A human operator would see a moving crane and say "that's a construction site, I'm going to go around". They would not fly directly under a crane even if it visually looked clear. In this case the crane was actively lowering something, so the drone not only missed the cable but it flew directly in between the crane and a visible object hanging in the air below the crane.
For concrete numbers, I would say stay 50 yards away from construction equipment, and always laterally or above, not below. Honestly these drones are enormous so I think "don't go under" can just be a blanket rule. They can't be going under trees or bridges or overpasses either, they're too big.
Edit: Also, the drones themselves should be far enough apart that if one crashes the other has time to react and stop or change course. I don't have a concrete number there, it depends on their speed and acceleration, but they shouldn't be flying so close that if one crashes they all will.
In non-drone aviation, we require vehicles to be separated from each other by 5 nautical miles horizontally and 2,000 feet vertically. Additionally every area of the planet they fly over has an MSA figure - Minimum Safe Altitude - which is supposed to guarantee 1000 feet clearance over any obstacles or terrain.
Both of these allow healthy margins of error, whether that error is from a human pilot or ATC, or from computer systems - either in the vehicle or the ground.
I'd argue these would be a great place to start for drone aviation.
If such limits make drone burrito or toilet paper delivery expensive, that seems fine.
At least in the US, minimum separation is for when you're talking to air traffic control, and MSA is for flying on instruments. Minimum separation when you can see outside and aren't talking to ATC is "don't hit other planes." Minimum altitude is 500ft, or 1000ft over "congested areas," plus 500ft distance from any obstacle. Unless you're flying a helicopter (or powered parachute or hang glider) in which case the only requirement is "the operation is conducted without hazard to persons or property on the surface."
It would make sense for a quadcopter to follow helicopter rules. Obviously it does not follow the "without hazard" requirement if you crash into cables, though.
Fair enough, I am not a pilot, just a nerd so I wasn’t aware of this!
I’d agree the helicopter rules seem most appropriate, though I guess I’d still feel like that would still rule out operating anywhere near a building under construction.
That said, a regular helicopter that suffers a loss of power or other fault, still has options like autorotation to at least attempt a landing without killing anyone on the ground. Do drones have any equivalent ? I.e. if battery is below x% it returns to safe landing spot?
"Without hazard" is pretty subjective, but I agree that it probably shouldn't include casually flying around cranes.
I don't know how drones are programmed, but landing immediately if the battery gets low certainly sounds like a sensible precaution. Electric motors might be reliable enough that you don't have to worry about gracefully handling failure of those. I hope so, because I don't think a quadcopter has much hope if any motor fails.
Was this particular operation careless and wreckless? Could be.
Someone else in the thread said the weather conditions included mist. I'm skeptical misty conditions also permits a minimum 3 miles of visibility, but what do I know I'm just a pilot.
But also, it's possible the waiver I assume Amazon is operating under could include visibility. I assume this because Part 107 requires visual line of sight operation, but Amazon's operation sounds like beyond visual line of sight (BVLOS). I don't know anything about that. I'd like to think the waiver and the operating requirements are public information, but I don't know that either.
Thanks for the context! Makes sense for traditional aircraft to be super conservative like this, especially given they tend to travel very long routes and sometimes have nothing more than a pair of human eyes paying attention to obstacles.
Do you know what are the rules for helicopters in a city? That seems like a closer analogue.
Why do people always get a hard on for bizarre drone rules? Some amateur pilot goes up in a fireball every other week, I think we put up their safety record to that of drones and there are going to be some hard choices to make - I think it's over for our amateur pilots out there.
TBH my concern is less for the fate of people who choose to fly planes or drones - they are the ones making the choice so if they hurt themselves that sucks, but none of my business. However I am concerned for the innocent bystanders who might be under them, who did not choose to risk death or injury by an aircraft, drone or otherwise.
The template for this ofc is how we handle (or don't handle) the danger posed by people inside cars, to people outside cars. In aviation we will do a lot to avoid even one death, in the air or on the ground. But in cars we mostly don't give a crap. It took decades for drunk driving to become unacceptable, but outside of that, we are still pretty ok with death by car. The only survivor of the collision just has to claim the person outside the vehicle "darted" and we all shrug and move on.
I would just love it if we could apply lessons of the past to new technology. Instead we just hand wave it all away. Then in a few decades, if enough people die and their surviving loved ones invest enough time & energy, maybe we'll make a few tweaks to the formula.
So no, we don't need 5nm separations for 20 lb drones. But we do need some sort of structure that recognizes the people under them didn't sign up to be part of the beta test. For bonus points it should also recognize that externalities exist and should be priced but I am not holding my breath.
At a mere (say) 240mph, 5 miles is 75 seconds. Talk to an experienced pilot about how short a time that is, when a bottom-10% pilot is trying to figure out some problem with his instrumentation, or has set his radio to the wrong frequency, or whatever.
100mph is probably around the top speed for home built FPV quadcopters with 5 inch propellers; that's neither uncommon or expensive. Amazon's drones are probably built for efficiency though so they will go slower than that.
5 miles at 50mph would give only Amazon 360 seconds to fix the bug that caused the first drone to crash. Or figure things out enough to get the second drone into manual override mode.
> Good politics is just being strategic about relationships and influence in the service of good outcomes.
Yeah, no shit dude. That's exactly the part that's disgusting. Using the word "just" here feels dishonest.
I was subjected early on to someone who viewed every single interaction in every single relationship as transactional and framed every decision around the question "what's in it for me?"
It really warped my worldview for a long time and it took a ton of therapy and self-reflection to overcome. I'm not going to sacrifice my principles just to get something I want.
This is what I took away from the post too. I'm not going to invite someone out for a coffee so I can use them down the road. I feel the same way about "networking". The people around you are not tools.
I want no part in "it's not what you know" kinds of situations. I'm paid for what I know. The author seems to think being apolitical means not giving your input or making decisions. If I'm not allowed to do that without sucking up to the higher-ups, I'll find another job. Everyone I respect is above politics.
There's no part of me that wants to maintain relationships for the express purpose of extracting value in the future for gain -- personal or otherwise.
I simply refuse to let the end justify the means, whatever that end is.
When you’re responsible for supporting people who refuse to receive patches like this one [1], and those same people have the power to page your phone at 11pm on the weekend… you quickly learn how to call a spade, a spade.
There is undoubtedly a better word than stupid. They're very likely not stupid. Careless, maybe. Inept, maybe. Irresponsible, maybe. Stubborn, maybe. More generously: overworked. Just probably not stupid.
That wasn't what he used the word for. I understood his point perfectly: there are AI teams that are not knowledgeable or skilled enough to modify and enhance the docker images or toolkits that train/run the models. It takes some medium to advanced skills to get drivers to work properly. He used shorthand "too stupid to" instead of what I wrote above.
Still, it adds an air of arrogance to the whole post. For a while the only pytorch code that worked on newly released hopper GPUs we had was the Nvidia ngc container, not Pytorch nightly. The upstream ecosystem hadn't caught up yet and Nvidia were adding their special sauce in their image. Perhaps not stupidity but lack of docs from nvidia
> For a while the only pytorch code that worked on newly released hopper GPUs we had was the Nvidia ngc container, not Pytorch nightly. The upstream ecosystem hadn't caught up yet and Nvidia were adding their special sauce in their image.
I'm sorry to come across as arrogant, but it's really just frustration, because being surrounded by this kind of cargo-culting "special sauce" talk, even from so-called principal engineers, is what drove me to burnout and out of the industry into the northwoods. Furthermore, you're completely wrong. There is no special sauce, you just didn't look at the list of ingredients. There never has been any special sauce.
The build scripts for the base container are incredibly straightforward: they add the apt/yum repos and then install packages from that repo.
The pytorch containers are constructed atop these base containers. The specific pytorch commit they use in their NGC pytorch containers are directly linked in their release notes for the container: https://docs.nvidia.com/deeplearning/frameworks/pytorch-rele...
Do I need to keep going? Every single one of these commits is on pytorch/pytorch@main. So when you say:
> For a while the only pytorch code that worked on newly released hopper GPUs we had was the Nvidia ngc container, not Pytorch nightly.
That's provably false. Unless you're suggesting that upstream pytorch continually rebased (eg: force pushed, breaking the worktree of every pytorch developer) atop unmerged code from nvidia, the commit ishes would not match. Meaning all of these commits were merged into pytorch/pytorch@main, and were available in pytorch nightlies, prior to the release of those NGC pytorch containers. No secret sauce, no man behind the curtain, just pure cargo culting and superstition.
A huge portion of journalism is in fact reporting what people say. An important part of a certain kind of journalism is investigating and reporting on those claims. Sometimes the facts are opaque but claims can be corroborated in other ways. The clue here is the "other experts." If multiple independent sources are making the same claims, that's newsworthy, even if there's no tangible proof.
Also keep in mind this is not an academic article or even an article for tech folks. It's for general population and most folks would be overwhelmed by details about prompts or methodology.
Multiple 'independent'* sources making up the same shit is known as 'manufactured consent'. Especially if it's at the behest of a regime with an agenda to push.
* Mass media is not and has never been independent. It's at the service of the owning class.
I'm a die hard linux user, and some years ago took a windows gig on a whim. I find powershell fantastic and the only thing that makes my role bearable. Now, one of the first things i install on Linux is powershell.
For the sake of clarity: I am not arguing against your point, nor am I defending Amazon or the tech in any way shape or form.
reply