• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Useful Applications for AI

Lol. I'm waiting for the day where AI locks that thread by itself, but leaves this (more complimentary) one alone.

Ahem.
Apologies.
 
JSmith said:
Yep, went there first mate... thread is locked. :)


JSmith
Click to expand...
So it is. Perhaps that should suggest to us that we shouldn't just bring discussion intended for there, to here. Perhaps we should just wait for that thread to be re-opened, so that we don't get this one locked too?

:-)
 
As the other A.I. thread is currently in 'lockdown'.......

Perhaps it would be useful to ASR members to compare answers to common, audio-related, questions from popular, free and paid A.I. services ?
 
Power Pop 23 said:
As the other A.I. thread is currently in 'lockdown'.......

Perhaps it would be useful to ASR members to compare answers to common, audio-related, questions from popular, free and paid A.I. services ?
Click to expand...
Not here.

If you're not happy to wait for the other thread, start a new one. But don't be surprised if it gets shut down. Read the AI policy.
 
Not sure why the other thread is still closed.

Perhaps we should have both a "Positive aspects of AI" and a "Negative aspects of AI", thread?

Anyhoo, back to my main concern.
A tad worrying, I would suggest.
As per: YMMV.

www.bbc.co.uk

Anthropic researcher believes more than 10% chance AI 'could kill all humans'

It is the latest in a series of increasing warnings about the safety threat posed by artificial intelligence.
www.bbc.co.uk www.bbc.co.uk
 
Last edited:
^ Publicity for the company, keep it in the news. ^

Job done
 
My first challenge to AI engine Gemini for search in ASR Forum with inquiries for which I knew correct answers and right information

I do hope it would be allowed posting this which is almost identical to my post #1,051 on my project thread shared today.


It seems that AI’s learning and exploration capabilities have been continuously making astonishing progress—even in a span of just one month.

So, for the first time, I decided to give a "challenge" to a specific AI engine Gemini. I wanted to see just how accurately and in detail Gemini could answer questions about the contents of ASR Forum specifically regarding matters where I already knew the correct answers— since I myself was the one who wrote the original material.

In a Windows 11 Pro PC which I seldom power-on, I newly installed Firefox web browser with no account making nor any log-in to an account, so that Firefox and the free Gemini do not know (I believe “not”) I am the author of the specific contents in ASR Forum for which I put my inquiries; also, in the PC, I have no files/documents relating to my ASR contents.

Then, I went into the free Gemini web site, without “account log-in”, starting my inquires/questions as follows shown under the spoiler cover bellow. In This post, I share only my inquiries —you may have LOL for some of them—, I dare not including the responses from Gemini which consisted of many sentences in proper segmentations for each of my inquiries, sometimes with headings and bullet points.

I was generally impressed by the responses from Gemini, except for only a few minor misunderstandings and lack of relevant information.

That is it; nothing more, nothing less; it was just my very first personal challenge to Gemini (and now Gemini knows “my” challenge after seeing this post).

I have no intention at all, at least at present, to use Gemini as “search engine” for ASR Forum; I continue searching ASR by myself “within ASR” including the “Search” functionality of ASR itself.

In case if ASR will officially incorporate its own “AI assisted advanced search”, I will be more than happy trying it.

My challenging inquiries to Gemini:
  • Hello Gemini, can you directly review and learn Audio Science Review Forum?

  • OK, would you please summarize the essences of Mr. dualazmak's multichannel audio project?

  • Can you summarize "pre-project historic sequences" of his multichannel audio project?

  • What was/were his primary motivation(s) to go into multichannel active audio setup?

  • Why does dualazmak still use rather outdated speaker drivers and cabinet of Yamaha NS-1000?

  • It looks that dualazmak "intentionally" uses several "integrated amplifiers" in his multichannel active setup; what are the background and benefits?

  • Can you summarize the background and reasons for Mr. dualazmak's selection of Japan domestic manufacturers' integrated amplifiers?

  • Are the availability of "domestic" reliable speedy maintenance and repair services also key factors for his selection?

  • It looks that dualazmak added subwoofer Yamaha YST-SW1000 and super-tweeter Fostex T925A; why and what are the benefits?

  • Can you summarize pros (and cons) of the large heavy subwoofer YAMAHA YST-SW1000?

  • Can you summarize the merits and benefits for Mr. dualazmak, and for other people, the use of EKIO as system-wide DSP center?

  • It looks that his latest system utilizes VB Audio MATRIX as system-wide ASIO-VASIO-VAIO routing center; what are the benefits?

  • Can you summarize Mr. dualazmak's approach and practices for 1 msec to 0.1 msec time-alignment among the speaker drivers?

  • Can you briefly describe the "three methods" dualazmak applied for such time-alignment tuning?

  • Can you summarize the background and reasons for dualazmak's unique physical "positioning" of super-tweeter Fostex T925A "under" the main SP cabinet, not on top of the cabinet?

  • Can you summarize dualazmak's unique configuration and benefits of Fostex T925A with reverse reflex dispersion using hard heavy random-surface crystal glass?

  • For the frequency response measurements in digital domain, analog line-level domain, SP high-level domain, as well as in room air sound at listening position, dualazmak uses "FFT analysis of accumulated/recorded well QC-ed white noise", instead of rapid sine-sweep; what are the scientific background and merits?

  • For reliable and reproducible FFT analysis we need rich raw acoustic data; does his "Cumulative Recorded White-Noise Averaging Method" also fit in this regard?

  • Has dualazmak objectively measured/confirmed frequency responses in various stages in the signal path, i.e. in digital domain, analog line level, SP high level after amplifier as well as room air sound at listening position?

  • Did dualazmak measure frequency responses in SP-high-level before and after the protection capacitors?

  • Why dualazmak uses the 22-ohm "parallel" (not in series) resistors in the SP high-level lines for mid-range drivers, tweeters and super-tweeters?

  • Can you summarize dualazmak’s “protective measures”, in digital domain and in analog domain, for his treasure speaker drivers, i.e., midrange driver, tweeter and super-tweeter?

  • Can you summarize the background and reasons for dualazmak's setting of -48 dB/Oct steep low-pass (high-cut) digital filter at 25 kHz for midrange driver, tweeter and super-tweeter?

  • Can you summarize the background and reasons for dualazmak's usual digital playback of all the music tracks in 88.2 or 96.0kHz 24bit by applying on-the-fly format conversion of JRiver MC?

  • What would be dualazmak’s “countermeasures” to eliminate/suppress undesirable low frequency air-conditioning noises of recording venue in digital tracks, in case if it is audible uncomfortable level?

  • If needed, can dualazmak directly “edit” the digital music track to delete or suppress the undesirable air-conditional portion(s)? If “yes”, what kind of audio editing software does he use?

  • For multichannel multi-amplifier active audio setup, we need multiple rather long speaker cables with proper and exact "identifications such as color coding" on them; what are dualazmak's actual practices in this regard?

  • Why does dualazmak carefully eliminate magnetic-susceptible metals, magnetizable metals, in speaker wiring chains?

  • Can you describe the background and actual case reports regarding his elimination of magnetizable metals in speaker wiring chains in detail?

  • Can you find any actual example case of such “signal distortion” due to magnetizable metal(s) in amplifiers within ASR Forum?

  • Can you find "the typical example case of issue with ferrous/magnetizable metals" associated with early version of BUCKEYE 3 Channel Purifi Amplifier?

  • It looks that dualazmak performs periodic cleaning of all the metal-to-metal connectors and contact points; can you summarize his practices of such intensive cleaning?

  • I remember dualazmak discussed that "we should avoid using CAIG DeoxIT, but it is better to use isopropyl alcohol (IPA) only"; what is the background in this regard?

  • A discussion in his thread shared that DeoxIT contains some kind of unsaturated fatty acid which would be polymerized into undesirable resin during long while; right?

  • It looks that Mr. dualazmak prefers “crimping” instead of “soldering”; can you find and summarize the reasons?

  • Why does dualazmak insists “we should never solder the crimped terminals”?

  • What are the background and rationales for dualazmak’s use of heat-shrink tubing covers on crimped terminals?

  • Are there any other discussions on "crimping is better than soldering" within ASR Forum?

  • Mr. dualazmak implemented/integrated his own DIY-built 12-VU-Meter Array in his multichannel system; can you summarize the background and actual utilization as well as benefits of such VU-Meter Array?

  • Mr. dualazmak also implemented "real time listening configuration" to vinyl LP disks in his multichannel audio setup; can you summarize the background, implementation, as well as his objectives in this regard?

  • How did Mr. dualazmak effectively suppress EMFI (electromagnetic field interference) noises coming into his relatively affordable phono preamplifier, Audio-Technica AT-PEQ30, especially while using MC (moving coil) cartridge?

  • How does Mr. dualazmak "organize" his digital music library?

  • Can you summarize dualazmak's policy and practices on "safe back-up" of his digital music library?

  • I once read that he backups his digital library at remote sites; his son's home and his daughter's home,,, right?

  • I once read Mr. dualazmak recommended taking high-resolution photos of the racks/shelves of physical disks; can you summarize the background and reasons?

  • Is his “taking photos” also his one of the preparatory measures against possible physical collapse of the shelves by earthquakes and/or by tornado?

  • I know Mr. dualazmak has prepared his own "Audio Reference/Sampler Music Playlist" and he has been utilizing the Playlist throughout his multichannel audio project. Can you summarize the concept and actual configuration of his Playlist?

  • How would Mr. Dualazmak's “room acoustic environments” contribute to "as if disappearance" of speakers in his audio setup?

  • How did Mr. dualazmak validate/confirm frequency response calibration-profile of his rather affordable BEHRINGER ECM8000 measurement microphone?

  • Did he cross-calibrate his ECM8000 with precisely calibrated professional measurement microphone?

  • What was/is actual deviation range from "flat", in dB, of his specific ECM8000 in 20 Hz to 20 kHz frequency coverage?

  • Can you summarize the dualazmak's "policy" and practice for sporadic (not frequently) integration of a TV unit into his multichannel audio setup?

  • Can you summarize the benefits of the USB wired volume controllers which dualazmak uses for master volume control within upstream JRiver MC?

  • Why does dualazmak not use SOTA audio racks or similar cabinets?

  • How much is the average electricity consumption by dualazmak's whole multichannel audio setup?

  • Why has Mr. dualazmak been sharing so many illustrations, diagrams and photos in sharing his multichannel audio project?

  • What would be the general acceptance of Mr. dualazmak's multichannel audio system within ASR Forum, and in other audio(-visual) communities?
 
Last edited:
Chr1 said:
Not sure why the other thread is still closed.

Perhaps we should have both a "Positive aspects of AI" and a "Negative aspects of AI", thread?

Anyhoo, back to my main concern.
A tad worrying, I would suggest.
As per: YMMV.

www.bbc.co.uk

Anthropic researcher believes more than 10% chance AI 'could kill all humans'

It is the latest in a series of increasing warnings about the safety threat posed by artificial intelligence.
www.bbc.co.uk www.bbc.co.uk
Click to expand...
Please keep this thread on topic and don't use it for posts which belong in the closed thread otherwise we will lose this one too.

this thread is for USEFUL APPLICATIONS OF AI

Doom mongering is not that.
 
Yesterday I asked to Gemini for an electronic music streamer, free, no ads and gave me a list of 5 or 6. Thanks to it I found hearthis.at as a web and as an app. Recently installed app and, hurra!, myriads of channels non interrumpible music to listen, again, no ads including the non paid version. Really amazing. Some yesrs ago I was fan of di.fm, but actually is thrash.

This evening, in the house channel, a guy mixing live with vynils. Wow!

Give them a try.
 
I revised my DIY speakers with HypexFusionAmps with AI.

I gave it all the measurements I made with REW sometime last year where I did the full spinorama in a gym that was big enough so I could gate with 9ms.

I exported those for every single driver without any crossover as frequency responses. On top I exported distortion of the 0 degree measurement for every driver.

I told AI to come up with the best crossover frequency based on directivity and distortion and it did that. I asked for all other filters needed for best response and it directly coded those into the config.xml that Hypex Fusion designer consumes after I gave my old as an example.
I said, that I want to look at it like on the spinorama website and it turned up a webserver with exactly that. I wanted to toggle each filter and live review the changes on that website, also done. It kind of replaced VituixCad. I like VituixCad so I wanted to look at the filters in there. As the VituixCad project files are XML AI generated me a VituixCad project file and I could just open it.

Most important this is the best tuning I ever got for my speakers.
 
Another one. Throw some emails at it, a little direction, get a detailed, formal letter of complaint.

So given your understanding of how I give each orgnaisation a unique email address, and from the attached emails, what can you tell me about my relationship with clifton cameras.
Click to expand...
Now look at these obvious scam emails I've received in recent days, having had nothing sent to that address, since 2016 when I ordered some binoculars from them.

What would you then conclude?
Click to expand...
I want to email Clifton Cameras. Inform them of what has happened, and point out their responsibilty. (regardless of how the leak happened it is either from them or thier supply chain - their responsibility). It seems like a recent leak/hack since I am now getting multiple scam emails in a few days after years of nothing.

I want to inform them of what steps we think they should be taking (identify leak, and close the security gap, inform/warn customers who now will be receiving scam emails like I am.

Tell them that I expect a response letting me know they are treating it seriously, and that if I don't get such a response, I will be rasing the issue with the ICO.

Please draft the email.
Click to expand...

Then the generated email.

Dear Clifton Cameras Data Protection Team,

I am writing to report an apparent exposure of personal data that I supplied exclusively to Clifton Cameras, and to ask that you treat this as a formal data-protection complaint and potential security incident.

I use a different email address for each organisation. The address XXXXXXXXXXXX was created specifically for Clifton Cameras and has not been supplied to or used with any other organisation. I used it when I registered an account and purchased a Canon camera from you in 2015, and again when I ordered binoculars in 2016.

After receiving nothing at this address for approximately ten years, I have now received four obvious credential-phishing messages within five days, between 7 and 11 September 2026. The messages use the precise Clifton-specific address in their recipient fields, subjects, bodies and malicious links. Copies of all four messages are attached in their original .eml format so that the full headers can be examined.

The sudden concentration of messages after years of inactivity raises a serious concern that this address has recently been exposed, or that previously obtained Clifton customer data has recently been acquired or activated by criminals. Either possibility warrants investigation.

Because this address was entrusted solely to Clifton Cameras, its presence in a criminal phishing list points to Clifton Cameras or an organisation within its supply chain. Your historical correspondence indicates that customer information was processed through several systems or service providers, including ecommerce, hosting, IT and email-marketing providers. If the disclosure occurred through a processor or supplier, that does not remove the need for Clifton Cameras to investigate and address the security of the customer data entrusted to it.

I expect Clifton Cameras to:

- record and investigate this as a possible personal-data breach;
- examine relevant current and historical systems, suppliers, data migrations, exports and backups;
- establish when and how the address was exposed, what other personal information may be involved, and how many customers may be affected;
- contain the incident and close any continuing security gap;
- preserve relevant logs and other evidence;
- assess its notification obligations under UK data-protection law, including whether the incident must be reported to the ICO; and
- warn affected customers promptly where appropriate, particularly because the exposed information is now being used in an active phishing campaign.

Even if your investigation is not yet complete, customers potentially receiving these messages may need an immediate warning so that they do not surrender email passwords or other credentials.

Please acknowledge this complaint, provide a case reference, and confirm that it has been passed to the person responsible for data protection and information security. I also expect a substantive response explaining what has been investigated, whether an incident has been identified, what remedial action is being taken, and whether affected customers and the ICO will be notified. I do not require disclosure of information that would compromise your security, but I do expect a clear account of how this matter is being handled.

If I do not receive a response demonstrating that Clifton Cameras is treating this seriously, I will raise the matter with the Information Commissioner’s Office and provide the ICO with this correspondence and the original phishing messages.

Yours sincerely,

Anthony Collins
Click to expand...
 
As critical as I am of AI, I have to say that I'm pleasantly surprised by my experiment. I conducted a series of measurements of the room acoustics at and around my listening position using REW. I then exported the various measurements and had AI interpret them. The recommendations are really good; no complaints so far.
 
maxmax said:
As critical as I am of AI, I have to say that I'm pleasantly surprised by my experiment. I conducted a series of measurements of the room acoustics at and around my listening position using REW. I then exported the various measurements and had AI interpret them. The recommendations are really good; no complaints so far.
Click to expand...
I must try that as I've just started with REW and there's so much info in there.
 
AudioSceptic said:
I must try that as I've just started with REW and there's so much info in there.
Click to expand...
Good idea to have AI help you with the tons of features of REW.

Be aware that AI sometimes tells funny stories. At one point in the interpretation of my REW measurements, AI wrote that it makes no difference if you toe in speakers. I then insisted and ask for further explanations about to toe in or not. At some point I uploaded new measurements and mentioned that I slightly repositioned the speakers to get a better sound stage. AI then responded that in fact the correct position of the speakers is more important than some PEQ settings :) Conclusion: better to have some insights in a topic to get the most out of AI.
 
An A.I. model trained on board games played in Roman times suggests a limestone artifact was used as the surface of a board game.

The board game is on-line and can be played against a bot.

gift link

www.nytimes.com

Archaeologists Decode Rules of an Ancient Roman Board Game

They scanned the pattern of microscopic imperfections on an ancient limestone slab, and used A.I. to figure out how a game might have been played.
www.nytimes.com www.nytimes.com
 
So here is the chat gpt summary of what I have just asked it for, and what it did for me. First quote box is my opening prompts, second is the ai generated summary of what happened.

Me : Can you see the linked webpage. It is a list of all my recent posts on the audio science review forum.

[https://www.audiosciencereview.com/...rch/59493800/\&c\[users]=antcollinet\&o=date)


Chat GPT : of course bro (Paraphrased)



Me : The forum is xenforo. I want to get a download of - lets say - the last couple of months of my posts into a text file. I know zenforo doesn't have a feature for that. I am thinking of creating a safari extension that will open each of the posts from that page, and download the content of my post only to a single text file. (one file for all posts)

Or do you have a proposal for a simpler way?
Click to expand...
15 minutes later - after first screenshotting an error message followed by ChatGPT providing a fixed version of the script, I had a markedown format file of all my last two months posts.

Of that 15 minutes, my time was maybe 3 or 4 of prompting. The rest was a short time for AI to create and fix the script followed by some minutes of rate-limited downloading.

It has used in total - including summary generation - about 7 percent of my five hour codex usage, and 1% of my weekly usage. (Chat GPT Plus Plan)


Here is ChatGPT's summary of what we did.

I wanted a Markdown archive of my recent ASR posts, but XenForo doesn’t provide a suitable export feature.

I asked Codex for advice and initially considered creating a Safari extension. It suggested a much simpler solution: a JavaScript exporter that runs directly from Safari’s Web Inspector console while viewing XenForo’s “all posts by user” search results.

Codex inspected ASR’s current XenForo page structure and wrote the complete script. It:

  • Works through the paginated search results until reaching a configurable cutoff date
  • Opens each post in the background using my existing forum session
  • Extracts only my post rather than the whole thread
  • Produces one human-readable Markdown file
  • Preserves dates, thread titles, permanent links, quotations, lists, formatting, images and code blocks
  • Removes duplicates, reports progress and continues past individual failures
  • Deliberately spaces out requests to avoid hammering the forum
The first run exposed a small difference in ASR’s URL format—index.php?search/... rather than /search/.... I supplied a screenshot of the error, Codex identified the faulty check, updated the script, and the next run worked.

A nice example of AI not merely answering a question, but helping select the simplest design, implementing it, diagnosing the real-world failure and delivering a working tool.
Click to expand...
 
I am currently experimenting with how far AI can be used in a practical way for the development of an active loudspeaker.

For a start, I deliberately chose a relatively simple and inexpensive project that should nevertheless offer good technical and acoustic performance. The plan is to build two active 2-way loudspeakers, each using one HiVi M8a and one FaitalPRO 4FE35. The enclosure will be around 28 litres bass reflex, and the crossover will probably be somewhere around 400 Hz. Amplification and DSP will be handled by a JAB5 with an ADAU1701.

Of course I am interested in the loudspeaker itself, but at least as much in the development process. My goal is to connect REW, SigmaStudio and ChatGPT within the Work environment on the PC and turn them, as far as possible, into an automated development workflow. I will initially perform the necessary measurements in REW and provide the measurement data to ChatGPT. The AI should then determine as independently as possible how the two drivers can best be combined.

This first involves analysing the individual measurements of the woofer and the mid/high-frequency driver. ChatGPT should then develop an appropriate crossover while taking into account frequency response, level, phase and acoustic summation through the crossover region. Where necessary, it should derive changes in level, polarity or delay from the measurements. The AI should also decide which additional EQ filters are useful. As initial target curves I intend to use a flat response and a Harman-type target.

The crucial part for me, however, is automation. In the longer term, I do not want to transfer every measurement manually between programs and then determine individual DSP parameters by hand. Instead, ChatGPT should take over or control as many of these steps as possible within the Work environment. This requires suitable interfaces between REW, the analysis process, SigmaStudio and DSP programming. Depending on what proves possible, this could involve measurement files, export formats, scripts, APIs or direct interaction with the applications themselves. Determining which of these interfaces can actually be automated reliably is itself part of the experiment.

The next step is to implement the decisions derived from the measurements as directly as possible in SigmaStudio. ChatGPT should therefore not merely explain that, for example, a filter, delay or level adjustment is required, but should determine as independently as possible how the ADAU1701 signal flow should be constructed and parameterised. The system would then measure again with REW, analyse the result automatically or largely automatically, modify the DSP configuration and perform another measurement. The aim is to create a closed iterative process consisting of measurement, analysis, optimisation and verification.

My own role should increasingly be limited to those tasks that cannot easily be automated physically: building the loudspeaker, positioning the measurement microphone, rotating or repositioning the loudspeaker where required, and connecting the hardware. I want to interfere as little as possible with the interpretation of the measurements and the decisions derived from them.

If this first, relatively simple 2-way loudspeaker works as intended, I would like to expand the system step by step. The longer-term goal would be to bring the overall process, as far as reasonably possible with commonly available hardware, closer to the methodology of a professional loudspeaker measurement system. The Klippel NFS is obviously not a realistic hardware comparison here, but rather a methodological reference.

One conceivable extension would be to perform systematic measurements at a large number of defined microphone or loudspeaker positions, acquire these measurements automatically and analyse them together. This would gradually shift the focus from a single frequency-response curve toward spatial radiation behaviour. From a sufficiently large set of repeatable measurements, it should become possible to derive horizontal and vertical off-axis data, directivity information and corresponding visualisations. These data could then in turn be included in the automated DSP optimisation process.

The real question for me therefore goes beyond “Can AI develop a loudspeaker?” What I find more interesting is how far relatively simple components — REW, SigmaStudio, a DSP, a calibrated microphone and ChatGPT in the Work environment — can be combined into a largely automated measurement and development system.

The ultimate goal would be a loop in which the AI defines a measurement strategy > processes the measurement data > derives changes to the DSP > implements them, and then uses new measurements to verify whether those changes actually produced an improvement.

If this works, I think it could become a genuinely useful application of AI (not only) in DIY audio. REW and SigmaStudio are already very powerful tools. The real difficulty for me lies between them > especially generating measurements systematically, interpreting them correctly and translating them into the appropriate DSP actions.
That is exactly the gap I would like to automate as far as possible.
 
dualazmak said:
If you would be interested, you will be also much welcome
Click to expand...
Thank you very much for your kind words and for the invitation.

I have actually followed your thread many times already. To me, it is a prime example of someone who is not only deeply interested in HiFi, but really treats the hobby as a passion while still approaching it with remarkable care, precision and attention to detail.

I have always appreciated the combination of enthusiasm and technical thoroughness in your posts.

If my AI-assisted project develops far enough to become worth documenting in that way, I would be happy to contribute diagrams, photos and the signal path to your thread.
 
You must log in or register to reply here.

Similar threads

napfkuchen
AI for art generation and image upscaling
Replies
3
Views
455
amirm
amirm
Brian Hall
AI Videos we have generated
Replies
19
Views
2K
Brian Hall
Brian Hall
Matias
Things you bought based on AI advice (good or bad!)
2
Replies
36
Views
3K
bravomail
B
J
Ai Signal Flow — a web app I built
Replies
10
Views
774
Juiced
J
Kain
12.2.8 immersive setup possible in a 12 feet long x 11 feet wide x 9.5 feet high room?
2
Replies
37
Views
2K
napfkuchen
napfkuchen
Back
Top Bottom