3 times the same flight:

I suggest people take this guy's test with a grain of salt.

Not long ago he predicted the demise of BATC, comparing it to Cyberpunk 2077. Just an hour ago he called the software a "piece of sh.." because he couldn't even get it installed https://www.avsim.com/forums/topic/644583-beyondatc-released-ea/?do=findComment&comment=5098882 and now he posts this, apparently having done extensive tests in less than an hour and BATC is, oddly enough, bad at almost everything.

Smells like an agenda at this point.

 

Edited May 5, 20242 yr by Sethos

Yeah, nice try, buddy. We will see what others will report. I just started and you are the first one who can call me an word not allowed when I am wrong. Until then - I would suggest to report a couple of facts or tests, or participate with substantial information and don´t behave like a fan boi.

And a look or two on BATC discord will also help. 

 

Edited May 5, 20242 yr by techman

How about some details to your test? Like, you know, airports? Aircraft?

30 minutes ago, Fiorentoni said:

How about some details to your test? Like, you know, airports? Aircraft?

Of course.

 

EDDW SOFED1L SOFED N125 EEL DCT LONAM N610 NATEB Y96 AGPED AGPED1E EGPH

Fenix

 

1 hour ago, Sethos said:

Smells like an agenda at this point.

Perhaps, but the bad points he is showing I also encountered during beta testing BATC. Those things won't happen every flight but I myself had a very hard time finishing a flight without encountering a bug.

I'd like to see if the EA is better but I can't because I can't purchase BATC (don't have a credit card). Quite odd to see all those people on Discord using BATC while I as a beta tester can't. Spend ALL my flying time during the beta on BATC and providing feedback and now it's out I can't even buy it. Yeah, I am a bit p i s s e d LOL

Edited May 5, 20242 yr by tup61

22 minutes ago, tup61 said:

Perhaps, but the bad points he is showing I also encountered during beta testing BATC. Those things won't happen every flight but I myself had a very hard time finishing a flight without encountering a bug.

I'd like to see if the EA is better but I can't because I can't purchase BATC (don't have a credit card). Quite odd to see all those people on Discord using BATC while I as a beta tester can't. Spend ALL my flying time during the beta on BATC and providing feedback and now it's out I can't even buy it. Yeah, I am a bit p i s s e d LOL

You were told on Discord to contact someone from the team. They seem to want to work it out and give you access, so just contact them.

2 minutes ago, Fiorentoni said:

You were told on Discord to contact someone from the team. They seem to want to work it out and give you access, so just contact them.

Ah, thanks for the heads up. I'll have a look!

For those people like me that struggle with acronyms, "SI" is "Say Intentions" and BATC is "Beyond ATC" ...

One conclusion can be made very quickly:

All 3 rely on Navdata provided by MSFS / Navigraph and all 3 struggle at some point. SI uses a generic approach for taxiing, BATC and FSHUD seem to "read" the local installation until they come to encrypted and/or incomplete sceneries. The Data quality regarding SID/STARs also seems to be a problem for all 3.

There are no other vendors for Navigational Data right now, correct?

 

 

@techman, has gone to the trouble of comparing the same flight with three different ATC programs. The best way to check if you agree or not is to fly the same route and report your findings. Remember to depart at the same time / date for an accurate comparison.

This is a completely objective test. Things either work or they don't. For those of you who have used Radar Contact v4 that would be a useful comparison if you still have P3D installed.

3 minutes ago, techman said:

One conclusion can be made very quickly:

All 3 rely on Navdata provided by MSFS / Navigraph and all 3 struggle at some point. SI uses a generic approach for taxiing, BATC and FSHUD seem to "read" the local installation until they come to encrypted and/or incomplete sceneries. The Data quality regarding SID/STARs also seems to be a problem for all 3.

There are no other vendors for Navigational Data right now, correct?

 

 

No, they are actually pretty different in that regard: BATC and FSHud use the local data (installed scenery and navigraph navdata) and therefore can give correct taxi instructions most of the time (most payware devs get the taxi data right these days).
SI on the other hand reads no local data, but uses their own cloud navigraph data. They cannot and will never be able to give correct taxi instructions, because those are not available via the common Navigraph API. And, as said, SI cannot read your scenery data.
So in regards to taxi instructions SI has a fundamental disadvantage.

Also your comparison is not quite fair: For example in SI you can just do wrong readbacks, fly all over the place, ignore ATC instructions etc. and it will not enforce them upon you. It will probably just say "Thank you" if you insult it. And SI can not really be "fixed" by coding: Instead of fixing bugs, they have to try to tell ChatGPT what to do, like raising a child. It will always have its moments where it utterly fails and/or hallucinates, even after years of development.
Also ChatGPT has no spatial awareness and can never have. It will regularly fail with vectors and literally give you the opposite ones. Again, this will likely never change, because ChatGPT is ChatGPT and has these flaws by definition.

11 minutes ago, Fiorentoni said:

No, they are actually pretty different in that regard: BATC and FSHud use the local data (installed scenery and navigraph navdata) and therefore can give correct taxi instructions most of the time (most payware devs get the taxi data right these days).
SI on the other hand reads no local data, but uses their own cloud navigraph data. They cannot and will never be able to give correct taxi instructions, because those are not available via the common Navigraph API. And, as said, SI cannot read your scenery data.
So in regards to taxi instructions SI has a fundamental disadvantage.

Also your comparison is not quite fair: For example in SI you can just do wrong readbacks, fly all over the place, ignore ATC instructions etc. and it will not enforce them upon you. It will probably just say "Thank you" if you insult it. And SI can not really be "fixed" by coding: Instead of fixing bugs, they have to try to tell ChatGPT what to do, like raising a child. It will always have its moments where it utterly fails and/or hallucinates, even after years of development.
Also ChatGPT has no spatial awareness and can never have. It will regularly fail with vectors and literally give you the opposite ones. Again, this will likely never change, because ChatGPT is ChatGPT and has these flaws by definition.

Regarding taxi: Thats the same I wrote just in other words. All 3 have a problem right now. None of them is perfect. FSHud fails at a lot of european airports when the line up point is not in line with the holding point (you can see it when AI traffic is trying to line up and goes 180 for 20 meters and turns back - EDDL 23 L or LSZH 14 for example - default scenery). 

Regarding GPTs: I am actually running an AI project myself (not regarding flightsim, but for business) and I can tell you that what you say is not 100 % true. You just have to design it properly. This is a pretty new technology and it developes very very fast. Its not an if/then/else thing, its a different approach to almost everything. 

 

15 minutes ago, techman said:

Regarding GPTs: I am actually running an AI project myself (not regarding flightsim, but for business) and I can tell you that what you say is not 100 % true. You just have to design it properly. This is a pretty new technology and it developes very very fast. Its not an if/then/else thing, its a different approach to almost everything.

I mean this would get far offtopic, but let me tell you that I know LLMs quite well and all I said is true. They have flaws by design that cannot be fixed, they have genetic defects from "birth" on if you will. Hallucinations are one of these flaws, the total lack of understanding of 3D spaces or states is another one of them. These flaws will always exist with ChatGPT.
That's not to say that a LLM is useless; quite the opposite, it's very useful for mondane tasks and efficiency-based tasks especially in businesses (you said it yourself).

Back on topic: Watch closely on Discord how things that Brian (with all due respect, he's a great person) claimes to be "fixed" appear over and over and over again, or how "old bugs" magically reappear every now and then, for three months now. This is because you can not really "fix" things with ChatGPT. You can try to minimize occurrences or simplify things on the backend, but - you said it yourself - since it's not classic programing it can not be fixed by classic programming. Working with ChatGPT is like raising a stupid, but highly efficient child: You can slowly make it do less mistakes if you're working hard with it over time, but it will never end up being "great". And I think it needs to be great to manage something like ATC, which can absolutely not have any mistakes, especially once AI traffic gets on board.
But let's agree to disagree now and see where it is in a year.

47 minutes ago, Fiorentoni said:

So in regards to taxi instructions SI has a fundamental disadvantage.

SayIntentions is looking to use crowdsourcing to collect taxiway data (see their recent stream/dev update). This is similar to how Pilot2ATC generates taxi instruction and not a bad way to do it if an editor of some kind is provided.