Get instant answers WITHOUT reading the Axe-Fx manual

By that logic, because the forum and support "answers questions", users shouldn't read the manual.


As mentioned in the first post, it's specifically framed as "someone who's already read the manual",
my point was: it won't "answer guestions" with accuracy and intelligence the way some unsuspecting users who have not read the manual think it will.
 
According to who/what?
because its a filter between you and the actual info and
it does not understand what its looking at in the way a human can - but, if it follows current A.."I" trends, It'll try n look like it knows what it's talkin about.

edit: would you trust it to interpret an airline pilot's emerg procedure manual during a mid flight engine failure?

common sense (often lacking these days) - I'll just RTFM thanks - I keep it right handy (read it while waiting in the grocery line).

Entertaining thread - thanks
 
Last edited:
because its a filter between you and the actual info
So is Bing Chat, but in my experience, it does a pretty good job of answering topical questions about stuff I have little familiarity with.

it does not understand what its looking at in the way a human can
Again, the same is true of Bing Chat, but that doesn't stop myriad people from using it as a learning tool.

edit: would you trust it to interpret an airline pilot's emerg procedure manual during an in flight engine failure.
Apples and oranges. That scenario is a matter of life and death. This isn't.
 
Apples and oranges. That's a matter of life and death. This isn't.
Nope - it's the exact same process - the airplane context just drives home the point wrt accuracy - it's not going to be as good / acurate as simply reading the manual - I think we all know that deep down - even those that hate reading manuals lol!

Sorry - have to get at my bed time reading now (even tho my spouse hates it when I read the Axfx Blocks Guide in bed 😳)
 
Last edited:
I'm tired of working on my grad course tonight, so I screwed around with the 2290 issue with ChatGPT. (edit: it's not an issue, it's just the way the delay is! It's part of the magic!)
Here's my conversation:
Screen Shot 2023-05-15 at 10.09.29 PM.png
It's answer:
Screen Shot 2023-05-15 at 10.10.06 PM.png
Obviously, it got it wrong, completely missing the phase issue. So I continued:
Screen Shot 2023-05-15 at 10.10.42 PM.png
So I asked it the exact same original question again, with identical wording. It's response:
Screen Shot 2023-05-15 at 10.11.26 PM.png

Of course, the big question is whether or not it will pass on that corrected advice to another user- I doubt it, but who knows... it seems like it would be extremely problematic to have ChatGPT so easily influenced by user input.
 
Nope - it's the exact same process - the airplane context just drives home the point wrt accuracy - it's not going to be as good / acurate as simply reading the manual.
No, it's completely apples and oranges. I'll give you an example; products used in the medical industry have to meet some of the strictest regulatory standards and undergo rigorous testing for safety, accuracy, and reliability because lives depend on them, but that hardly means that retail medical products for home use aren't safe because they don't meet the same standards. Given that the output of askmypdf.com is limited to the information provided in the uploaded PDF, it's not a big concern, in my opinion.
 
Last edited:
I think it’s “interesting”, but it’s not great. That the text output is relatively useful is largely because @Admin M@ does a good job writing it. The system is spitting out the text, acting as a fancy search engine with a relatively rich natural query language, but it doesn’t really understand the topic.

Many concepts in the manual require the associated graphics; it’s the ages old thing that a picture is worth 10,000 words. That it returns the text from the section is nice but it can’t show us the graphic nor can it tell us what page to look at, at least not using the queries I gave it.

I think an interesting test would be to upload multiple revisions of the FX3 and FM* manuals. And what would it do with the blocks guide which is mostly modeler agnostic, but it will need to ask which version of the firmware and which modeler do we need information about. I suspect the underlying software expects to have final forms of the document, not works in progress.

It can work with non-PDF…
AskYourPdf now supports the following document extensions: '.pdf', '.doc', '.docx', '.txt', '.ppt', '.pptx', '.csv', '.epub', '.rtf'
Heh. PowerPoint docs. Great.

Let’s see if it can survive ingesting the Emacs documentation. :)

It’s not going to be something I use or recommend for a while.
 
No, it's completely apples and oranges. I'll give you an example; products used in the medical industry have to meet some of the strictest regulatory standards and undergo rigorous testing for safety, accuracy, and reliability because lives depend on them, but that hardly means that retail medical products for home use aren't safe because they don't meet the same standards. Given that the output of askmypdf.com is limited to the information provided in the uploaded PDF, it's not a big concern, in my opinion.
Accuracy is accuracy, regardless of underlying subect matter.

Besides - to many folks here, Axfx is the essense of life itself lol! - little margin of error permitted

(guess I should have used another analagy: ie: Hetfield asks newbie tech to reconfig Axfx modifier 5 minutes before show - (If I'm Hetfield, I'm hopin newbie tech read the actual manual or remembers CC training, not going by Q+A thru a hit and miss AI))
 
Last edited:
The system is spitting out the text, acting as a fancy search engine with a relatively rich natural query language, but it doesn’t really understand the topic.
No AI model (including GPT-4) comprehends its output, nor does it need to in order to be useful.

Many concepts in the manual require the associated graphics; it’s the ages old thing that a picture is worth 10,000 words. That it returns the text from the section is nice but it can’t show us the graphic nor can it tell us what page to look at, at least not using the queries I gave it.
I can probably count the number of questions I've seen posted by new users that require graphics on one hand.

It’s not going to be something I use or recommend for a while.
I'll definitely be using it for pinpointing information. It's a lot faster than skimming a document, and it beats standard search functionality hands down, in my opinion.
 
TBH,

I think this is really fascinating material, but I'm wondering if AI will ever progress to visual responses, since many students and end users are visually oriented, and sometimes have trouble reading words (which truthfully can become monotonous after a while).

What would be really awesome is for some enterprising software developer to encapsulate the FM9-Edit software into AI, and use it as a teaching tool, with directed arrows, highlighted areas, close captioned and bold study points text, voice, etc.

In this way, students of FAS might enjoy the learning process more often (no discredit to CC, who has spent hours of categorizing FAS' teaching points). And, it might free up working musician's extra time to pursue their art more often. CC has done great job of putting together a well-thought-out course, but my only concern is that it's kind of a step-by-step learning process, and not able to answer specific questions we might have unless you actually view the entire course and glean the learning points as you view.

A visualized editor that can answer specific questions, provide visuals, a voice, and a transcript of your inquiry would be something many people might appreciate.

Of course, feel free to call me lazy, and that when we were kids we all read books in school, that's how we learned. In fact, we're still doing it now. My feeling is that teaching has come a long way since book education.

What do you think?
 
Accuracy is accuracy, regardless of underlying subect matter.
As I said, the difference is it doesn't require 100% accuracy to be acceptable any more than home medical products require the same regulatory standards as products used in the medical industry. However, as I've pointed out over several posts, the AI limits its output to information provided in the document, so the accuracy is undoubtedly much, much higher than the native version of ChatGPT. For general use, its accuracy is perfectly acceptable, in my opinion, at least in my experience thus far. I mean, Google Maps is generally accurate and most people use it, but would you use it if your life depended on the directions being 100% accurate? Probably not.
 
Last edited:
As I said, the difference is it doesn't require 100% accuracy to be acceptable any more than home medical products require the same regulatory standards as products used in the medical industry. As I've pointed out over several posts, the AI limits its output to information provided in the document, so the accuracy is undoubtedly much, much higher than the native version of ChatGPT. For general use, its accuracy is perfectly acceptable, in my opinion. I mean, Google Maps is generally accurate and most people use it, but would you use it if your life depended on the directions being 100% accurate? Probably not.
when there's the same or less effort (ie just RTM) to get more accuracy - I'll go with more accuracy.

The other nuance we've not touched on is retention and value. In the case if Axfx (and many others I expect), reading the manual gives one a solid foundation of knowledge which can save a 100 questions in the future. I see it here and everywhere all the time - people just want the quick easy answer - but they'll want that 100 times for a 100 different scenarios which reading the manual section carefully once could have avoided as it provides a complete understanding. AI is feeding into that: "give em a quick n easy / maybe not even right but it'll keep em happy for a SM minute" thing.

I think it could be worthwhile if calibrated by an SME somehow (ie Fractal in this case) - but I wouldn't trust it to really think something thru for me without any "credential" at all behind how it's pulling stuff together - specially when the actual manual is sitting right there and written with me as a user in mind.
 
Last edited:
when there's the same or less effort (ie just RTM) to get more accuracy - I'll go with more accuracy.
In my experience, I haven't seen it cite anything from the manual that I'd consider inaccurate as of yet, but feel free to offer an example. In fact, it would've taken me a lot longer to find the same answers simply by skimming or searching.

The other nuance we've not touched on is retention and value. In the case of the Axfx (and many others I expect), reading the manual gives one a solid foundation of knowledge which can save a 100 questions in the future.
Did I or did I not specifically say, "No one's recommending that users not read the manual"?

I see it here and everywhere all the time - people just want the quick easy answer
Yes, because not everybody, me included, is particularly keen on sitting down and reading the entire manual to find the answer to a simple question any more than I'm keen on reading the entire manual to Reaper (DAW) in order to understand how to resize a window. If you have all the time in the world and want to spend it skimming/searching, knock yourself out. I don't.

which reading the manual section carefully once could have avoided.
Perusing material doesn't equate to 100% recall. I mean, the manual is huge, and unless you have a photographic memory, you're likely not going to remember everything. I certainly haven't. In my experience, asking the AI a how-to or explainer question is often a hell of a lot faster than skimming/searching manually.

I think it could be worthwhile if calibrated by an SME somehow (ie Fractal in this case) - but I wouldn't trust it to think something thru for me with any "credential" at all behind how it's pulling stuff together - specially when the actual manual is sitting right there and written with me as a user in mind.
Again, in this instance, most people are perfectly comfortable with high, albeit imperfect, accuracy (e.g. Google Maps, Weather.com), in my opinion. Though again, with respect to AskmyPDF, I haven't seen issues with accuracy thus far.
 
Last edited:
Back
Top Bottom