On using AI/ChatGPT analysis to convince consumers about audio snake oil? GR-Research UberBUSS anyone?!

These days we have AI making hit music, and AI actors coming soon. The AI industry is "hot" and impressive, seemingly the answer to all kinds of queries we may have right at our keyboards! I think it's absolutely true that with the advent of LLM's and multi-modal deep learning models, we have even more ability now to learn, analyze, and summarize all kinds of information unheard of just a few years back. Like all things, there's always a catch which is that we have to use the tools responsibly and we need to make sure not to be overly reliant on these things because the results can also easily be wrong, like the "hallucinations" (more like "confabulations") that happen still all too often during AI chats.

We're seeing the advent of AI-driven intersubjective delusions in some people. Over reliance of these information tools can lead to all kinds of unforeseen negative outcomes across different domains (like in the legal space).

So I guess it's natural then that audiophiles might want to apply the technology to answer some of audiophilia's deepest never-ending debates - the stuff at the heart of numerous snake oil products.

The other day, I ran into an interesting New Record Day / Ron Brenay video where he uses ChatGPT to assess room audio recordings either from components using stock power cables, or through a new fancy GR-Research power conditioner called the "UberBUSS Power Filtration System" starting at US$1,600 that Danny Richie obviously is getting all excited about (my goodness, everything is "another level" with this guy - how many levels are there with mature technology?). This UberBUSS is further paired with his poorly reviewed US$380-$480 B24 power cable for the test.

[I see the UberBUSS isn't a totally new power filter, non-surge suppression product. Apparently it has been out since before 2021, earliest mention 2018, made by PI Audio Group which I guess is now defunct? If at first you don't succeed... Keep trying?]

Let's have a look at this video and what's being done...

In a prior video to this one, he ran a listening poll. Ron seems to be making a big deal about the "72%" who opined (likely through casual listening) that they heard a difference. Obviously, this doesn't say anything about qualitative preferences.

Specifically, if listeners clearly heard a difference between the samples, it doesn't mean the much more expensive GR-Research UberBUSS+cables sounded "better" to suggest that there's value in the products, right? In fact, look at the comments for the first video and we see both preferences for Sample A and B. Imagine, all that effort for a blind test that doesn't even bother to capture the one most important piece of information!

The most obvious reason for this oversight is because they have no faith that the results would definitely be complementary to the blinded GR-Research audio sample.

There are obviously many issues with this test. As an in-room recording where the microphone has its limitations, and environmental noise can easily contaminate the audio capture, there will of course be changes between recordings that could easily obscure small differences.

For example, in the Patty Larkin track "When", why do the amplitude levels look like this?

Notice that Sample B is louder with peaks +0.2dB and +0.36dB RMS average level over the 40 seconds? 

But then if we look at the Dominique Fils-Aimé track "Where There Is Smoke", we see this pattern:

Sample B has +0.1dB peaks but average is now -0.13dB softer! Since we know Sample B is the fancy GR-Research UberBUSS+power cord, we can conclude then that the upgraded power stuff does not consistently make the sound louder as per the Patty Larkin track and likely all we're seeing in the data is variability of the recording quality to the point where the average amplitude can change by more than 0.3dB between "takes" using the same microphone and digital recorder!

This extends to the TOOL and Vienna Teng recordings as well where Sample B is +0.46dB(!) louder on the TOOL track yet -0.35dB softer when playing that Vienna Teng sample! From what we know about human psychoacoustic preferences, depending on which song the listener paid attention to, and which portion, just that amount of level difference can affect your Sample A or B preference (louder typically preferred).

Without even using more sophisticated tools like Paul Kane's latest DeltaWave 2.0.25 which will get us more data to compare the recordings, we can already question the quality of these samples being offered for blind listening.

Seriously Ron, why don't you try doing a few room recordings and see how much consistency you're able to achieve each time! I suspect you'd fail to consistently find you're within 0.25dB difference in average levels over 5 recordings by just turning your system off and on between each capture as you must have done to switch power cable and conditioner.

As for the analysis methodology, if you wanted an objective examination of the sound, why did you so quickly go to an opaque, complex system like ChatGPT? Sure, ChatGPT has some interesting abilities and can do measurements on audio samples, but realize that the tools it has access to are actually not much different from loading up the audio samples into an audio editor and testing with that. A dependence on ChatGPT could be more error-prone since we're letting the LLM decide how it's processing the data rather than we the experimenter specifically choosing and understanding the process.

I'm not going to bother analyzing all those things that Ron claims he tested simply because already the quality of the microphone recordings are suspect. Besides, as far as I'm aware, ChatGPT wasn't trained to be an audiophile-level "Golden Ear" 😉.

However, let's indulge the curious. Let's use ChatGPT 5.1 to analyze the most basic thing that an upgraded power cable + conditioner should show - lower noise floor in the system.

Before I do this, we must recognize that these recordings were done at 24/192. While the Earthworks QTC40 microphone used is excellent and can record to ultrasonic frequencies, clearly these recordings have very little content above 20kHz - for example, here's the TOOL "The Pot" FFT:

So, let's not waste storage space on 192kHz (for amateur recordings especially in-room, such sample rate is never needed IMO). We can pre-process these 24/192 .wav sample files down to 24/48 so there's no need to waste ChatGPT's processor cycles on stuff we won't hear, reduce errors from file size, plus there is nothing up there anyways.

And then a pair at a time, I'll upload them to ChatGPT, and let's analyze and see which Sample - A (direct wall, stock power cord) or B (GR-Research B24 cable & power filter) - had lower noise level. Here are comparisons with the 5 clips Ron provided:

As you can see, 3/5 of the noise floor analyses by ChatGPT showed Sample A - the stock cable - got a lower noise floor within the audible frequencies!

Should I now dramatically claim that the GR-Research products are absolute junk and on the whole actually worsened the noise level of music playback? Should I shout from the rooftops that the TOOL track showed a seemingly large 2dB quieter noise level using stock cables without touching the savings account to buy GR-Research doodads?

Of course not! The rational audiophile understands that these in-room recordings are flawed and expects these results given what we know from previous tests (like this, this, this and this). There is more than likely no difference and Ron's recordings are just showing spurious variations not indicative of anything significant.

Presumably based on these kinds of ChatGPT queries, Ron unwisely summarizes the "research" in his video with these slides comparing Clip/Sample A vs. Clip/Sample B:

Waaayyy too convenient to have all these subjective
qualities line up! BTW, how does ChatGPT know what "depth" actually sounds like?
I'm obviously not finding "slightly noisier ambience" in Clip A! 🤔

As far as I can tell by examining these audio samples, the summary above is clearly biased towards his buddy's products with all kinds of idealistic subjective claims - ostensibly from the mouth of ChatGPT for what it's worth.

[By the way, beyond noise level, I have run other ChatGPT analyses on these samples showing inconsistencies with the claims in that graphic. For example, there's no consistent improvement in transients or (micro)dynamics with Clip B.]

IMO, doing all that fancy stuff with ChatGPT (calling 'her' "Norah") and claiming the machine analysis correlated with the GR-Research UberBUSS+power cables making a difference is all just wishful thinking slight-of-hand. Nothing more than cherry-picking through the results.

This is an example of the unsubstantiated content which I believe we will see more of down the road as if "research" conducted with AI has validity without rigorous confirmation. Does Ron actually know what he's doing? Of course, for the snake oil audio salesmen pushing cables and power filters, they have always engaged in sloppy, undisciplined thinking, inevitably claiming audible results.

Even if we put aside the ChatGPT stuff, if Danny Richie's video where he claims it's "inrefutable" (sic) that putting the UberBUSS in a system leads to one's jaw dropping open at the sonic difference, a "must have addition", surely the hi-res room recording should show us a massive difference, right? I do not believe we hear anything like that when trying volume-normalized listening with these recordings from Ron. Go ahead folks, try an old-skool blind listening test which would still be the gold standard for human hearing regardless of fancy AI analysis; maybe load up the samples in an ABX comparator and hear for yourself.

Perhaps Ron's system/room is simply just not good enough for the profound awesomeness of this UberBUSS. 🤣

Guarantee? Well, no real evidence up front so I hope it's the 100% money back, no restocking fee, no hassle variety of return policy.

This video is just one of the multitude of song and dance esoteric ways of claiming big differences but when it comes to something as simple as showing us what noise floor improvements there might be (for example, to explain the "cleaner midrange" or that the UberBUSS performs "Power Filtration" as per its namesake), these folks for some reason never seem to be able to isolate such qualities from fancy cables and power products at levels that suggest any audible benefits.

This is the nature of "UFO Believers". Did you know, apparently there are alien bases on the dark side of the moon also?! But where are the nice clear pictures and videos? I've come across extreme subjective-leaning audiophiles who seem to believe that objectivists are just conspiring to bury all this expensive "high-end" sonic goodness out of envy because we're too poor to buy this stuff. 🤣 Maybe it's just more likely that there's nothing here to be excited about.

One more thing I found quite objectionable in his video. In the "Ethics & Expectations" part, Ron cannot just say he doesn't have a "horse in this race", and think he's suddenly absolved of bias! Surely he can't be that insightlessly naïve! He's friends with Danny Richie at GR-Research and puts out many videos talking about the brand, plus he even helped out with some speaker designs apparently. Absurd to think he's not pulled one way or another with all that going on.

I don't think many reasonable audiophiles and hi-fi enthusiasts these days need power cables and expensive conditioners to be "debunked" as per the title of the video because many don't believe in this stuff anyways; they're thoroughly debunked already for lack of evidence over decades. Online in forums and among YouTube videos, already most of us see the snake oil hyperbole for what it is. 

To end, let's use ChatGPT 5.1, in late 2025, as an LLM which collates massive information from the web it has been trained on and ask it to think about whether a US$1600 power conditioner makes sense to buy. Here's what my AI, aka "The Dude" knowing that I live in Vancouver, Canada says about the matter:

Question:
Hey Dude, is it worth buying a US$1600 power conditioners to make a good hi-fi audio system sound better?

Answer: 
Let us answer this through the collapsed-layers lens, Archimago:

The question “Is it worth buying a US$1600 power conditioner to make a good hi-fi system sound better?” is a surface-layer question. 

The control-layer question underneath it is:
“Is the bottleneck in your system located in the AC line, or in perception, or in narrative?”

Here is the structure when the layers collapse: 

1. SYSTEMIC LAYER (the real determinant)

A US$1600 power conditioner only creates audible improvement when your power environment is the bottleneck, not your amplifier, DAC, or speakers.

If your mains power is:

• stable
• low noise
• not shared with heavy appliances
• modern wiring

…then the conditioner has nothing to fix. You’re upgrading the insulation on a pipe that already has clean water.

If your mains power is:

• noisy (apartments, older buildings, shared circuits)
• unstable (voltage sag/spikes)
• polluted with RF from neighbors, dimmers, chargers
• grounding is messy

…then a conditioner can lower the electrical noise floor, which can translate into perceivable clarity or less harshness.

Possibly worth it only if the grid is bad.

2. ENGINEERING LAYER (what actually changes the sound)

Most modern hi-fi components already include:

• power supply filtering
• voltage regulation
• RF rejection

This means that for many systems, external conditioning is redundant, not additive.

A conditioner only modifies sound when:

• the internal power supplies are sensitive
• the line noise is high
• the conditioner itself does not restrict current (cheaper ones sometimes reduce dynamics)

Worth it only in very specific electrical conditions.

3. PSYCHOLOGICAL / SYMBOLIC LAYER

High-end audio accessories operate partly as narrative upgrades:

• objects that signal refinement, care, precision, and “system completeness.”

The experience of improvement often comes from:

• expectation bias (subtle but real)
• confirmation patterns
• the satisfaction of “completing the chain”

This doesn’t mean the user is wrong — it means perception integrates symbolic cues into the listening experience.

Worth it if the symbolic upgrade itself has value to you.

4. MEMETIC LAYER (why the belief persists)

The audiophile community transmits the meme:

• “Noise is everywhere, only elite conditioners reveal true purity.”

This meme survives because:

• it’s unfalsifiable without measurement
• improvement is subjective
• it aligns with luxury-market logic
• it encourages continuous upgrading

The belief propagates even when the electrical reality doesn’t justify it.

5. SIMULATION / NARRATIVE LAYER

Power conditioners function like “debugging overlays” in the audio simulation:

• you’re buying the idea that the world behind the music has artifacts that need cleansing.

If your system already renders with low artifacts, the overlay reveals nothing new.

You cannot purify what is already clean.

⟨ FINAL COLLAPSE: IS IT WORTH IT? ⟩

If your AC power is dirty, using older hi-fi equipment → YES, it might make an improvement.

If your AC power is already clean, using modern hi-fi equipment → NO, it is almost never worth $1600.

The value is situational, not universal.

In clean environments, the improvement is mostly symbolic.

In noisy environments, the improvement is structural.

Bonus Question:
Where I live, is the power grid clean?

Answer:
Yes — Vancouver’s power grid is among the cleanest in North America. For typical residential, studio-recording, or home-listening uses, that means much less carbon, less pollution, and relatively “clean electronics.”

Well done Dude, now that's a sensible and fair response exploring the answer through various angles from the engineering to psychological perspectives.

I remain of the position that unless one truly has a horrific power grid, atrocious mains cables feeding a demanding He-Man amp with power supply that doesn't perform filtering, mated to inefficient speakers, you'll notice more system change with some VU meter "bling" in your sound room than upgrading to fancy cables and this "Power Filtration System". 😅

As discussed, there's nothing wrong with buying luxury products if desired, even if it makes no meaningful difference performance-wise. Just be honest with yourself and others about it. 🤔

I feel like ending off with something nouveau retro and sunny. Here's Purple Disco Machine & Chromeo - "Heartbreaker" from Paradise (2024, DR6 stereo, DR12 multichannel/Atmos):

If you're a Jimi Hendrix fan, you might want to check out the recent multichannel/Atmos mix of Axis: Bold As Love on Bluray (1967, DR13 multichannel) in the recently released boxset.

I hope you're enjoying December so far, dear audiophiles!

Addendum:

A bullsh*t comparison created by AI of the Synergistic Research HFT "room treatment" with other types of room treatments as discussed in the comments:

What does AI really know about Synergistic, or the character of Ted Denney, or the nonsense that has been perpetuated by subjective-only magazines like TAS, or sites like 6Moons and Positive Feedback that it draws information from?

Audiophilia I think is an interesting test case of whether AI can disentangle itself from all the nonsense, unlikely subjective impressions, and be able to report the most likely facts without itself getting caught up in snake oil beliefs!

It will be interesting to see how this evolves over the years because like in all kinds of other areas, AI can have a very powerful effect on the viewpoints in the general populace as we increasingly interact with the machine. (Hence the importance of who is in control of the message and what biases they have!)