• The move to the new server is done. There are some software and database maintenance updates in process. This has us passing the hat around to help out. We appreciate any donations. Seriously, even a dollar helps. The payment page may be found here - https://www.audiokarma.org/support.html

AI to the rescue

Brice

Brice

AK Subscriber
Subscriber
Hi guys,

For fun I asked the ChatGPT AI a vacuum tube question:

how to determine what impedance an output transformer should have to run a push pull of 6L6G vacuum tube with a 8 ohm load?


The response was:

If the load impedance of the output stage of your amplifier is 8 ohms, then you will need to select an output
transformer with an impedance ratio that matches the impedance of the 6L6G tubes. For example, if the
primary impedance of the output transformer is 8 ohms, then the secondary impedance should be designed to
match the impedance of the 6L6G tubes in push-pull configuration. This will typically be in the range of 2-4
kilohms, depending on the specific characteristics of the tubes and the amplifier.


Quite remarkable isn't it?
 
Register to hide this ad
I can see where it messed up already. It said the primary was needed to be 8 ohms but it is actually the secondary you want to be 8 ohms. :rolleyes:
 
I’m surprised it did that well. I would guess this is the first tube related question it tried to answer. LOL.
 
It's no substitute for someone who knows what they're doing. Never will be.
People who rely on it are damn fools if they don't understand how and why its wrong.
I cringe at the implications of this on my life.
Soon enough this technology will be cooking your food. Are you ready for that?
 
This is a garbage in, garbage out problem. The real reason it did so poorly on an easy problem like this is it has not been trained with the appropriate data sets. Train an AI model with good data and it will produce fantastic results most of the time.
 
In a very general sense (today) and depending on the task AI is chosen to do, "bias" can creep into the learning process and skew the results. This bias needs to be removed or eliminated at some point so AI can continue to process its task correctly. Akin to raising a child, when the child does something wrong or inappropriate, you have to correct or discipline it.
 
You asked for a procedure to be described and what you got was random information pulled from the internet which is highly generic and fundamentally incorrect.

It gave you the wrong answer and didn't show its work. In every math class I've ever taken that earns you an F. If this passes for intelligence then I'll remain blissfully dumb.
In School we were warned that copying and pasting from the internet was not a useful skill and was actually criminal. You can't take credit for someone elses work that you simply copied and called your own. That's all this AI can do! It plagurizes faster than any 100 individuals can read.
Becoming an expert in a topic takes time dedication and lots of hard work. AI is never going to be able to replace a true professional or expert in a trade or craft.
When they start trying I'm going to steer clear of those business which employ it, I already do...
 
kward said:
This is a garbage in, garbage out problem. The real reason it did so poorly on an easy problem like this is it has not been trained with the appropriate data sets. Train an AI model with good data and it will produce fantastic results most of the time.
Click to expand...

That's the biggest problem with AI now. The inability to effectively gather and evaluate (with emphasis on EVALUATE) new information.

Until an AI can successfully go out and acquire its own new data sets, and be able to discern true info from false- it's not really mature "intelligence". It's more like an infant, who has to have everything provided...

And as mentioned above- showing its work- as in citing its sources, and explaining how it got to its answer- is critical, to having meaningful, trustworthy results. However, even humans fail at this every day... urban legends, rumors, "fake news", etc. There's a lot of people out there that don't know how to properly vet a piece of data- so, expecting a machine to be able to do it, is asking a lot...

But, as of now- I don't see the fundamental difference between the output of the ChatGPT AI, and something like the Web Economy Bullshit Generator.

https://www.dack.com/web/bullshit.html

... because, without a knowledge of context and verification of data- it's all bullshit, anyway... :D

Regards,
Gordon.
 
Last edited:
I guess we should all get ready to bow down to our ro it overlords soon. I think AI is a great idea for some things. But as we all know, governments and militaries will abuse it as it progresses. Ya, I got a little dark there.
The real problem now and in the near future is people getting used to it and coming to depend on it. Glitches, bad data in, etc., does already and will continue to cause problems.
 
Can it really be called AI before it can learn? Will it do better on the 25th question about tubes? They are getting AI(ish).

To me this is just a far more elaborate version of the Apple IIe program I had that could slightly 'converse' with you by using what you typed to create its response. Just with a massive source of data to pull answers from instead of "I am happy" "why are you happy?". in nearly 4 decades we have advanced to Amazon's bot asking me about my problem where I received an empty envelope and telling me "sorry that item can't be returned".

If you have to train it with datasets is it truly an AI or just processing information as programmed.

That said, I think it could be fascinating to talk with the programmers to know just what 'it' is doing to find and process the answer. It would be cool to 'see it's work' for where it got the information and why it chose to use what it did. Is it possible for it to 'learn' if people pointed out the errors in its answers and showed it sources for the correct information? With so much wrong information on the net how do you guide it to avoid that?
 
So ChatGPT (Generative Pre-Trained Transformer) just had a major upgrade on March 14, moving from the 3.5 engine to the 4.0 engine. I am told the 4.0 engine is much improved. So I thought I'd ask it the same question @Brice originally asked above, but note I reworded the question so it would have a better chance, while still keeping to the same intent of the original question.

What is the optimal primary impedance of an output transformer when driven by a push pull pair of 6L6GC vacuum tubes, and with an 8 ohm secondary load?

Answer:
The optimal primary impedance of an output transformer when driven by a push-pull pair of 6L6GC vacuum tubes and with an 8 ohm secondary load depends on several factors, including the desired power output, the circuit design, and the characteristics of the tubes.

However, as a general guideline, the optimal primary impedance for this type of setup is usually around 5,000 to 6,000 ohms. This range provides a good balance between power output and frequency response, and is commonly used in many guitar amplifier designs.

It's important to note that the optimal primary impedance can vary depending on the specific application, and other factors such as the tube biasing, the output stage topology, and the type of music being played can also affect the performance of the amplifier. Therefore, it's always a good idea to consult the amplifier design specifications and experiment with different transformer primary impedances to achieve the desired sound and performance.

===

No more gobbledygook! It went from a Kindergarten education to a High School Diploma with that upgrade. That's pretty impressive in and of itself.
 
You must log in or register to reply here.
Back
Top Bottom