From:         "peter (p.j.) ashwood-smith" <>
Organization: Bell-Northern Research, Ottawa, Canada
Date:         15 Jun 95 14:25:26 
References:   1 2 3
Followups:    1
Next article
View raw article
  or MIME structure

|> I didn't say that we are ready to suppress pilots right now.
|> So OK, maybe a fully automatic aircraft is not safer than the
|> current pilot-system _cooperation_ (but systems replaced
|> humans for many functions, right ?)
|> Nonetheless, it's an absolute certainty that it will come.
|> In 30 years from now, ATC will have to be totally automatic.

   Don't be too sure of this. Just because we want to do it does
not mean we are capable of it. What makes you think we can do
anything with computers we want to? I hit the limits of what we
can get the damn things to do every day and its pretty humbling.

|> This raises much more difficulties than the automatization of
|> aircraft.
|> There will still be unpredicted cases, and they will for sure
|> cause losses. But at some point it will become safer to accept
|> these losses than pilot induced losses.

   This is currently not the case. We are seeing lots of perectly
airworthy planes crashing. This stuff should not be tested on
paying passengers. Besides about the only way serious software gets
developed is by incrementally adding to its functionality. Each
upgrade is thouroughly tested etc. and we step-wise refine our way
to a solution. The Airbus approach is a big "jump".

|> I think the main point is that systems are careful 60 seconds
|> per minute and 24 hours a day. And software reliability is
|> increasing, and will keep increasing.

   There is no question that the software will do the same thing
over and over again (right or wrong). The problem is that when it does
fail it crashes a perfectly airworthy plane and kills everybody.
The pilot may fail far more frequently but usually the results
are pretty benign. Again if you have an event with probability
P of occuring but its cost is C the expected cost is PxC. Pilots
have a higer P of failure but a much lower C. Automation has a
lower P of failure but the cost of that failure is an almost
guaranteed maximum C. Also, what makes you think software is
getting more reliable? My perception is quite different.

|> I understand that most people will feel unconfortable if
|> they hear that a computer is holding their lives, but it's
|> an irrational feeling.

   Sorry, I am a very rational person and my dislike of
computers holding my life in their hands comes from many years
of trying to get them to do what I want them to do. I would not
want to trust my own life to my own software and have met very
few programmers (if any) that I would trust my life with.

|>  I also understand that pilots feel
|> self-confident enough, and won't be pleased to see their
|> job evolve in that way. To me a pilot should even regret
|> "fly-by-guts" DC6 :)

   Flying is an art form to most pilots. It is an expression
of ones flair and skill to fly a plane well. It is a very
satisfying experience and it is that experience that drives
young people in droves to try for a career as a pilot.
The managment style of flying required for an Airbus is going
to attract a different kind of person.

|> In another post somebody cited nuclear stations as a system
|> that worried him. But all major nuclear failures were human
|> failures, and there would have been much more accidents
|> without automation.

    I don't have any information on all these accidents but you
are probably right. I am not against automation but I am against
any automated system which does not allow a human to override it.
Hell, instrument landing systems have saved millions of lives I'm
sure and I would not want to get rid of them.

|> One of the consequences of automation is that at some point
|> humans loose their knowledge of "what's going on", and therefore
|> become useless. Even if something goes wrong, they won't have
|> the ability to react correctly.

   Well that's the point where you are in seriously in trouble.
If you are running a nuclear power station and can't figure out
what is going on because the computers always took care of it
you are in for a nasty surprise when the computers go screwy.

   Again, I have two problems with Airbus. The first is the non
overridable envelope and the second is lack of feedback in the
side stick, throttles and rudder pedals. Nothing you have said
has made me feel any better about these issues.