Topic by DonSmith

  • Avatar
  • DonSmith
  • Member
  • Offline
  • Registered: 24-09-2008
  • Posts: 414

Topic: I've been trying out ChatGPT by Open AI

I've been trying out ChatGPT by Open AI and started asking about physical modeling sound synthesis. It's so intuitive I ended up asking it:

Question:
What is Fourier analysis or wavelet analysis? What would I be looking at if I were using them?

Fourier analysis and wavelet analysis are both techniques for analyzing the spectral content of a signal, which is the distribution of energy in a signal as a function of frequency.

ANSWER:
Fourier analysis involves decomposing a signal into a set of sine waves of different frequencies using the Fourier transform. This allows you to identify the frequencies and amplitudes that make up the signal, and can be used to create a spectral representation of the signal that shows how the energy is distributed across different frequencies.

Wavelet analysis is a similar technique that involves analyzing a signal using wavelets, which are small wave-like functions that are localized in both time and frequency. Wavelet analysis can be used to identify the time-frequency structure of a signal, which can be useful for analyzing signals with non-stationary properties, such as musical sounds.

Both Fourier analysis and wavelet analysis can be used to extract information about the spectral content of a signal, which can be used to create mathematical models of the sound source for use in physical modeling sound synthesis. For example, you could use Fourier analysis to identify the frequencies and amplitudes of a recorded sound, and then use that information to create a mathematical model of the physical properties and behavior of the sound source.

Overall, Fourier analysis and wavelet analysis are important tools in digital signal processing and are widely used in fields such as audio engineering, music production, and scientific research.

Reply by xinaes

  • Avatar
  • xinaes
  • Member
  • Offline
  • Registered: 12-09-2013
  • Posts: 40

Re: I've been trying out ChatGPT by Open AI

I just spent ages writing quite a lengthy reply to this and the forum software barfed.

It is interesting, and I think it already shows signs of being more useful than Google for a lot of things - I think over the next small number of years, it will develop considerably and deeply permeate our lives.

It can also be confidently wrong a lot of the time.

I had a really interesting conversation with it about Messiaen's "modes of limited transposition". Apparently they've been used a fair bit by minimalist composers. It gave me quite a detailed explanation of how this applied to Steve Reich's Tehillim, and I'd love to be corrected on this, but when I tried to find any supporting sources for this (or listen / look at the score and make sense of it that way), I wasn't convinced. Indeed, this particular piece seems very much more traditionally functional harmony.

It seems like a logical thing for minimalists to make use of these scales, and ChatGPT was very eloquent about describing in terms of theory and aesthetics how this has been done and what the results sound like. I'd like to know more about that, because my attempts at Googling haven't got me anywhere.

When it came to spelling out the notes in one of these scales, it also got it totally wrong (with a set of intervals that don't even add up to an octave) and fairly adamantly insisted that there may be other interpretations, but both were valid. Not really the case here, sorry.

It can be a bit like talking to a politician sometimes.

Still, I think we'll see quite rapid improvement soon and it is already pretty mind-blowing.

Last edited by xinaes (17-02-2023 14:16)

Reply by hesitz

  • Avatar
  • hesitz
  • Member
  • Offline
  • Registered: 07-06-2019
  • Posts: 128

Re: I've been trying out ChatGPT by Open AI

You can search 'chatgpt music' on youtube for some interesting vids.  Below is an interesting/amusing one, where Chat GPT ends up not being reliable, but there are plenty of instances where it gives impressive results.  Often, though, it needs to be used by someone who has some knowledge of what they're doing.

https://www.youtube.com/watch?v=dcjhUI3-n6I

Last edited by hesitz (17-02-2023 18:32)

Reply by Antonio M

  • Avatar
  • Antonio M
  • Member
  • Offline
  • Registered: 23-09-2020
  • Posts: 94

Re: I've been trying out ChatGPT by Open AI

xinaes wrote:

[...]I think over the next small number of years, it will develop considerably and deeply permeate our lives.[...]

Just a couple a moths back someone was suggesting Modartt to make use of AI for Pianoteq "or else..." (someone else would do it).

I dislike string pianos (I think they bit me when I was a kid) as much as I dislike the "old world".
Welcome to the new technologies.

hesitz wrote:

You can search 'chatgpt music' on youtube for some interesting vids.  Below is an interesting/amusing one, where Chat GPT ends up not being reliable, but there are plenty of instances where it gives impressive results.[...]

I remember several year ago (many year ago if you are young) that some Chopin "enthusiast" was upset he could not distinguish Chopin's own composition from some AI generated ones... (I though "Chopin enthusiasts" would know every thing piece  by him)

Reply by levinite

  • Avatar
  • levinite
  • Member
  • Offline
  • Registered: 07-08-2020
  • Posts: 133

Re: I've been trying out ChatGPT by Open AI

So, sometime it gives the wrong answers. How do you know when you can trust it and when you can't?

Reply by Antonio M

  • Avatar
  • Antonio M
  • Member
  • Offline
  • Registered: 23-09-2020
  • Posts: 94

Re: I've been trying out ChatGPT by Open AI

levinite wrote:

So, sometime it gives the wrong answers. How do you know when you can trust it and when you can't?

well, you can't trust anyone, anything, anywhere, anytime, anymore. so...?



(...sorry...)

Reply by wonner

  • Avatar
  • wonner
  • Member
  • Offline
  • Registered: 27-09-2019
  • Posts: 25

Re: I've been trying out ChatGPT by Open AI

levinite wrote:

So, sometime it gives the wrong answers. How do you know when you can trust it and when you can't?

A growing number of organizations are providing plugins for ChatGPT that allows you to use their knowledge bases. For example, Wolfram (the company that makes Mathematica) has a plugin. If you set ChatGPT to use the Wolfram plugin and ask mathematical questions, you are likely to get solid results.

I have experimented with ChatGPT a lot in the past few months and overall been extremely impressed with the results on a wide range of topics, even when not using plugins.

W

Reply by wonner

  • Avatar
  • wonner
  • Member
  • Offline
  • Registered: 27-09-2019
  • Posts: 25

Re: I've been trying out ChatGPT by Open AI

Here is an example of a question and answer from ChatGPT.

Question: What is the common eight bar jazz chord progression?

Answer:

The most common eight-bar jazz chord progression is the "Rhythm Changes," which is based on George Gershwin's song "I Got Rhythm." However, it's worth noting that "Rhythm Changes" is typically a 32-bar AABA form, with each section being 8 bars. The A section of this progression is the one that's most frequently associated with "Rhythm Changes."

Here's a basic version of the A section of "Rhythm Changes" in the key of Bb:

| Bb6   | G7    | Cm7   | F7    |
| Bb6   | Bb7   | Eb6   | Ebm6  |

The B section, or the bridge, is typically a cycle of dominant 7th chords:

| D7    | G7    | C7    | F7    |

Then the A section is repeated twice more.

Many jazz musicians have written tunes based on these changes, and it's a common progression for improvisation. If you're studying jazz, becoming familiar with "Rhythm Changes" is essential!

Reply by Pianophile

  • Avatar
  • Pianophile
  • Member
  • Offline
  • From: The Netherlands
  • Registered: 07-06-2007
  • Posts: 229

Re: I've been trying out ChatGPT by Open AI

wonner wrote:

Here's a basic version of the A section of "Rhythm Changes" in the key of Bb:

| Bb6   | G7    | Cm7   | F7    |
| Bb6   | Bb7   | Eb6   | Ebm6  |

The B section, or the bridge, is typically a cycle of dominant 7th chords:

| D7    | G7    | C7    | F7    |

Then the A section is repeated twice more.

That's a pretty garbled version of the 'Rhythm changes' chord progression as a whole. You wouldn't get very far with this in an actual band context. Chat GPT often gets a lot ot stuff right but not without also getting a couple of crucial things wrong, or omitting them altogether.

Reply by sharpnine

  • Avatar
  • sharpnine
  • Member
  • Offline
  • Registered: 01-11-2020
  • Posts: 67

Re: I've been trying out ChatGPT by Open AI

Pianophile wrote:
wonner wrote:

Here's a basic version of the A section of "Rhythm Changes" in the key of Bb:

| Bb6   | G7    | Cm7   | F7    |
| Bb6   | Bb7   | Eb6   | Ebm6  |

The B section, or the bridge, is typically a cycle of dominant 7th chords:

| D7    | G7    | C7    | F7    |

Then the A section is repeated twice more.

That's a pretty garbled version of the 'Rhythm changes' chord progression as a whole. You wouldn't get very far with this in an actual band context. Chat GPT often gets a lot of stuff right but not without also getting a couple of crucial things wrong, or omitting them altogether.

I agree with Pianophile.
Yes, the B section is correct. But the A section is way off. Just look at the wikipedia page on Rhythm Changes. For example, it shows bars 7 and 8 are actually the chords of bar 6, and it omits the chords from bars 7-8. Also the comment about the structure "then the A section is repeated twice more" is obviously incorrect for a piece (correctly) noted as AABA. Overall the suggestions are a mess and make no sense.

Last edited by sharpnine (05-09-2023 19:04)

Reply by sharpnine

  • Avatar
  • sharpnine
  • Member
  • Offline
  • Registered: 01-11-2020
  • Posts: 67

Re: I've been trying out ChatGPT by Open AI

sharpnine wrote:
Pianophile wrote:
wonner wrote:

Here's a basic version of the A section of "Rhythm Changes" in the key of Bb:

| Bb6   | G7    | Cm7   | F7    |
| Bb6   | Bb7   | Eb6   | Ebm6  |

The B section, or the bridge, is typically a cycle of dominant 7th chords:

| D7    | G7    | C7    | F7    |

Then the A section is repeated twice more.

That's a pretty garbled version of the 'Rhythm changes' chord progression as a whole. You wouldn't get very far with this in an actual band context. Chat GPT often gets a lot of stuff right but not without also getting a couple of crucial things wrong, or omitting them altogether.

I agree with Pianophile.
Yes, the B section is correct. But the A section is way off. Just look at the wikipedia page on Rhythm Changes. For example, it shows bars 7 and 8 are actually the chords of bar 6, and it omits the chords from bars 7-8. Also the comment about the structure "then the A section is repeated twice more" is obviously incorrect for a piece (correctly) noted as AABA. Overall the suggestions are a mess and make no sense.

Not to mention that the original question asked about the most common 8-bar progression, and the answer is a 32-bar progression! I am dubious that the initial 8-bars  is the most common 8-bar progression--it is pretty obviously rhythm changes, whereas there are plenty of generic 8-bar progressions which are likely more common (say an 8-bar blues, or the first 8 bars of a 12-bar blues).

Reply by Kawasertan

  • Avatar
  • Kawasertan
  • Member
  • Offline
  • Registered: 07-12-2023
  • Posts: 1

Re: I've been trying out ChatGPT by Open AI

The exploration of "Improving Computational Essays: Interactive Feedback and Validation" was not just enlightening but also resonated deeply with my passion for writing. The intricate blend of interactive feedback and validation mechanisms showcased in the essay provides a dynamic approach to enhance the learning experience. The utilization of computational elements adds a new layer of engagement, making the educational journey more immersive. It's heartening to witness innovative strides in essay development that transcend traditional boundaries. As a student who values creativity and progress, this insightful piece reaffirms my belief in the transformative power of academic writing. For more inspiration, I often explore diverse essay styles, and platforms like https://essaypro.com/essays-for-sale offer a rich array of perspectives and ideas.

Last edited by Kawasertan (08-12-2023 20:36)

Reply by Joseph Merrill

  • Avatar
  • Joseph Merrill
  • Member
  • Offline
  • Registered: 24-10-2023
  • Posts: 36

Re: I've been trying out ChatGPT by Open AI

I never got a better answer from a person than the most basic of all GPT's, Bing ChatGPT, on explaining and resolving the apparent paradoxes with relativity theory.  Initially it gave me the same basically wrong and bs answer that most people give but when I pushed back and clarified what the real issues were, it fixed its answers and provided great answers.   When you query a person on this, they can't see past the nose of their own arrogance just talk right past you thinking you don't know anything and regress deeply into ears-off mode, but basic b Bing ChatGPT actually comprehended my corrective follow up questions, took them seriously and dialed right in on resolving the apparent paradoxes with expecting a set of rules that appear symmetric but are not symmetric.   The ultimate answer is considering acceleration resolves the apparent paradoxes, that the real application of the relativity mathematics does not result in these weird symmetrical situations that cause contradictions, whether its special relativity without gravity considerations or general relativity with gravity.

Reply by Joseph Merrill

  • Avatar
  • Joseph Merrill
  • Member
  • Offline
  • Registered: 24-10-2023
  • Posts: 36

Re: I've been trying out ChatGPT by Open AI

Bing ChatGPT can also be horrible at offering up total fabrication, especially when you try to get it do an analysis and provide specific results in a specific format.  In that situation, a basic GPT will start to show signs of being what it actually does, which is just try come up with plausible response to your text prompts.   Yes, I think tons of capabilities emerge within such a program to accomplish the task, so there is emergent capability beyond the narrow confines of the task, but at the end of the day the basic GPTs are just training to plausibly respond like other people have, and any extra capabilities are unpredictable side effects.  Based on my experience with most basic GPTs I've played with, I'd expect the results of asking for specific chords resulting from an analysis to give you straight up BS.