Cant believe this

[nickhod 425]

6 hours ago, jabloomf1230 said:

Here's Pete Dowson's latest thoughts on FSUIPC and MSFS:

It's a shame so much software relies on an outdated and mostly undocumented protocol to be honest.

I'm sure we wont escape its use with MSFS. If Pete himself doesn't port it to MSFS someone will create something that does the same job, as they did with X-Plane and XUIPC.

 

[eslader 1,779]

I would love to see a realistic ATC that does not require me to interface with real humans, if for no other reason than that I sim in the same room as my wife's computer and I don't want her to kill me if I won't shut up with the airplane talk.

 

But, I mean, my Google Home is still a digital moron. It's gotten better, but there's a lot of phrasing that it simply doesn't understand, and that's with a big team of people from a more innovative company than Microsoft who do nothing but teach it to communicate all day.

I still have to say "turn off the living room lights and turn off the kitchen lights." If I say "turn off the living room lights and the kitchen lights" it turns off the living room lights and then gets confused and apologizes.

If it can't figure out a very short string of simple conversational on/off commands, there's no way a functioning AI ATC that's actually realistic is possible yet, from anyone, at any price.

 

[jabloomf1230 2,445]

3 hours ago, eslader said:

If it can't figure out a very short string of simple conversational on/off commands, there's no way a functioning AI ATC that's actually realistic is possible yet, from anyone, at any price.

I disagree, the speech recognition in Windows 10 is top notch. VOXATC and Pilot2ATC use it. The trick is to train the SR engine using ATC phrases and use a really good microphone. What's really odd about ASOBO not using SR is that the online MS SR is just as good as the local flavor. My gut feeling on this is that like virtual reality, the lack of speech recognition was a performance consideration.

[ARCO 245]

To add to jabloomf1230 comment above on Windows 10 speech recognition (SR), a number of flight simulator add-ons have been using SR extensively and successfully for years now, I too used VoxATC, PF3, Radar Vectors and Pro ATC,  but dropped all of them because of the voices, beside those already mentioned, MultiCrewExperience (MCE) uses SR exclusively and has been on the market for several years (FS9 and FSX then P3D, and one can bet FS2020 will be the next) and even started with Windows XP which was less sophisticated. I have been using MCE on all the above simulators for interactions with Ground Crew (GSX Pushback and boarding staff), Cabin Crew, CoPilot (Checklists, and crew coordination during the flight) and ATC - stock or add-ons - (from flight clearance to full stop), that eliminates the use of the keyboard. The SR training is limited to a few simple short sessions and will remain valid for as long as you dont change your equipment which will include a headset with a good microphone. Ambiant noises - if picked up by the microphone - will degrade the quality of SR, this is probably why your Google Home does not perform up to your expectations.  

Regarding performance, depending on the way the add-on is designed, it won't affect the flight simulator performance significantly - if at all.

Edited June 13, 2020 by Jean-Claude

[Dominique_K 10,133]

5 hours ago, jabloomf1230 said:

. My gut feeling on this is that like virtual reality, the lack of speech recognition was a performance consideration.

Do not underestimate that they build the darn thing up from scratch. They had to make priorities. If I understand well, the ATC was not one, early on. They decided to add it because of the outcry of the community. They obviously took over the old FSX ATC and improved it a bit (something we still await from the P3D team btw). SR that I very much look forward to, is kind of the cherry on the icing on the cake. 

[RALF9636 1,920]

2 hours ago, Jean-Claude said:

... beside those already mentioned, MultiCrewExperience (MCE) uses SR exclusively and has been on the market for several years (FS9 and FSX then P3D, and one can bet FS2020 will be the next) and even started with Windows XP which was less sophisticated. I have been using MCE on all the above simulators for interactions with Ground Crew (GSX Pushback and boarding staff), Cabin Crew, CoPilot (Checklists, and crew coordination during the flight) and ATC - stock or add-ons - (from flight clearance to full stop), that eliminates the use of the keyboard. 

Another happy MCE-user here.

I agree the voice recognition with MCE is close to perfect.

If MCE will be made compatible with MSFS (and I am confident it will), I don't see a need for voice recognition built into MSFS itself. MCE covers voice interaction with crews and ATC - and offers so much more I don't want to miss in MSFS.

[jabloomf1230 2,445]

5 hours ago, Dominique_K said:

Do not underestimate that they build the darn thing up from scratch

That's why they didn't rewrite the ATC and AI aircraft logic from scratch. They probably believe that the demand isn't there and the performance hit would be too great. It's not that including realistic ATC has an inherent performance penalty. It's that writing something so complex from the ground up is fraught with issues. And LM and LR haven't done anything differently with the default ATC in their sims. I guess that it's like Catch 22. If everyone else is doing it, I'd be a fool to do otherwise.

[charliearon 3,017]

On 5/30/2020 at 8:15 AM, DaWu said:

 Its AIDS? 

Acquired Idiot Diversion Syndrome??  Let's keep the discussion on track  Try to avoid  using medical terms to embellish your posts!

[B777ER 2,484]

On 5/29/2020 at 9:50 PM, Cyrex1984 said:

Just fly on pilotedge or vatsim. Problem solved.

Have we heard anything from MS regarding things like Vatsim?

[eslader 1,779]

On 6/12/2020 at 7:57 PM, jabloomf1230 said:

the speech recognition in Windows 10 is top notch.

The speech recognition on Google Home is pretty good too. The problem comes with Google Home knowing what to do in response to that speech.

 

The ATC might well know that I said the words "request left 15 for weather," but that doesn't mean it will know how to react appropriately to them.

 

[jabloomf1230 2,445]

I think we're talking about two different components of speech recognition. The way that Google Home and Alexa work is that they have an extremely limited vocabulary. If you deviate slightly from what the assitant expects it won't respond correctly. Both the Google and Amazon servers create training profiles, so over time the SR improves for a specific user's voice.

The ATC apps work in a somewhat similar manner. For example, VOXATC has two modes.The standard mode allows no deviation from accepted ATC phraseology. The expert mode allows a bit of flexibility in what spoken terminology will work. The Windows 10 speech recognition is only responsible for converting your analog spoken words to digital text equivalents. It is still up to the ATC app to match those words to the limited vocabulary. Both Pilot2ATC and VOXATC have training utilities. Prior to your flight, you enter the name of your flightplan file and the utility creates a specialized training session just for that flightplan. I have found that the best approach is to use a headset to reduce ambient noise and sim produced sounds, combined with a high quality desktop mic.

Personally, anything else is basically back in the dark ages with FSX. 

Edited June 14, 2020 by jabloomf1230

[Wobbie 938]

And we know, as we have been told, that some aspects/parts of FSX have been used in MSFS. So, not such a surprise at all.

[eslader 1,779]

On 6/13/2020 at 11:20 PM, jabloomf1230 said:

I have found that the best approach is to use a headset to reduce ambient noise and sim produced sounds, combined with a high quality desktop mic.

And even then, it's limited. I use Pilot2ATC. I generally like it, but sometimes it's frustrating. Real pilots do not always say "descend and maintain one niner thousand."  Sometimes they say "down to one-nine-thousand" or even "down to nineteen thousand."  The response to a traffic advisory is often "we're lookin'."  These commonly used phrases do not work in P2ATC even though Windows knows what I said.  And that's not getting into the sometimes bizarre actual ATC decisionmaking that P2ATC does, like occasionally thinking I'm gonna descend from FL400 to the initial approach fix in 15 miles.

So, I think we're talking about the same component of speech recognition; Speech recognition has gotten pretty good at recognizing the dictionary words that you say. But it still sucks at being able to figure out what to do with those words unless you luck out and someone specifically programmed a response to the specific phraseology you use.

 

In short, like I said before, with current speech-recognition-and-response technology, you will not have a 100% realistic AI ATC in any flight sim, for any price because it is absolutely impossible for the programmers to anticipate every single possible sentence a sim pilot might say and program in an appropriate response to it.

That level of AI ATC isn't going to happen until true AIs happen, and I don't mean Watson which has to be specially set up for whatever demo it's doing this week - I mean computers on the level of those in Star Trek.

 

 

Edited June 15, 2020 by eslader

[michal 44]

On 6/15/2020 at 10:59 AM, eslader said:

Real pilots do not always say "descend and maintain one niner thousand."  Sometimes they say "down to one-nine-thousand" or even "down to nineteen thousand." 

 

Other common responses are: "At one seven thousand descending one three thousand" or  "one seven thousand descending one three thousand" or "leaving one seven thousand for one three thousand". I was often taught to avoid using "maintain" as redundant. Any "realistic" ATC is limited nowadays to a live human on the other end. In full agreement with your assessment.

Edited June 19, 2020 by michal

[Will Fly For Cheese 1,710]

9 minutes ago, michal said:

Other common responses are: "At one seven thousand descending one three thousand" or  "one seven thousand descending one three thousand" or "leaving one seven thousand for one three thousand". I was often taught to avoid using "maintain" as redundant. Any "realistic" ATC is limited nowadays to a live human on the other end. In full agreement with your assessment.

And if it gets too complicated because there's a risk of spilling a coffee in the cockpit they'll generally say; "Unable".

Never, ever underestimate the power of saying "Unable" to ATC - it throws them into a massive fit. It blows their plan for your APP away and makes them have to think. Which is good.

Lots of fun. Especially if you're flying something Heavy. Like a 777.

Edited June 19, 2020 by Will Fly For Cheese