Hacker News Reader

Comment by DrNosferatu | Hacker News Reader

It's technologically trivial, today, to perform an automated secondary backup (local or otherwise - but ideally remote) of days - months, years - of this telemetry and voice recorder data.

(as well as to keep it private & encrypted, and only accessible with a warrant from a judge)

If the FAA does not mandate, at least, this from now on, it will just add to the pile of evidence that they are in Boeing's pocket.

PS: Secondary backup. No existing system has to change. Just an outer layer of backup tapping into the existing data recording loop.

View on HN

Aviation has a healthy blameless culture which has created willingness on the part of air crews to self-report deviations. If you make a mistake and own up to it you suffer no consequences. Keeping recordings forever may disincentivize crews from communicating freely if their comments could come back to haunt them years later. This may still be a good idea but it’s not as simple as changing the retention period on recordings.

Yes, the blameless culture is a critical safety feature and it could be ruined very easily if the failed leadership in the industry starts looking for scapegoats.

It's sad too because analyzing every cockpit conversation with AI to highlight things that may cause common confusion could be invaluable. Instead, the short-sighted leadership in today's business world will use it for (job) performance analysis and to penalize workers for failing to act like robots :-(

Possibly. But as a technologist I don’t trust AI to actually deliver such a result in our generation. The system will inevitably be created by modern technologists without intimate domain knowledge. It will be motivated by short-term deadline chasing and MVP culture. The cockpit of an airliner is no place for bleeding edge technological innovation. I don’t need an Airbus with Siri. Pilots already spend hours in training every year. Is reviewing the output of this AI system really a better use of their time? What problem is it even solving?

What would it take to make you trust AI?

Decades of development and successful deployment first in non-safety-critical settings then further demonstration in progressively higher stakes environments. In addition demonstration that the technology has any benefit to this domain at all.

My concerns have more to do with the technology industry than AI itself. AI is a buzzword and the technology as it exists is of dubious benefit here. We have an unhealthy tendency to deploy technology too early and for the wrong reasons. These are undesirable impulses in aviation.

Analysing cockpit voice recordings for confusing terms (using AI or not) does not strike me as something that would affect the operation of the aircraft in any shape or form?

You’re assuming the AI works and does what you expect. Neither of those are reasonable assumptions.

Everything has an opportunity cost. To be beneficial AI has to be better than the status quo to be. In aviation that’s a much higher bar than “not nothing”. It has to be better than the time saturation flight crews already experience.

I’m not assuming the AI works at all. Maybe we are talking about different things?

If the flight is over, and some analysis spits out a bunch of garbage, how does that hurt anyone? Nobody is going to tell pilots that they should stop using the word ‘and’ even if the AI spits that out as a possible point of improvement.

If the AI doesn’t work it is a waste of everyone’s time to even look at the output. As I already said nothing is free. There’s always an opportunity cost. Collecting data for unknown benefit has a cost in dollars, time, and potentially human life. “The AI doesn’t have to work” is exactly the thoughtlessness that aviation avoids and technology breeds.

Thank you for your honesty. Are there any other industries that you hold at a similar standard?

I can't speak for GP, but "safety critical" is the term relevant to this thread I think. Any situation where people die if you screw up, where that includes screwing up at a higher level, such as introducing ill-thought-out process changes.

Thank you. That's a good grounding that helps orient toward what I feel is being conveyed.

I sense there's a lot of overlap with SCADA-enabled infrastructure: transportation, nuclear, industrial control, &c. And we should be hesitant to deploy AI in those environments. Am I getting that right?

Only warrant-issued-by-judge-mandated disclosure.

Full privacy preservation.

Aviation is safe because the participants all trust each other. Blanket recording for undefined future use undermines that trust. This has a chilling effect on communication. Crews will be reluctant to communicate with each other. The risk of a warrant being issued at all creates an adversarial relationship between crews and investigators. This is a horrible idea and suggesting it betrays a cliche misunderstanding of the domain that is the stereotype of modern technologists.

> If the FAA does not mandate, at least, this from now on, it will just add to the pile of evidence that they are in Boeing's pocket.

I don't see how it's in the interests of Boeing to keep the mandate at 2 hours. If it gets extended, it's likely that in future incidents they will have more evidence of pilot error than they do now since that is the primary cause of accidents (even for Boeing), plus I am sure they can print some nice invoices for the costs of the upgrades to existing fleets.

Well, could be, but another possibility:

- most pilot errors resulting in an incident occur close to the incident and thus the # of times it's their fault will drop off with longer recording time

- the longer recording time may allow for some complaints about the Boeing aircraft or some clunking noises to be identified which could indicate an issue with the aircraft

The key concept here is: “clunking noises”.

(aka “aircraft defects”)

PS: I mean it.

Not all clunking noises are defects and not all defects make clunking noises.

Free hint: “metaphor”.

Or they could just implement the 25-hour recorder like they have in Europe. This was a policy choice, not a technical limitation.

It’s not that simple. Europe has a different regulatory environment. Do pilots in the US have the same protection from retaliation for example? Has the 25 hour recording period improved safety? Have there been undesirable side effects?

PPS: Give me USD 15M + 10K per aircraft + 10 cents per transported/flown passenger ever protected by this System, and I'll design and implement it for you - with quadruplex redundancy.

(disclaimer: Founder tier pricing. General pricing may vary - up)

Someone just invented DropBlackBox.

I'll make sure you get some royalties for that name suggestion ;)

BBaaS

Just get an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem.

...rsync

rcs. you meant rcs. and uucp.

But it needs to be more enterprise-y, X.400 with maybe X.25 should do nicely.

Just dont use Azure....

This comment is peak HN arrogance, or pretty close to it...

Well I think that one flew right over your head. Something Boeing couldn't achieve, so clearly the poster must have worked for Nasa.

Err - it is an EASA requirement that Boeing already conforms to, I believe.

I think he was punning about the flying over the head being something Boeing couldn’t achieve

I read it more like an annoyed take on the aviation industry's change management process.

Based on my experiences with modern technology I don’t want aviation learning anything from tech.

Oh, but you haven't seen nothing yet - what about, on top of that, even some extra credibility:

Former NASA Engineer, baby B-)

Impressive. Just tell me, when did NASA develop and build civilian aerospace components?

Thanks for the venture offer, but I'm busy frying bigger fish right now.

Not the point - these are just peak HN arrogance credentials ;)

Absolutely the point, because none of what you propose is easy, let alone trivial... As you would know if you came close to civil aerospace development, certification or change management.

You also ignore that it is more a policy than a technical problem, Europe has 25 hours and not 2 like the US. No idea how easy it is to retrofit the 25 hours into US certified aircraft with 2 hour recording, which might pose another problem.

Bit hey, use your NASA credentials, apply to YC or any other VC, and launch your "trivial" tech solution.

Startup culture is basically entirely about disruption and challenging status quo. It's not really any more arrogant than general startup mentality, which admittedly is very often arrogant (facebook for dogs, seeking $1B in investors!)

Actually I found it makes a great point.

> peak HN arrogance

Take peak HN arrogance, balance it against peak bureaucratic promises plus peak actual cost, and let me know what you come out with.

Is it arrogant to wish to survive a common civilian airplane flight?

No, but pretty ignorant to propose something that makes it easier to investigate crashes and incidents while having zero impact on incident prevention.

Assuming zero impact is the arrogance. These systems are complex and human. You can’t make changes without side effects. Nothing is free and actions can have unintended consequences.

All the same, not changing anything changes nothing.

And any change involving tech/gear are preferable to changes involving procedure (either more of it, or changes to it).

Extending the recording length on voice recorders to a higher number seems pretty straightforward (and already done elsewhere).

Extending to 25 hours is a lot different than extending it forever. Using “cheap” and “easy” recording devices is frankly reckless. Look around this comment section at the ignoramuses willing to throw SD cards at commercial airliners. Everyone’s an expert.

Yes, voice recorders should last long enough to capture everything that happens on a flight. But at the same time pilots should have protections to make them feel safe from retaliation.

Aviation safety is an incredible achievement that we as technologists should learn from. It’s not fertile ground for self-proclaimed “engineers” to piss their pet theories upon.

You're the one thinking of SD cards - no one else (that I'm aware) mentioned anything of the sort.

The aviation industry, and its monopolies, have to stop being treated as sacred cows and be subjected to thorough - consequent - scrutiny and accountability at all stages of the aircraft lifecycle. It certainly hasn't been the case with the 737 MAX - the system, as is, clearly failed the almost 400 people that already died. And if the MBA-types continue to have their way, the deaths won't stop there.

Finally, I'm not going to engage in insults, they only qualify yourself.

To be fair, investigating incidents and crashes is how you get incident prevention. Most safeguards today are a result of the lessons learned from NTSB investigations.

I had assumed it was satire so you’re absolutely right if it was not!

How dare they!?

> If the FAA does not mandate, at least, this from now on, it will just add to the pile of evidence that they are in Boeing's pocket.

It's the pilots' union that is opposing this. I doubt Boeing cares either way. If anything they'd want the extra data.

I think you would find pushback from far more than just Boeing. Airbus, APA, ALPA, etc. Nobody in the industry would find it reasonable to broadcast something that isn’t meant to come out.

I think the reasons for the missing data need to be established before we go calling for sweeping change.

EDIT: It appears the recorder had continued operation and thus overwrote the data. This is a failure in procedure. Procedures are in place for securing evidence in incidents such as this and for some reason the ball was dropped.

It seems that the main disadvantage of the current system is a 2-hour loop of storage, which can be overwritten in long flights and hence missing. The change needs to be either longer storage which would be a approvals nightmare, or something like OP proposed which isn't too ridiculous. I have certain compliance paperwork that needs to be retained for 5+ years, why not cockpit audio also?

Plenty of other commenters have established why it’s ridiculous. Cost, bandwidth, privacy, questionable value.

The ICAO 25 hour standard seems reasonable to me as someone in the industry.

>> only accessible with a warrant from a judge

What's your technical solution to this aspect? The solution must be "technologically trivial" which I take to mean implementable today or in the near future with no change to how current regulations or laws work or to how current workstreams outside of the secondary backup system work (i.e., no change is required like having the judiciary start using crypto). We are also using the strict definition of "only." It should be technologically impossible for any person or entity to access the data without a warrant.

Asymmetric cryptography isn't a rocket science. Airline keeps encrypted recording and public key, certain government institution keeps private key, you need both to decrypt.

This violates rule 1 by requiring a government institution to keep a private key: no change to how current regulations or laws work or to how current workstreams outside of the secondary backup system work (i.e., no change is required like having the judiciary start using crypto)

This violates rule 2 by not assuring that a warrant is necessary because somebody at the government institution has the private key (assuming that institution is not the judge): It should be technologically impossible for any person or entity to access the data without a warrant

> It should be technologically impossible for any person or entity to access the data without a warrant

I don’t understand why you’d impose restrictions on this system that do not exist elsewhere? What makes cockpit voice recordings sensitive enough to warrant more secure storage than people’s bank accounts?

No it doesn’t - it will just cost you the BoTaaS service I’m providing, payable in Swiss Francs.

I mean my System will have a layer of encryption protection. And if the FAA, Unions, etc., so wish, it can be made to completely preserve privacy and only unlocked when the applicable judiciary (or whoever) sees fit.

So you don't have a technical solution. Got it.

But I absolutely do:

Elective during (and after) the System's procurement, you can also option it with a "Board-of-Trustees-as-a-Service" to decide when to open the Seal of Privacy and disclose the System's Recordings - if ever.

(prices for this option are in Swiss Francs)

why do you think the lack voice recordings protect Boeing instead of the people being recorded?

There are human lives on the line - thorough records are clearly warranted.

(not same as open-access: judge must authorize playback)

> There are human lives on the line - thorogh records are clearly warranted.

I'm not suggesting otherwise.

> it will just add to the pile of evidence that they are in Boeing's pocket.

I don't understand the connection between the lack of voice recordings and how that is proof of the FAA protecting Boeing rather than the lack of voice recordings as proof of the FAA protecting the pilots.

FAA has the final word - over manufacturers, unions, etc.

Reply on news.ycombinator.com