[Sutripta]

So should not be used interchangeably.
Should have started with

Originally Posted by Jeroen

They are two very different thing all together.

AI is often used as sort of the next level of automation. Technically that is incorrect. AI is a capability (provided by several components, e.g. computers, software, storage etc) . What is true is that the combination of automation and AI gives us unprecedented capabilities, far beyond what we can achieve with automation without AI.

Alphazero/ Deepmind learned chess by playing against itself. What would constitute its dataset?

Regards
Sutripta

[Jeroen]

Originally Posted by Sutripta

Alphazero/ Deepmind learned chess by playing against itself. What would constitute its dataset?

I know very little about these chess-computers. Apparently this one takes chess computers to a whole new level.

Past chess computers relied more on raw computing power (just crunching almost endless move combination to find the best) and a set of rules/guide lines given / mirrored from real chess players.

From the little I know Alphazero develops its own set of guidelines based on million and millions of simulated chess plays. So it appears it was given one basic data set, which would constitute of the normal basic chess rules and movements. E.g. knights move diagonally, pawn can move left/right/up/down etc.

Next it runs millions of simulations and has the ability to learn along the way and come up with what apparently is a very novel and surprising way of playing a game of chess.

Not sure how it does it in any more detail though.

Jeroen

[srini1785]

Originally Posted by Sutripta

Would someone explain to me the difference between automation and AI. (In ones own words please. Not links to learner articles which would just further confuse us laymen).

I have a distinct feeling that you are not a layman . Reason See below :

Originally Posted by Sutripta

Alphazero/ Deepmind learned chess by playing against itself. What would constitute its dataset?

My words :

Automation : Any piece of machinery / system that simply follows a pre programmed sequence. No uncertainties / non programmed conditions involved (as "human intervention" in Mr. Jeroen's post). Eq a Paint shop Robot.

AI : Same as above , except ability to deal with uncertainties.

[Sutripta]

^^^ (Jeroens)
One can't just increase depth of search. The combinatorial explosion quickly overwhelms even supercomputers. Some sort of branch and bound has to be used. Which lines of play to discard becomes important. Deep Blue did this deep search and evaluate better than others.
Strictly even Deep Blue should not be called AI.
AlphaZero works differently I think.

Anyway MCAS is simple if seen as a control loop. It's problems were elsewhere - in the design assumption and constraints of its human designers. Implementing a reliable MCAS does not need AI. (For the time being let's not get into 'a properly designed aircraft should not need MCAS'. )

As said before, AI can be a great cockpit resource. Esp in high pressure critical situations. In deciding what nformation to present to the pilots. With commentary. An expert system if you will.

Initially put AI in the simulator, and run the flight software and pilots against it.

Regards
Sutripta

[Sutripta]

Originally Posted by srini1785

I have a distinct feeling that you are not a layman .

One of the great advantages of being a layman is one can ask really stupid questions. And in a forum of experts it is stressful being another 'expert'. Layman suits me just fine.

Strangely I know some actual experts in their field on this forum. Who comment and opine on various subjects, but are absolutely mum when it comes to their actual (bill paying dayjob) field.

And no, AI is not my field.

Regards
Sutripta

[Thad E Ginathom]

Not only a layman, but out of touch with modern computing developments for nearly twenty years.

My guess is that a chess-playing computer could provide its own data by playing many games with itself and finding out what won and what lost. Just as with instructing another human, someone has to tell it what the objective is and what the rules are.

Some time ago I read about computer management of silicon-chip manufacturing facilities. Our membership is so diverse, I bet we have a member or two that knows about this! Every aspect of the machines is monitored, and maintenance is usually predicted before the fault develops. I guess there must be some dataset as a starter; otherwise, the system is learning symptoms and outcomes as it goes along. I don't think my brain is big enough to understand this!

[Jeroen]

Originally Posted by Sutripta

As said before, AI can be a great cockpit resource. Esp in high pressure critical situations. In deciding what nformation to present to the pilots. With commentary. An expert system if you will.

Initially put AI in the simulator, and run the flight software and pilots against it.

It will be interesting to see where and in what roles AI will be introduced in the cockpit. I am not so sure AI is the best way forward when it comes to dealing with real emergencies. High pressure critical situations are not necessarily emergencies. Pilots work loads tend to be very high during take off, climb out and approach and landing. Lots of interaction between the pilots, ATC, lots of systems to be set, adjusted, checked, monitored, etc.

When pilots gets behind the curve as they say, that is where an incident and or accident is in the making.

Automating the ATC instructions so we don’t have to use crappy VHF radio technology would be good. AI should be extremely good at juggling, spacing and separating traffic in busy airspace. Much better than your typical air traffic controller. Why not have it instruct the plane directly, without pilot intervention?

I have this horrible feeling we might see AI introduced primarily as a cost saving. Do away with a pilot and you save yourself a lot of cost.

Originally Posted by Sutripta

Strangely I know some actual experts in their field on this forum. Who comment and opine on various subjects, but are absolutely mum when it comes to their actual (bill paying dayjob) field.

Maybe they prefer their hobbies and other interests to the day job?

[Sutripta]

^^^
Or maybe keeping the dayjob requires obeying the first two rules of the fight club!

[Jeroen]

This 737 MAX is not going to take to the skies any time soon

https://www.forbes.com/sites/marisag.../#722f945a8888

The FAA has announced it will need to certify each 737 Max individually. An unusual approach, but it will continue to do so until:

Citing the large number of 737 MAX planes that will need vetting that have piled up in storage as Boeing has continued production despite the global grounding of the plane, the FAA said it will “retain authority to issue airworthiness certificates and export certificates of airworthiness for all 737 MAX airplanes” for an indefinite period, until the Administration is satisfied that Boeing has “fully functional quality control and verification processes.”

Jeroen

[Sutripta]

^^^
Compromise of some sort.
Previously I think FAA wanted to certify the planes but not the the pilots till they had extensive and rigorous retraining.
Australia's ATSB was apparently proving recalcitrant. ATSB has a stellar record, and having no real domestic aircraft manufacturer more immune to pressure.

Regards
Sutripta

[Jeroen]

Huge back log of Max’s they will need to check:

https://www.youtube.com/watch?v=hwejf6Yb3tA

[itwasntme]

Nice one, had seen a similar one earlier when they still had the operating car park on the other side.

They'll have to repaint that Jet Airways one :-)

[Sutripta]

Originally Posted by Jeroen

The FAA has announced it will need to certify each 737 Max individually. An unusual approach, but it will continue to do so until:

What about the Maxes already with operators, but now grounded?

Regards
Sutripta

[Jeroen]

Originally Posted by Sutripta

What about the Maxes already with operators, but now grounded?

As far as I know it is always the local (country or for instance EU) regulator that issues the relevant airworthiness certificate. So it is up to them. With the FAA taking this position, it will be interesting to see what position the other regulators take.

I do not think they can take a more accommodating stance than the FAA. If the FAA does not trust Boeing why should any other regulator trust Boeing? And let's face it, all the regulators also distrust the FAA.

It is going to be hugely important for Boeing how this will play out. The majority of Maxes is with operators outside the USA!

Irrespective of how it will play out exactly it is going to take a long time to inspect all these planes. It is not as if these regulators have an army of inspectors that are capable of doing this. So there is going to be serious resource constraint.

Jeroen

[itwasntme]

Originally Posted by Jeroen

Irrespective of how it will play out exactly it is going to take a long time to inspect all these planes. Jeroen

@Jeroen, confused layman here:

If Boeing rectifies the MCAS firmware to the approved version including inputs from both AoA sensors, installs the optional 'AoA mismatch' cockpit indicator light etc, recalibrates the AoA sensors either themselves (or through approved third parties) and checks all relevant hardware, WHY has the FAA still reserved the right to inspect and certify each aircraft?

I mean, what else is bespoke about individual MAX 8s that could differ from airplane to airplane?