-4
$\begingroup$

I was asking ChatGPT a few questions about NOTAMs, and decided to ask it what pilots use MATONs (NOTAM backwards) for. It "knew" from context that it had something to do with NOTAMs. It said this:

MATON (Military Aeronautical Telecommunications) is a system used by military pilots to access NOTAMs (Notice to Airmen) information. NOTAMs are issued by the Federal Aviation Administration (FAA) to provide important information about changes or potential hazards in the airspace, such as closed runways, temporary flight restrictions, and communication outages. Military pilots use MATONs to access NOTAMs information relevant to military operations, such as information about temporary flight restrictions (TFRs) or other airspace restrictions that are established for military exercises or training.

MATONs are used by the military to access NOTAMs information in a secure and encrypted format, in order to protect sensitive information about military operations. The NOTAMs information provided through MATONs is used by military pilots to plan their flights and navigate the airspace safely.

It's worth noting that while the FAA is responsible for issuing NOTAMs, the military also has its own NOTAM system, the MIL NOTAM, which is used to disseminate information relevant to military operations to military pilots and air traffic controllers.

Google couldn't find any military aeronautical telecommunications systems called MATON. Also, it would be an unusual acronym for Military Aeronautical TelecommunicatiONs; "MATEL" would make more sense. Did ChatGPT just make up an answer to my question? (It certainly isn't unheard of for ChatGPT to fabricate information, but this would be a particularly egregious case; it appears to have invented an entire military communications system based on only some context from a discussion of civil aviation and an acronym.)

I then asked it "Does MATON really exist?" and it said this:

I apologize, upon further research, it appears that MATON (Military Aeronautical Telecommunications) is not a system used by military pilots to access NOTAMs. My previous response was based on my knowledge cut-off and I apologize for any confusion this may have caused.

The FAA is responsible for issuing NOTAMs, which provide important information about changes or potential hazards in the airspace, such as closed runways, temporary flight restrictions, and communication outages. Pilots and air traffic controllers use NOTAMs to plan flights and navigate the airspace safely.

Military pilots also use NOTAMs to plan their flights, however, military operations may have specific and unique requirements that are not addressed by the civilian NOTAM system. To meet these requirements, the military may use a different system, such as the MIL NOTAM, which is used to disseminate information relevant to military operations to military pilots and air traffic controllers.

Again, I apologize for any confusion and any inaccuracies in my previous response.

Clearly one of its responses is wrong.

$\endgroup$
7
  • 6
    $\begingroup$ ChatGPT doesn't "make up" or "invent" anything. That is giving it far too much credit and demonstrates a very dangerous and unfortunately widespread misunderstanding about what ChatGPT is (and isn't). It strings together words that it has seen appear next to each other. That's all it does. Have you ever tried typing a message on your phone by only ever selecting the first suggested word on the predictive keyboard? Have you noticed that what comes out typically makes no sense but still sounds at least remotely like a message you might have written? That's how ChatGPT works, except with a … $\endgroup$ Commented Jan 12, 2023 at 16:55
  • $\begingroup$ … way bigger training set. But the principle is the same: GPT-3 gets fed pretty much everything ever written in the history of mankind, breaks it up into pairs of tokens, for each start token, it analyzes what's the most likely following token, and stores that. Then, when you ask it to produce a text, it will just start with a token, predict the most likely next one, then the next, then the next, and so on. $\endgroup$ Commented Jan 12, 2023 at 16:57
  • 1
    $\begingroup$ I'd call that "working as intended". It is designed to pass the Turing test, i.e. to fool people, not to be correct, precise, or accurate. In fact, that would defeat the point, since people aren't correct, precise, or accurate, so you would be able to tell it's an AI by its correctness, precision, and accuracy. $\endgroup$ Commented Jan 12, 2023 at 17:06
  • 2
    $\begingroup$ As my oldest kid the software developer says, AI is a hype term for just really really fancy software. Still garbage in/garbage out. BTW they just changed the "Airmen" part of NOTAM to "Air Mission" to achieve certain social language objectives, which now appears to be the FAA's priority. $\endgroup$
    – John K
    Commented Jan 12, 2023 at 17:19
  • 1
    $\begingroup$ @JohnK having developed a program that is on the simple end of what many people would consider "AI" (far simpler than ChatGPT), I definitely agree that it's not the big deal people think it is. "Machine learning" is a more meaningful term IMO. $\endgroup$
    – Someone
    Commented Jan 12, 2023 at 17:21

1 Answer 1

9
$\begingroup$

The OP has the necessary info to answer the question:

Google couldn't find any military aeronautical telecommunications systems called MATON.

No, MATON isn't a thing.

As Stack Exchange notes in the Why posting GPT and ChatGPT generated answers is not currently acceptable page,

Stack Overflow is a community built upon trust. The community trusts that users are submitting answers that reflect what they actually know to be accurate and that they and their peers have the knowledge and skill set to verify and validate those answers. ... Currently, contributions generated by GPT most often do not meet these standards and therefore are not contributing to a trustworthy environment.

GPT isn't a person, and it is written to provide a "looks good" answer, not necessarily an "is correct" answer. To say that is has zero integrity and is willing to lie convincingly isn't wrong, but that's as much because integrity & a concept of doing the right thing is only applicable to humans, while GPT is a program. It's doing what it's programmed to do.

This experiment demonstrates vividly why GPT is not to be trusted, and why SE is right to ban it.

Note that the policy of suspending users who post GPT content without attribution (i.e. the OP is fine here) is a zero tolerance & no warnings policy; a few accounts have already received significant suspensions for posting GPT-generated content on Aviation.SE.

$\endgroup$
2
  • 4
    $\begingroup$ Note that the original ban is only about questions and answers generated by GPT. However, there is also currently a discussion about the scenario we have here: asking about something generated by GPT. For this case, the consensus also seems to be to ban it completely, because a) the answer is always the same: Is it true? "No!" Why did GPT say this? "Because it just randomly strings together convincingly-sounding gibberish". and b) the fact that you can, with zero effort, generate an infinite flood of such questions, each of which require non-zero effort to answer, is essentially a DoS attack $\endgroup$ Commented Jan 12, 2023 at 17:11
  • 2
    $\begingroup$ Such questions also almost always qualify for downvotes as "does not show any research effort" since ChatGPT's limitations are clearly listed on its introductory blog post under the heading "Limitations", one of which is "plausible-sounding but incorrect or nonsensical answers". $\endgroup$ Commented Jan 12, 2023 at 17:20

Not the answer you're looking for? Browse other questions tagged .