The emperor has no clothes, 21st century edition

[sungfw]

Audition list of orchestral excerpts according to ChatGPT

[BGuttman]

I read Doug's blog post from which this is drawn.

He is right, most of the trombone solos depicted in the "audition list" are vaporware; exception is Saint-Saens, but the trombone solo is not in Movement 1 -- it is later in the piece.

There is a trombone solo in the 2nd movement of Beethoven's 9th symphony. It's for the Bass Trombone and occurs 3 times. It consists of a single note: D above the bass staff (D4) held for 4 bars. Not particularly lyrical or expressive; nor is it difficult except for counting the 571 bars in a rapid 1 waiting to play it.

[Matt K]

This is generally the problem - at least for now - with these AI products. They are extremely confident and appear competent. Accuracy, however... well, that remains to be seen in a consistent fashion.

[JohnL]

There is a trombone solo in the 2nd movement of Beethoven's 9th symphony. It's for the Bass Trombone and occurs 3 times. It consists of a single note: D above the bass staff (D4) held for 4 bars. Not particularly lyrical or expressive; nor is it difficult except for counting the 571 bars in a rapid 1 waiting to play it.

Ah, but if one can demonstrate "a strong sense of melody and phrasing", along with "the ability to execute rapid, virtuosic passages with a clear, focused sound" while playing a single held note? That'd be pretty darned impressive.

[officermayo]

When they add the 2nd Bone solo from Sammy Nestico's "Switch In Time" give me a call.

[BGuttman]

Part of the problem is that these AI programs get their basis from the Internet. If all the data on the Internet was truthful, they might have a better chance. Still, they have no way to validate the question. Yeo asked it to research Homer Rodeheaver (a gospel trombonist) as a violin player and it just gave a nice (but wrong) report on Rodeheaver's violin playing.

They need to build some humility in these programs. As Yeo said, "I don't know" is a better answer than some of the bilge that the AI program generated.

[Matt K]

When they add the 2nd Bone solo from Sammy Nestico's "Switch In Time" give me a call.

Incidentally, I noticed that the first piece on PSO's list for associate tromonist audition (I'm presuming Rebbeca is retiring?) is Marie. It's a pops arrangement from the Boston Pops IIRC, but still, I thought that was an interesting choice for an orchestra audition.

[Posaunus]

I think I could handle that audition. With no mistakes. Beautiful tone. Strong sense of melody, phrasing, and musicality - just as written in the referenced scores. Not to mention that, in addition, I am a certified musicolologist!

[MStarke]

Great to see this gets some attention in the music world as well. Doug Yeo definitely comes up with some very valuable ideas and research.

In my daily job (management consulting) we just did some pragmatic evaluation on ChatGPTs relevance and applicability. We came to some interesting use cases, but also a LOT of limitations. But it is definitely fantastic in doing what it is supposed to do. This is not to do reliable research.

[JohnL]

They need to build some humility in these programs. As Yeo said, "I don't know" is a better answer than some of the bilge that the AI program generated.

It doesn't know when it doesn't know.

[brassmedic]

exception is Saint-Saens, but the trombone solo is not in Movement 1 -- it is later in the piece.

Actually, it is in the first movement. That symphony only has 2 movements.

[VJOFan]

Isn't ChatGPT out in the world in beta mode right now?

The whole point of it being live is so it can learn from trying and failing. The makers know it isn't perfect or all powerful. That is why there is a thumbs up/down with each generated response and a request for the user to submit an ideal response if it a thumbs down.

There is a healthy debate to be had about the technology itself and where it will ultimately lead, but not about how smart it actually is. It's a toddler right now. Of course we can make it say stupid things. Eventually it won't be so easy to fool it if it is properly taught- but should we teach it?

I think what we need to wonder about is how good do we want a technology like this to be? What does the world need something like this for? What guardrails should be in place so these fancy bots don't cause more harm than good?