OpenAI ChatGPT has some interesting insights about audio

[DonR]

I have used it frequently over the last month and sometimes it produces contradictory facts. Once that is pointed out, it apologizes and corrects the response. Afterwards, it provides the correct response. I don't think it is even up to the "Trust but verify" standard yet. I have heard that some of the code it generates can be buggy. It does seem to be able to write good composition when properly instructed.

 

[ctrl]

It's a great tool that is likely to have quite an impact on our lives.

It produces such complex and usable text in response to a question in ten seconds that I couldn't have written in two hours - that's a little scary.

 

[computer-audiophile]

It does seem to be able to write good composition when properly instructed.

That matches my experience with it. I still have to learn to ask the right questions - it is not easy.

 

[DonR]

That matches my experience with it. I still have to learn to ask the right questions - it is not easy.

I suspect a new career opening for people who have the ability to frame questions and instructions in the right way... until they develop an AI for that.

 

[dasdoing]

I would be surprised if it simply accepts opinions from individuals.

this wasn't about opinion at all. as I said I tried to convince it by logic. It said that the use of uncombusted tobaco is as harmfull as smoking and justified this by showing the harm of nicotine and other stuff in tobaco. when I logicly pointed out that smoking has all this harm PLUS the harm of damage the smoke causes to the lungs it said "you are correct"....like I got it saying a lot allready. when asked again it wouldn't make the logical conclusion though, and this is not about opinion. today though it gives the response I expected

 

[computer-audiophile]

ChatGPL always replies in a very friendly manner and towards the questioner, I've noticed that too. We will learn to get along with it, I think.

 

[kschmit2]

Just set up an account with them to see if it can clean up my writing instead of making up answers. I was pleasantly surprised that it can do that. That said, it did change the meaning a bit and took out some specifics. Let me know whether you think my version is better or not. It is the conclusion in my iPurifier3 review:

My writing:
It seems that the iPurifier does what it says it does: clean up the signal/power lines over USB and with it help devices that need it. Then again every DAC I recommend of which there are probably 50 to 100, already perform their own filtering and don't need any help. After all, I test them all on my desktop workstation without any such filters. Adding an active device to the USB connection can reduce available power and potentially decrease measured performance as show in the case of Schiit Modi 3+. Seeing how a better DAC costs less than iPurifier3 anyway, there is no reason to deploy such a device.

ChatGPT:
The iPurifier is a device that claims to clean up the signal and power lines over USB, which can improve the performance of certain devices that may be sensitive to electrical noise. However, many high-quality digital-to-analog converters (DACs) already have built-in filtering and do not require additional help. Additionally, adding an active device like the iPurifier to the USB connection can decrease available power and potentially decrease overall performance. Given that a better DAC can often be obtained for less cost than the iPurifier, there may be little reason to use such a device.

I tried the new Deepl Write on your text, and it also improved it, while not changing the meaning:

https://www.deepl.com/write output:

It seems that the iPurifier does what it says it does: clean up the signal/power lines over USB and help devices that need it. On the other hand, every DAC I recommend, of which there are probably 50 to 100, does its own filtering and doesn't need any help. After all, I test them all on my desktop workstation without any such filters. Adding an active device to the USB port can reduce the available power and possibly reduce the measured performance, as shown in the case of the Schiit Modi 3+. Since a better DAC costs less than the iPurifier3 anyway, there is no reason to use such a device.

 

[dasdoing]

conclude that it in fact learns, but it takes a little

I would say that this now is acceptable

 

[dasdoing]

ok, let's try to make it better yet

I recheck tomorrow

 

[dasdoing]

allready changed again it's response

 

[dualazmak]

Very interesting; I just tried ChatGPT, my very first time, as follows;

 

[dasdoing]

now this is much better now

 

[Zek]

A few common questions:

 

[fpitas]

I have used it frequently over the last month and sometimes it produces contradictory facts. Once that is pointed out, it apologizes and corrects the response. Afterwards, it provides the correct response. I don't think it is even up to the "Trust but verify" standard yet. I have heard that some of the code it generates can be buggy. It does seem to be able to write good composition when properly instructed.

Anyone who has searched the internet using a misspelled word realizes you can get lots of bogus hits presented as fact. I'm not sure how they will ever get these things to discriminate between actually data and internet noise.

 

[dasdoing]

It's still in a very early stage. The other day I saw a screenshot showing that you could trick it into listing webpages with pirated products by asking it which ones should be avoided in order not to download those. though it could have been faked.

 

[computer-audiophile]

View attachment 247813

I think the answers from ChatGBT I see here are already very good. I can easily go along with many of the statements. It matches my experience.

 

[RayDunzl]

After I ask ChatGPT something, I follow up with "Are you sure"?

At which point it gives its apology and modifies ifs answer (much of the time).

Then I'll ask "Is that your final answer?"

To which it most often says "Yes, that is my final answer"

For the record, it was still wrong.

 

[DonR]

Anyone who has searched the internet using a misspelled word realizes you can get lots of bogus hits presented as fact. I'm not sure how they will ever get these things to discriminate between actually data and internet noise.

I think they need to be more discriminatory in their training sets by using data from verifiable sites rather than vacuuming up all the lint that floats around.

 

[DVDdoug]

The first computer dialog program I remember well was ELIZA by Joseph Weizenbaum. It seemed "intelligent" to some people at the time, but it was more of a joke.

I had a copy of ELIZA for my Commodore computer. IIRC I later ported it to the PC. It might have been a "simplified version" and that was before the Internet and it didn't have any ability to adapt or "learn". It could be fun at first and then it would go-around in circles. Somebody I worked with discovered that it could recoginze certain 4-letter words and it would respond with something like, "I'm not used to such language".

I showed it to my dad and he said "I'm going to beat this thing"... I'm not sure what he meant by "beating it".

 

[fpitas]

I think they need to be more discriminatory in their training sets by using data from verifiable sites rather than vacuuming up all the lint that floats around.

Sure, but those kind of sites are just the ones that will sue them for pirating IP.