by conducting detailed censuses of the population, courtesy of IBM: https://en.wikipedia.org/wiki/IBM_and_the_Holocaust