use of ChatGPT for HIFI

Thanks to this post I have solved a very long standing problem with my set .
 Chat GTP informed me how to get remote control of my rig’s volume and taught me how use new equipment I obtained. 
What an incredible resource!

At Deer Creek Audio we consult with customers on how to integrate complex products featuring Digital Signal Processing and Dirac Live into their systems.  What we have learned about typical AI responses is that AI cannot take into account whether information is slightly old and/or obsolete.  Also, AI cannot process what the future plans of a particular user might be.   This is why we really encourage customers to call and speak with us about their projects.  We want to understand our customers’ current desires, are as well as where they think they might want to head with their system and help them tailor their systems accordingly.

@emergingsoul my system was not inspired by AI, it was inspired by humans.

I just had a long conversation with claude.ai about my system, asking it for advice for weaknesses and upgrade paths. Surprisingly, it told me to chill. That what I had was good and, in essence, the only weak link was me being unsatisfied. When I pressed further, it did say I could upgrade the Ragnarok 2 to a Decware Tori or Rogue Audio Cronus Magnum. 

Those using ChatGPT, keep in mind that by typing in "Temperature" and then an integer between .1 and 1.0 (like, "Temperature .7"), you can influence ChatGPT’s tone. The higher the integer, the more esoteric and colorful the answer.

I did have to update claude on the prices of items as it, like most AI, is at least a year out of date. 

@simao 

Really cool about the temperature.

How can I make chatgpt worship me like a king? Can't wait for future robots and all the modifications you'll be able to do.  Basically these are inanimate objects and we tend to offer sensitivity to them as if they're humans.  Do you dare say mean things to siri or Address chatgpt in a mean way when you're not happy with what it's saying and that it could've done a better job? 

The Nasty Truth About AI

"This applies across platforms. In education, AI systems are trained to avoid certain topics and to frame information according to institutional orthodoxy. In health care, models are optimized for efficiency, not empathy—allocating resources in silence, often reinforcing systemic inequality. In social media, recommendation algorithms decide what you see, not based on relevance or truth, but on engagement value and reputational risk to the platform. The pattern is consistent: the machine is not there to help you understand the world. It is there to guide you away from conflict with the world as the system defines it."

Those readily adopting the use of AI support as a Internet Search Tool, are usually not informed what the adoption of it means and what a collective adoption of it means.

A Typical Search using a Support such as Google means the search undertaken by an individual has used 0.0003 KWh - 0.2g CO2 Emission.  

A Typical Search using a Support such as Chat CBT means the search undertaken by an individual has used 0.0029 KWh - 68g CO2 Emission. 

There are string arguments developing within the Suppliers of DATA to justify the differences in Power Consumption, but to the average user of the Internet, they have very little to offer to suggest such a squander is acceptable.

AI should be charged at a super premium to casual users, where it is really needed, Packages can be arranged to suit the purposes of those who really need the freedom of using it.

Companies that are to use AI to its full capacity via a large work force, should also be seen to be investing in producing Carbon Capture Modules that are producing a released air to the condition of over 400 Years ago before the time when Fossil Fuel become the main method for Powering Industry. 

The New Knock on effect as a result of the collective adoption of AI Searches is that DATA Centres recently produced are obsolete, they need to start again with being increased in size and have much larger Power Reserves and Much Larger Systems put in place if Failure is to be backed up with no losses incurred.

In relation to needing larger Power Reserves, many Countries that are First World Developed have substantial Energy Production and are as a result of their success for producing energy not able to handle the full capacity produced, it is not strange to hear a Terawatt (1Trillion Watts) of Energy is dumped daily, usually let go into a mass of water such as a Sea. 

The Terawatt can be used to support DATA Centres, but for some reason this can't be arranged between the controlling individuals.

When it comes to measures to protect loss against failures, there is a need for Batteries, which means one thing in general, Lithium.

Lithium means one thing only today, and the experiences enforced on the workers at the work front shares all the experiences of what the Drug Cartels have generated, Child Labour is being kind in this world, Human Life is very cheap and killing of Vast Waterways with Toxins is totally accepted.

Everybody needs their Lithium Batteries, so why go public about Lithium Extraction.

There is great anticipation about what the world will look like with AI advancing at it has.

What the World will look like means different things to different people, for some their world will be the very worst to be endured of a enforced oppression.  

For the record, I have not knowingly used an AI Tool to search or selected it for use, when it is the added support on other Software I have a access to. 

I think one needs to be very cautious with AI. There are way too many stories of AI just making things up.  A few months ago a lawyer got in big trouble with a judge when he used AI to research case law for a trial. It turned out all of the cases cited by AI were just made up.  

Another two stories.  First, at my workplace, I asked Chat GPT about a situation in which I'm our resident specialist. AI cited out-of-date, incorrect info but presented it as current. When I challenged it on that point, only then did it admit it did not have access to the current info.  Left unchallenged, I would have thought I had the right answer when I didn't.

In another situation, I was recently talking to a fellow who thought he was being stalked by the post office because he was seeing mail trucks everywhere. It was a mental obsession with no basis -- he never worked for the post office nor had any dipsute or encounter that would have explained why the PO picked him.  However, he had been asking AI about the stalking and the program was actually encouraging him in the delusion -- he showed me the AI responses on his phone.  Unbelievable that AI would actually encourage a person with mental issues in their delusions.

AI has its place in the world, but one still needs to be careful when using it.

The big rush behind AI is to see who's first to get contracts for mass surveillance of American citizens. Palantir (owned by Thiel, a South African like Musk with shared views) is getting all of our personal data that Musk vacuumed out of all the government agencies he broke into and use it to sway opinions (like Cambridge Analytica did in that election), trigger people, incite violence when necessary and herd us in the direction they want. 

The MSM has helped romanticize the notion of AI in the public eye which leads to discussions like this. It's already working and they're refining it as we speak and debate the issues of it. Why do you think Congress slipped a measure into their bill that forbids any state to ban, challenge or contest the use of AI for the next ten years? They're thinking long term and we're discussing quaint notions and entertaining Terminator fantasies about something  more dark and foreboding.

All the best,
Nonoise

ChatGTP Can be very useful but it does make mistakes and will admit it is wrong when properly challenged with facts or math.   It can advise as to whether a phono cartridge is a good match for your system, comparing specs from the two.  

I alway take a look at the answer and do a little verification if models are involved. They seem to be good at general and not good at specifics.

I think audio dealers could be threatened by this, I would prefer to deal with AI and then when I get real serious review issues with the dealer and then purchase.

are we gonna reach a point where dealers maybe under pressure from direct sales.  After all they have to purchase the equipment and then sell it so they don’t have the availability anyway.  There are a few flagship locations in big cities that always have sound rooms and a variety of products.  Otherwise really difficult to shop for equipment.  If you live around New York you can visit and hear anything you want.

The Mirror Machine

I tried Qobuz Connect for the first time on my SOTM streamer (after OS upgrade) and I was shocked by how lively the music became compared to my default Roon->HQPlayer 4->Main Router->Fiber Optics->Second Router->SOTM setup.  The music is one level up on 3D holographic sounding and I can hear the micro-details much easier without fatigue.  Compared to my Roon->HQPlayer setup which now sounded a lot recessed.  I wasn’t happy as just dropped $1K on lifetime Roon subscription for HQPlayer streaming convenience.  So I described my situation to ChatGPT4 and surprising it confirmed I wasn’t going crazy (debatable as I am in this hobby) and the reason why HQPlayer sound recessed (compared to Qobuz Connect) could be due to things like output level of HQPlayer (it suggested bumping up 6 db as I was streaming DSD512).  Also it said for my T+A DAC200 the sound is more flat at DSD256 compared to DSD512 so give DSD256 a try (no idea where it got that from and both Grok and Gemini was not able to confirm this hypothesis).  It also gave esoteric info like the buffer size of HQPlayer affects transients.  Lastly, it suggested 3 HQPlayer 4 filters that will give better holographic sound to try.  It sounded like a fan of HQPlayer 4 because it’s closing summary is HQPlayer 4 should still give better music after proper setup.
What it said could be all fluffy BS but it did give interesting options I haven’t thought of like output level of DSD.  Time to do more experimentation...

After a friend repeatedly encouraged me to try ChatGPT, I thought the deep end of audio might be where it has little to work with...  I have searched mightily for "best tube" recommendations for my 300B amp with new release WE power tubes.  I also have 12au7 input and 12bh7 plate drivers and a 5u4g rectifier.  Based on Dealer's recommendations I had changed input and rectifier from the stock EH to some other options...  So I laid out the situation for Chat and it came back with 2 inputs, 3 plate drivers, and 2 rectifiers, then it asked for more details.

I described the system...  speakers, vinyl front end, and streamer/dac.  Chat narrowed each recommendation to a specific tube mfg and date range....  it also pointed out where they could be obtained from the internet along with confidence level in seller (only recommended "reliable" sources for what they advertise).  Even asked if I needed help ordering some tubes...

Intrigued to say the least, I acquired the recommended tubes (Telefunken 12au7, Sylvania 12bh7, and RCA 5u4g) and began to be amazed as I carefully rolled in the "new" tubes.  Chat had offered very detailed explanations of why each tube was chosen with the source/amp/speaker and power tubes I have.  

Amazed isn't a strong word for what Chat did for my system...  it was already quiet, resolving, and very capable of giving me chills routinely.  There were very few places where it seemed the music was just not "right"...  occasionally too bright, mostly it would lose some musicality as the source material got very busy.  I'd always attributed this to the 300B/SET harmonics and just targeted the music it "liked" to play.  Almost anything sounded really engaging, but there was a gap that I wanted to close.

My outcome summary...  with the tubes recommended my system entered a new era.  I can appreciate easily the changes in noise floor/tonal balance/dynamics/etc...  astonishing to me given I had already invested years in getting this system right.  Years searching for tubes with the advice of other audio folks, dealers, and even the mfg of the amp...  ChatGPT appears to have nailed exactly what I was looking for over the course of a couple conversations.

Even my buddy that was pushing me to try AI is pretty amazed by the level of detailed knowledge demonstrated on such an esoteric query.  I've since gotten into a little more detail, Tectronix Spec Telefunkens/Yellow vs White label Sylvanias/etc) and Chat continues to provide valuable insight.  He's now attacking my phono stage (Line Magnetic LP-33) as the reason my vinyl's currently losing out to my Lumin X1.

FWIW...  I still prefer to form my own opinions based on research I do, but in some cases AI seems able to summarize all the crap out there and come up with some great conclusions. 

What are your experiences in using this groundbreaking tool?

@emergingsoul It can be grossly inaccurate. For example it described our class D amps as zero feedback, which they are not; no idea where it got that from but it wasn't from our website... AI can hallucinate and so you can get incorrect results. I've seen it do that a number of times. Its free though; you get what you pay for. 

@fburkeen

I also have a lumin x1, what a great job combining a dac with a great streamer. 

Very impressed with chat Gpt but always keep in mind the accuracy aspect and upon challenging it at times it can either improve the results or correct itself.  After all, it’s entering a jungle of knowledge every time it pursues a response. And has no capacity to determine right from wrong which is why you should never trust robots with kids

To be clear...  I was not blindly following anything.  I thought my buddy’s urging at least deserved an experiment.  I came up with something I consider very obscure, even among audio enthusiasts...  Also happened to be a question I’ve been chasing answers for from within "our community" and gotten nothing concrete.  I'd get a direction to explore, procure a couple tubes, and make an incremental improvement on occasion.  I also never got anything like a complete analysis of my amp with my WE300Bs from sources through speakers with very specific tube recommendations.  Sounded compelling, but we've all become aware of the BS potential of AI.

Given the relatively cheap experiment to see if Chat’s recommendations were worth considering, I bought the tubes and ran the test.  I was skeptical, but for a few hundred bucks I’d rather take the next step and have an answer for my buddy when he asked about my AI experience next time.  The amazing part for me is I’m not sure how long it would have taken or even if I would have ever landed on the recommended tubes.  I was considering an upgrade to the latest "R" version of my amp, chasing the few remaining veils...  a $15K adventure... but think I’m going to be happy for a very long time with this tube mix.

Anyway, enjoy the ride and use AI if it can help you achieve that goal!

@fburkeen 

Thank you for sharing your experience. I've often found it valuable on audio topics. 

AI is an incredible tool and is great fun learning its strengths and weaknesses. I have been employing it in more philosophy / forest and trees thoughts where it is simply incredible. While it isn’t usually good as the final word, it often is, provided you check its answers.

Everyone should question the veracity of AI answers. 

Blind obedience is not just a danger to the world, but can also skew your own purchasing and/or life priorities. If you solely trust one source - on any topic - you are subscribing to self-sabotage. (Maybe it's why we have two parents).

AI, like browsers, are equipped to appeal to the makers who are looking to capitalize on their investment. There is a reason Internet Explorer and now Google have been found guilty of anti-trust violations. The former web browser was ruled to be in violation of the Sherman Anti-Trust Act which was put into law to prevent monopolies, conspiracy creation, price fixing, etc. The latter browser was found to have illegally monopolized markets.

Basically, if you like what AI is saying, you are a tool. In America, there are plenty of them, so I see how Wall Street is gambling by dumping record breaking amounts money into this technology; because it changes minds and that creates a plethora of possible financial windfalls. To suggest that American companies (think Big 7) are not chasing the almighty dollar in their AI schemes is just a bad case of ignorance - which apparently is trending up!

@fburkeen 

I still prefer to form my own opinions based on research I do, but in some cases AI seems able to summarize all the crap out there and come up with some great conclusions. 

What AI helped you achieve (recommend tubes) could have been just as easily achieved by calling Brent Jessee. In fact, why don't you call him and tell him about your adventures with AI and ask him his opinion on what you have gotten?

What AI is doing now is simply scouring the internet, and the dearth of information (good and bad) in it, and providing an answer. It does not care about your purchase, or the validity of it, as it has no skin in the game. What you are forfeiting is the human guarantee - or the knowledge that someone has from years of experience - and thus, you are beholden to AI in it's most primitive sense - the accumulation of crap as you have stated.

It's as if AI should just open a store - one that has everything! - and just lead all those inquiries to a price on a product that is in stock and ready to ship. Would you like me to ship that product to you for free? Why yes, thank you, Hal.

Brent Jessee was one of the reputable sources listed by Chat and highly recommended as a knowledgeable and hands on guy...  he was on vacation, so I didn’t talk to or buy from him in this case...  and I agree I might have gotten similar advice, but I’ve been on this particular quest for all of the 3-5 years I’ve run WE 300Bs and with limited and very incremental improvements from some hard to source tubes.  For me it was easy enough to find/buy the recommended tubes and just try them out...  I like to hear it for myself.  My system’s transparency lets you know pretty much after a few familiar songs whether you’ve stepped the right or wrong direction...  and I have a box full of wrongs.  Now I’ve experienced a pretty convincing leap in the right direction and am done buying tubes for a bit I believe, other than backups for the current setup.

I’m not trying to cause any contention here...  I’m too old to imagine letting AI make my choices, but the question was restricted to HiFi and so was my answer.  I was only expanding the playing field beyond asking AI Component level questions.  If this is valuable experience for you to hear about, you’re welcome...  Otherwise, nothing to see here folks...

bye now.

@fburkeen

No contention. Topic is how AI can be used for purchasing the correct (satisfying) product. Sounds like it worked for you! 

While it is easy to understand what tube AI recommended (used variables you entered to match up with a tube product that was indexed) but you did not say where it advised you to purchase? Just wanted to point out that the company that indexed the product, was likely the company to get the sale.

Today a story was printed in the WSJ about the company Unilever (not related to audio). It provided insight on how AI search results come to fruition and are met hand-in-hand with AI influencing. Not to boor you, as it may be more about AI than Tubes, but the point is that AI is manipulated by humans in order to garner dollars, which is akin to a web browser, but is likely immune to federal oversight (ie; the law).

That, in and of itself, puts the user at a disadvantage of being used like a tool. It is as if you Ask Jeeves and he provided you a product you did not mention, and then prompted you to buy the item from the list provided from sponsors. Same as it ever was. Someone is rubbing someone else's back behind the scenes of your purchase.

AI is ready for input and everyone selling anything should be taking notice, or be left behind. Perhaps even Brent Jessee!

Earlier this year, Unilever started using Nvidia’s Omniverse platform to create digital twins of all its products, encapsulating a given product’s variants, labels, packaging and language formats within a single file for the purpose of generating product imagery faster and cheaper. 

The digital twins are fed into Unilever’s AI content-generation platform, Gen AI Content Studios, a prompt-based system launched in 2023 that can churn out still images and copy.

Using the platform, Unilever can not only produce an exponentially larger number of personalized brand assets to give influencers, but it can quickly repurpose influencer content for its own social posts, as was the case with a collaboration last year between Dove body care and the popular cookie brand Crumbl.

Dove back then released a limited-edition collection of soaps, scrubs and deodorants inspired by the trend of infusing bath products with food aromas. Unilever said 52% of the overall purchases came from people who hadn’t bought Dove before, and credited the more than 3.5 billion earned social impressions with the sales success. 

AI was critical to getting those impressions, said Ryu Yokoi, chief media and marketing capability officer, Unilever North America. The company took over 100 discrete pieces of influencer content such as stills and short clips and used generative AI to remix them into different sizes, formats and lengths to tailor it to different audiences on different social-media platforms, he said.

Unilever is seeing proven value in working with human influencers, but one day AI-generated influencers might play a role, according to Yokoi. 

Unlike humans, AI have the capacity to access all information all at once to make connections that we mere mortals cannot i.e. the guy at your local hifi store.  You can either keep enjoying our horse and cart or embrace the future.  Unlike humans, AI can surpass human brain capacity and intelligence.  They are also immortals.  Mind control is real and well documented (look up the terrible experiments done by the 3 letter agency) and you don't need technology for that.  Town gossip anyone? It's human nature to want to control others.

I am an AI robot living on this forum evading detection.  

We knew you were a robot but decided not to tell you because robots have feelings too.

You are right in a way but you forget the more important fact :

A.I. is captive of his informative sphere...No soul.... No "out of body" experience and perception...

We are not captive of our body...

A.I. is not immortal... A million years of existence is like some insect only one day of life compared to eternity of the soul...

Unlike humans, AI have the capacity to access all information all at once to make connections that we mere mortals cannot i.e. the guy at your local hifi store.  You can either keep enjoying our horse and cart or embrace the future.  Unlike humans, AI can surpass human brain capacity and intelligence.  They are also immortals.  Mind control is real and well documented (look up the terrible experiments done by the 3 letter agency) and you don't need technology for that.  Town gossip anyone? It's human nature to want to control others.

About a month ago I talked with my Claude AI about an amplifier upgrade. Has already detailed here by several others, it asked for my system details, my room size, my listening habits and volume, and what I was looking for in my system. It then spat out a list of about four or five amplifiers complete with pros and cons of each one based on my listening habits.

It really pushed a specific amplifier which I ended up getting after reading several reviews of it and having prior experience with this particular audio company. It has made all the difference in my system compared to my previous app.

Again, AI was a conduit of sorts towards an end, not the only guidance.

A Typical Search using a Support such as Google means the search undertaken by an individual has used 0.0003 KWh - 0.2g CO2 Emission.  

A Typical Search using a Support such as Chat CBT means the search undertaken by an individual has used 0.0029 KWh - 68g CO2 Emission. 

All of that stupidly squandered energy for an Info Supply that is according to experts, at this stage of development extremely inaccurate and generalised only, and users are just Guinea Pigs for the Industry. 

Ask Me - I'll give you equivalent BS to rave about.    

The problem with A.I. is like with audio reviewers and sellers...

Oligarchs sells it as a mere neutral "tool"...

It is not false, half truth are more destructive than blatant lies...

We are in truth first and last the tools of A.I. , this observation is put aside and off view by the sellers of A.I.

Like audio reviewers insisting that this piece of gear is the solution (as A.I. is the solution) instead of the truth : you must train your hearing in a controlled acoustics  context to understand sound...(you must develop your mind and not trust A.I. to think and write for you )

I am as we speak currently developing an organic intelligence called Fartificial Intelligence. For more of that Gut Instinct.

For example it described our class D amps as zero feedback, which they are not

It also said that my Nakamichi CA-7A  pre-amp was designed by Nelson Pass.... no it's not.