View Single Post
Old 09-23-2019, 09:50 AM   #9  
Line Holder
501D22G's Avatar
Joined APC: Jun 2011
Posts: 67

Originally Posted by rickair7777 View Post
From a technical, economic, and political perspective it can't be a minute less than 50 years...

This is true, and many folks don't get the implication of that....

The development costs for single pilot are pretty much the same as no pilot. But you're only saving the cost of the FO (and not even all of that if you have backup pilots on the ground) since there's no conceivable scenario where you go directly from two pilots to no pilots.

The hard part is not the automation per se, although there is significant technical uncertainty in the "last mile", ie getting all the way to 10(-9).

The hard part is the economics of developing and implementing, plus the certification hurdles. Google "Deterministic AI" to learn more about the later.

1. Nobody knows how to implement a true generalized AI. And if you did, there are serious doubts among the top minds as to whether you really should.
2. Non-deterministic systems which attempt to model that cannot be certified for complex safety-critical applications (you never know exactly what it will really do in a given scenario, and the response could change from day to day).
3. Deterministic AI is just software. If it's programmed wrong (by a human or another AI) it's just like MCAS.

Equally hard is making the business case to spend the development money in the face of uncertainty in areas of technical, regulatory, and public acceptance. No manager who starts writing checks for this will ever see a dime of short or mid-term gain, the timeline is too long. Technical movement in that direction will happen, but for other reasons. Automation of weapon systems for example, but all military planning and research revolves around teaming between manned and unmanned systems. There's no technical potential for totally autonomous weapon systems (aka terminator), too risky.

It will really need to happen with cars first. Cars are not easier than airplanes, they can always pull over to the side of the road if something goes wrong. Although a car's "taxi environment" is less predictable a plane will still need to be able to avoid errant ground vehicles, illegal aliens, and crazy folks running across the airport surface. Less common than for automobiles but liability cannot just be wished away. No company can afford to sell or employ something which cannot be insured, nor can a regulator certify it.

An autonomous airline will actually have to be demonstrably BETTER than manned airliners, in order to justify the cost and sway regulators and the public.

Also you'd need to completely re-engineer ground handling and ATC systems $$$$$$$$$$$$$$$. Who's going to pay for that? The fed? Why, to eliminate 100,000 good paying union jobs? That's not how the world really works.
You said it better than I could. I was typing a reply, but decided to see if there were any other responses first.
501D22G is offline