AI (Audiogon Intelligence)

@mapman - the responses to a question on AG are tempered by peoples personal experience or preferences - not to mention the various tangents people go off on.

Factor in Gear combinations, room acoustics etc..and I think the AI response could end up being a little confused/complicated?

Also, buiding the AI database would have to contain the various components/cable permutations and user preferences.

On the surface it seems like a good fit, but the existing and expanding information base would probably be largely incomplete, so it would not really offer viable responses.

But then, I could be wrong.

I'm sure it will happen someday, but presently searching AG and forming your own opinion based on responses and user might be the best way to go

Regards - Steve 

I would think that Audiogon was one of the sources that ChatGPT was trained on. The dataset it was trained on was absolutely enormous. Chat GPT does a pretty good job on questions like “what are the sound characteristics of say McIntosh equipment. When you ask questions like what are the best sounding DACs and what are the sound characteristics, in general it does well, but you have to verify. It may very often pick streaming DACs or some odd items that are not quite in the category.

Bard / Gemni can be far less useful. The other week on Sunday I asked if the stock market was open tomorrow (it was not)… it gave a long explanation… and was wrong! It said it was open.

So, the training set matters, and verification is required.

I can’t imagine the forum making much money… so it would be hard to see the huge investment needed to implement AI likely. But it would be fun.

I noticed my Linn turntable was making some noise yesterday so I google searched the topic in Microsoft Edge and was able to resolve. MS Edge provided an answer from copilot that was also useful.

I work in technology development for a major US financial services company. We have started to leverage copilot there as well to a limited extent initially. That particular AI app seems pretty good with knowledge in the technical domain. The information is usually accurate. I also have started to dabble with Poe app and find that very useful in general as well.

My first real job back in the 80s was in developing pattern detection and signal processing algorithms for government funded mapping systems of the day (hence my moniker). I have good familiarity with how these things work. They should definitely be capable of providing reliable answers to many questions in the electronics and high fidelity audio domains, including identifying patterns in subjective as well as objective knowledge. We live in very interesting times in this regard!

I’d invite folks to try asking your favorite AI app some hifi audio questions and sharing the answers and your thoughts on those here. 

@ghdprentice no doubt how the network is trained matters.

One thing I like about Poe for example is that when asked a question it often puts a disclaimer at teh top that its knowledge only covers up to a certain date, so for example it would likely say what it knows about Stock Market hours but qualify that it does not have recent information to base the answer on. That’s awesome! How may people are able or willing not just to tell you what they know but also exactly what they don’t know? That’s the right way to do things!