In the example above, I didn't ask for an ILS frequency at KLAX. But ChatGPT will give you that information as well if you ask. Try it out. ChatGPT is also very good at figuring out what you want even if you express your request poorly. For example, I mis-typed IFR as IRF but it still understood my intention. I'll bet you $5 that, within one year, ChatGPT (a Microsoft product) will be incorporated into MSFS to tremendously boost the quality of ATC.

EDIT: The web address is https://openai.com/blog/chatgpt

Edited March 11, 20233 yr by David Mills

It even understands the typo 'IRF' 

2 minutes ago, Rimshot said:

It even understands the typo 'IRF' 

My piloting is of equal quality to my typing.

I mean, it's wrong, but it's not entirely off-base. I wonder where it got a lot of this from. MERIT is incredibly heavily used at JFK being, I think, the most used departure gate because of all of the transatlantic traffic, you see OCN V23 LAX in a ton of flight plans because it's on a TEC route. 

Definitely don't go file it on VATSIM because it's a stupid route, but I am super impressed that it gave you something that really could get you from Point A to Point B. I tried to get it to generate KBOS-KJFK, and it gave me something that a newbie would file if they generated a flight plan "VOR-to-VOR" in FSX, but it even reminded me to check the NOTAMs. 😶

I'm impressed by the technology, but we have to remember it's a "dumb" AI, really more of an algorithmic search, and it hallucinates all the time. The Bing Bot, Sydney, proved that if the algorithm doesn't have protections you can probably convince it you killed JFK, and it will say that's why it loves you, and then it will get really insecure about not being the first AI to kill JFK. So, it's really early days lol

 

Here's a low-altitude flight plan from my small, hometown airport to LAX, including ILS landing frequency.

 

4 minutes ago, David Mills said:

 

Here's a low-altitude flight plan from my small, hometown airport to LAX, including ILS landing frequency.

 

And... as I mentioned in the other post... here's your AI hallucination. H270 off of SLI will bring you out to the ocean, nothing out there, and 109.9 is the freq for 25L, not 110.95. But, again, being able to compose something salient like this is impressive. 

The algorithm is too broad to be able to do complex and specific tasks without hallucinating, but it does show promising applications if it's developed specific to tasks. For example, the same model being taught flight planning rules, because you can use the same type of "big language model" applied to flight plan generation and then give it a ton of rules, could become basically a routing wizard. Give it input from real time weather, etc., and rules around desired aircraft performance, and it could be really brilliant. 

There are a ton of applications for this eventually, the technology just isn't mature yet. It's cool though.

Edited March 11, 20233 yr by mspencer

I haven't actually flown any of these ChatGPT-created flight plans. But I'm suspicious that flying from West Virginia to California should take you over Georgia as the plan directs.

Edited March 11, 20233 yr by David Mills

I'll stick with Simbrief. 

9 minutes ago, David Mills said:

I haven't actually flown any of these ChatGPT-created flight plans. But I'm suspicious that flying from West Virginia to California should take you over Georgia as the plan directs.

nor that 400+ NM detour via PXV. But if you trained it on the entire database inside of FlightAware, etc., it would do much better. 

32 minutes ago, jrw4 said:

nor that 400+ NM detour via PXV. But if you trained it on the entire database inside of FlightAware, etc., it would do much better. 

Right - exactly. It's been trained on a general database set. If you trained it on something specific like this, it would probably surpass any dispatcher. It would be imbued with essentially generations of flight plans.

Given the right rules and applications, this technology is already pretty important, even though it's still just an algorithm, and not true AI.

1 hour ago, mspencer said:

but we have to remember it's a "dumb" AI, really more of an algorithmic search,

This is the whole thing.  You're right, it's a simply a search engine.  Our AI isn't really getting smarter; our working definition of "AI" is getting dumber.

This thing isn't actually intelligent.  It cannot learn, it cannot "figure something out" if it was never programed to do the thing etc.  It's not intelligence.  It's just a search engine that is pretty good at regurgitating information in plain language.

Edited March 11, 20233 yr by Stearmandriver

50 minutes ago, Stearmandriver said:

This is the whole thing.  You're right, it's a simply a search engine.  Our AI isn't really getting smarter; our working definition of "AI" is getting dumber.

This thing isn't actually intelligent.  It cannot learn, it cannot "figure something out" if it was never programed to do the thing etc.  It's not intelligence.  It's just a search engine that is pretty good at regurgitating information in plain language.

Yeah, exactly. 5 years ago "AI" was what we call general artificial intelligence - Skynet stuff. It can think, learn, and solve problems.

Now, corporate America has adopted "AI learning" as speak for algorithms that can interpret data sets and work with them... so... an algorithm? I'm in telecommunications and it's particularly bad - at a convention last year there was a company pitching their "AI-driven SD-WAN." Long story short - it was just SD-WAN.

This is a really good chat bot, but there's little more to it.

1 hour ago, Stearmandriver said:

This is the whole thing.  You're right, it's a simply a search engine.  Our AI isn't really getting smarter; our working definition of "AI" is getting dumber.

This thing isn't actually intelligent.  It cannot learn, it cannot "figure something out" if it was never programed to do the thing etc.  It's not intelligence.  It's just a search engine that is pretty good at regurgitating information in plain language.

Lol, somehow you almost sound threatened by it.

2 hours ago, Wildblue said:

Lol, somehow you almost sound threatened by it.

By it?  ChatGPT?  Not at all.  I may be a little threatened by the further dumbing down of the general population though.  I've seen Idiocracy.  😁

Edited March 12, 20233 yr by Stearmandriver

2 hours ago, Wildblue said:

Lol, somehow you almost sound threatened by it.

To be fair, there are a lot of people who are absolutely in danger of being replaced by automation; MidJourney and Dall-E are real threats to anyone in the creative industries; God help Hollywood once it can be used to make videos. Even pilots are walking a fine line; I doubt anyone here will tell me there will be airline pilots in the Flight Deck in March of 2123. I think it's important we as a society have a sit down and really decide how far we'd like automation to help in our day to day, because while we can get these things to do almost anything we need, I'm of the opinion that one of our needs is to be needed. And AI threatens that.