Software authors have professional discipline of cute puppy compared to aviation(threadreaderapp.com)
threadreaderapp.com
Software authors have professional discipline of cute puppy compared to aviation
https://threadreaderapp.com/thread/1027539289750429696.html
14 comments
Absolutely. The stakes for aviation is life and death. Software generally does not have life and death stakes that necessitates the additional discipline cost.
One thing about aviation is accidents likely require high management level attention by manufacturers and operators.
Compare with automobiles, the CEO of GM doesn't get a phone call every time a car kills someone. And you can see the behavior of CEO's when things become an issue, they're annoyed and totally unprepared for it. Vs aviation managers who see this stuff as a fact of life.
Notable management of internet companies act exactly like auto exec's being called on the carpet for failures. Downplay and redirect blame.
Quote from the CEO of BP during the Deep Water Horizon accident: 'I want to get my life back'
Compare with automobiles, the CEO of GM doesn't get a phone call every time a car kills someone. And you can see the behavior of CEO's when things become an issue, they're annoyed and totally unprepared for it. Vs aviation managers who see this stuff as a fact of life.
Notable management of internet companies act exactly like auto exec's being called on the carpet for failures. Downplay and redirect blame.
Quote from the CEO of BP during the Deep Water Horizon accident: 'I want to get my life back'
Partially this, partially the cost of buggy software rarely falling on those who actually make the software. It's pretty typical for software companies to disclaim all liability for the results of using their software and offer no warranty that the software they sell will work. Not a lot of other products can get away with such an absolute disclaimer of liability.
Agreed, some companies care when there is a large financial risk but others do not. There’s also a lot of talk about automated testing all the time but a lot of it is lip service. The investment is often woeful.
Risk, and Gresham's Law dynamics, greatly influence this. Absent some means of enforcing a minimum standard of design and support, as well as internalising risk costs to the vendor, these problems are unlikely to be resolved.
On one hand, yes. On the other, given different stakes between what most software companies produce and planes/elevators/medicine it is fine that we don't have same level of safeguards. Text editor failing without loosing too much data is already failsafe.
We could do much better at data safety, privacy etc. But if you worry about particular data abuses by Facebook, that won't get better without regulation - the issue is not tech maverick, the issue is that the company earns money on that. Aviation became more safe only after regulations too.
And desinformation on Facebook is yet another entirely different issue. First, Facebook dont really care which is why there is no report. Second, I am not yet convinced that they have to care nor that asking sociopathic corporation to safeguard discussion is good idea.
We could do much better at data safety, privacy etc. But if you worry about particular data abuses by Facebook, that won't get better without regulation - the issue is not tech maverick, the issue is that the company earns money on that. Aviation became more safe only after regulations too.
And desinformation on Facebook is yet another entirely different issue. First, Facebook dont really care which is why there is no report. Second, I am not yet convinced that they have to care nor that asking sociopathic corporation to safeguard discussion is good idea.
it's too bad the ACM isn't wider spread in the industry itself among private companies: https://ethics.acm.org/2018-code-draft-1/
This may not be a bad thing.
Aviation as an industry has been very carefully managed by the US and UK (and maybe other) governments. This has led to a very, very low accident rate, but it's also led to a very, very low innovation rate.
Maybe this is a good thing, maybe it's a necessary thing, but you can't just take one facet of a complicated issue in isolation. Tradeoffs exist on industry- or nation-wide scales.
Aviation as an industry has been very carefully managed by the US and UK (and maybe other) governments. This has led to a very, very low accident rate, but it's also led to a very, very low innovation rate.
Maybe this is a good thing, maybe it's a necessary thing, but you can't just take one facet of a complicated issue in isolation. Tradeoffs exist on industry- or nation-wide scales.
Ironically, software is replacing pilots and their checklists. Check lists are essential in planes because they are complex machines with traditionally low levels of automation. This creates extremely high workloads for pilots. The only way to deal with that is checklists.
The solution is automation. Compare a modern fadec engine with a 50 year old piston engine. The engine manages itself. You basically turn it on and it regulates itself. There are still checklists of course but they are a lot shorter. Similarly fly by wire, auto pilots, etc. have reduced workloads. There is still plenty of opportunity to fix things.
Software could do with a lot more checks and balances. But you have to balance the workload, overhead with the added value. Also, you need to think about automation. Continuous deployment is basically doing exactly that. It can only be done if you have automated tests and deployments.
The solution is automation. Compare a modern fadec engine with a 50 year old piston engine. The engine manages itself. You basically turn it on and it regulates itself. There are still checklists of course but they are a lot shorter. Similarly fly by wire, auto pilots, etc. have reduced workloads. There is still plenty of opportunity to fix things.
Software could do with a lot more checks and balances. But you have to balance the workload, overhead with the added value. Also, you need to think about automation. Continuous deployment is basically doing exactly that. It can only be done if you have automated tests and deployments.
This gets to the problems of excessive automation: forgetting, skill loss, and detraining, among experienced operators, and a lack of experience gain amongst new entrants.
We're starting to see this in particular in areas without entrance barriers or continuous training requirements, noteably automobile operation. This is compounded by novel and nonuniform interfaces.
We're starting to see this in particular in areas without entrance barriers or continuous training requirements, noteably automobile operation. This is compounded by novel and nonuniform interfaces.
It's sad at a sentimental level but at the same time, the aviation industry provides plenty of statistical material on safety.
The general trend of getting humans out of the way through automation is a good one. Checklists are critical to make sure humans don't fail to do something. Anything they don't have to do is an item less on the checklist and one less thing that can go wrong. With many modern planes the startup check list is "push the start button". Of course you still have the pre-flight checklists and other checklists but engine management is typically fully automated. That means no more complex failure modes because of cowl flap management, wrong mixture or prop settings, over torquing the engine, etc.
The general trend of getting humans out of the way through automation is a good one. Checklists are critical to make sure humans don't fail to do something. Anything they don't have to do is an item less on the checklist and one less thing that can go wrong. With many modern planes the startup check list is "push the start button". Of course you still have the pre-flight checklists and other checklists but engine management is typically fully automated. That means no more complex failure modes because of cowl flap management, wrong mixture or prop settings, over torquing the engine, etc.
Aviation controlls the operator class.
Automotive does not.
Automotive does not.
Related:
Computing’s Hippocratic oath is here
https://www.fastcompany.com/90215922/why-we-spent-two-years-...
Computing’s Hippocratic oath is here
https://www.fastcompany.com/90215922/why-we-spent-two-years-...
In other words, I think software can be way better in terms of quality, but nobody's interested to pay for incurred costs.
Does it mean programming should be regulated like aviation was regulated? Not sure about that at all.