Better Presentation of Preference Score?

[amirm]

I asked ChatGPT Pro to give scores and plot the components of preference score for Clarity 66 speaker and compare it to Ascilab C8C. The results were quite a bit more understandable to my eyes:

The key to visual clarity is the normalization of each component as to best it could be.

What do you think?

 

[kemmler3D]

Seems like a sensible translation, I guess my main question would be whether this distorts the contribution of each factor too much? I don't remember how the equation goes but they don't have equal weighting, do they? Maybe it's ok for this anyway. It definitely seems more intuitive / informative as a radar plot.

 

[Ogoript]

Interesting. Does the PIR matter as much if the actual is good? Makes the 66 look quite a bit worse because of the PIR, but the are the same in the "actual" room smoothness. If I'm reading that correctly.

 

[GD Fan]

We used to have a system that generated a 'spider web' graphic like this for asset risk factors. Definitely a handy single visualization.

For a preference score though does it visually equilibrate inputs that in reality have different coefficients in the score equation?

 

[beeface]

not bad. would prefer to see the grading changed to panthers, though

 

[pablolie]

"smoothness" seems subjective to me. which measurements give that away? balanced presence? Distortion at higher SPL?...?

 

[DMill]

May I ask what prompts you used?

 

[pablolie]

May I ask what prompts you used?

not directed at me, clearly. but whose prompts for what?

 

[DMill]

not directed at me, clearly. but whose prompts for what?

My mistake. It seems Amir used AI to create a graph based on pre-existing preference data. I had read it quickly and thought AI was generating an opinion on preference. Ive found AI likes to agree with us based on prompts. But to create a graph like this seems a great use of AI. That said I’d still be curious?

 

[pablolie]

My mistake. It seems Amir used AI to create a graph based on pre-existing preference data. I had read it quickly and thought AI was generating an opinion on preference. Ive found AI likes to agree with us based on prompts. But to create a graph like this seems a great use of AI

AI - unless prompted otherwise-- will just tell you what it thinks you want to hear. Works better for subscription renewals.

 

[kemmler3D]

"smoothness" seems subjective to me. which measurements give that away? balanced presence? Distortion at higher SPL?...?

I think this is smoothness of the various frequency curves, it generates a single metric. There's an algorithm for calculating it... I couldn't describe it to you but I did find the source code. https://docs.rs/autoeq-cea2034/latest/src/autoeq_cea2034/lib.rs.html#408-420 ... the patent goes into some detail but it's quite dense: https://patents.google.com/patent/US20050195982A1

 

[SigbergAudio]

I am always a bit vary about trying present something inherently complicated in a simple way. A bit like the preference score itself. It's certainly easier to compare, but it also removes nuance.

 

[Pauli]

AI - unless prompted otherwise-- will just tell you what it thinks you want to hear.

I see ...

I just asked ChatGPT if I can be totally sure that I will live to be 80 at least.

Guess what – I actually did not want to hear the answer.

 

[Matias]

I am always a bit vary about trying present something inherently complicated in a simple way. A bit like the preference score itself. It's certainly easier to compare, but it also removes nuance.

Well in a way that is what these plots do, give better presentation of the components other than the single number of the score itself.

If only we could have other devices like amps and DACs with multiple metrics being plotted the same way...

 

[rolfrander]

But to create a graph like this seems a great use of AI. That said I’d still be curious?

Creating a graph like this in excel is also quite simple:

 

[Gorgonzola]

I asked ChatGPT Pro to give scores and plot the components of preference score for Clarity 66 speaker and compare it to Ascilab C8C. The results were quite a bit more understandable to my eyes:


View attachment 548895
The key to visual clarity is the normalization of each component as to best it could be.

What do you think?

I like the general idea but is "smoothness", presumably of frequency response, the be-all-and-end-all of performance? So what of distortion which, IMO, often seems to be the neglected parameter.

Also, how are the raw numbers determined, and someone remind me what "PIR" stands for.

 

[amper42]

I asked ChatGPT how to fix a shadow double entry in USB drive on RPi4 Moode. Of course, it proceeded to act like it knew what SSH commands to do up until the commands it provided completely erased the 98 albums on the USB drive.

AI is just as likely to give bad data than not and act like it was totally right the whole time.
I wouldn't give much validity to this AI chart.

 

[staticV3]

AI hallucinated scores? Yeah no thanks.

@Maiky76's graphs are just fine.

 

[respice finem]

It is possible to present the scores in more ways than one.
The lazier ones can take a quick look, the more diligent can "dig deeper".

 

[Pauli]

I asked ChatGPT Pro to give scores and plot the components of preference score for Clarity 66 speaker and compare it to Ascilab C8C. The results were quite a bit more understandable to my eyes:


View attachment 548895
The key to visual clarity is the normalization of each component as to best it could be.

What do you think?

Looking at charts like this I am always suspicious regarding the impression that the sizes of the areas covered have on me. Let’s say model A’s »pentagon« has a size double that of model B. That would seem to claim that model A is »twice as good« as model B ...