Interesting video from the cockpit of an A350 flying from Copenhagen to Bangkok. "The challenge on this route is like the jamming and the spoofing. But we know how to deal with that now. You'll see later on when we come close to Ukraine a lot of our systems will fall out"
"The worst thing about this is the spoofing so uh because of the jamming the airplane may think it's it's a different place than where it is right so and then it'll it could give us some failures that we have to handle."
"That also happens in the northern part of uh Norway and actually sometimes going close to Kaliningrad."
[They start to lose GPS.] "It's only a problem if you are going to land somewhere." "For en route navigation is not a problem we use this system. The IRS, the inertial reference system, that's super reliable."
"Turning off the terrain warning system, just because now we get so many of these jamming problems that we want to make sure that we don't get these terrain warnings on the screen in the middle of the night."
"Right now we have no satellites at all, all the satellites are gone so when we see that the satellites are coming back then we can turn it back on again."
Video:
Systems listed in orange are inoperative. RNP AR, GNSS 1, ADS-B TRAFFIC 1+2, APU, GLS AUTOLAND, GLS 1, SLS 1.
"I was sitting here one day, and I saw the clock going backwards."
Watching these pilots deal with systems they can no longer use or trust, I'm reminded of a quote from a major airline's Chief Operations Officer about GPS interference: "Navigation is not my problem. My problem is normalization of deviance."
I think this is the flight: SAS973 on 2024-03-19 (thanks, @giammaiot2). The bad GPS jamming happened over the Black Sea, about 10 minutes before entering Turkey.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Short thread on how I made this map, which shows areas where aircraft are being affected by GPS spoofing, and the locations they're being spoofed to (not the location of the spoofing transmitter):
Wired retracted the op-ed claiming Google alters search results to make more ad money. The story fed into existing beliefs about Google and tech and spread super fast. But the stories we want to believe are the ones we should scrutinize most. wired.com/story/google-a…
First I was skeptical because it fit so well into those existing beliefs. Then I realized the story had one source–I couldn't find any other reporters at the Google antitrust trial who mentioned what was clearly a bombshell. That's when I figured it was a misunderstanding.
Thinking through how ads would get matched to search queries, I realized I could easily imagine query rewriting being part of that, just as described, except not affecting the search results. That seemed to make a misunderstanding even more likely.
OpenStreetMap has an amazingly powerful query language, "Overpass". Sadly it was designed by a madman and is hard to learn. Luckily GPT-4 knows it well: "Give me an overpass query that finds all buildings that straddle the boundary between Glendale and Burbank in California."
Last night there was a loud explosion in LA. I figured it was a transformer explosion. A friend had video showing a 9 second delay between flash and boom so I used GPT-4 to write an Overpass query to find all electrical substations about the right distance to cause that delay.
(If you are the mad person who designed Overpass, I thank you and it's incredible—best super power ever. I just can't learn it.)
I saw a somewhat astonishing thing today. GPT was asked a question that it needed to write code to answer, and given access to a Python REPL. It wrote buggy code, then based on the error message it fixed its own code until it worked (and gave the correct answer). It debugged.
The question posed was "What is the 10th fibonacci number?" The GPT-based agent's first attempt was "fibonacci(10)"—"Action Input" is what the LLM is sending to the Python REPL. This resulted in an error because it's not a standard function. The LLM figured that out.
It then defined a fibonacci function, called it, and got the right answer. All fully automated, with no human intervention (using @LangChainAI). Technically simple but kinda mind-blowing: It debugged its own code until it worked.
An experiment in trying to recreate the aesthetic of 50s/60s aerospace photos in Dall-E, Stable Diffusion, and Midjourney. First up: Dall-E. Prompt: "Photo of SR-71 aircraft flying over the Mojave desert faded ektachrome 100sw vintage retro nasa archive"
Dall-E gets pretty close to the overall aesthetic I'm looking for. It knows an SR-71 is black, but not much else.
Next up: Midjourney. The aircraft is very well defined, though is further from an SR-71. Kind of a mix of the military aerospace photograph and 70s science fiction book cover aesthetics.